{
  "schema_version": "pct-article-edition-1.0",
  "publication_date": "2026-11-08",
  "export_scope": "articles_only",
  "source_edition_id": null,
  "rights": "PCT reserves rights in original editorial content; third-party rights retained. No general source-data reuse license.",
  "articles": [
    {
      "id": "pct-2026-11-08-company-runpod",
      "content_type": "article",
      "desk": "company",
      "title": "Runpod’s next test: earning its share of the GPU bill",
      "publication_date": "2026-11-08",
      "observation_date": "2026-11-08",
      "editorial_voice": "Company Desk | AI-assisted, measured business columnist",
      "body": "A property photograph is a useful place to start understanding Runpod. The image has to be processed, returned and paid for. The customer does not gain much from knowing that a GPU was available somewhere if the finished picture arrives too late for the listing. That distance between renting a chip and completing a useful job is the business Runpod is trying to occupy.\n\nNodalview illustrates the opportunity. In a September 9 case study published by Runpod, the property-media company reported processing almost 1.5 million photos a month. Up to 70% arrived within one afternoon hour. It moved the GPU layer to Runpod but retained AWS for its web platform, storage, authentication and Kubernetes cluster. Production inference uses Serverless; development and training use Pods. [Nodalview case study](https://www.runpod.io/case-studies/how-nodalview-chose-what-to-move).\n\nThis is a more plausible sales proposition than replacing an entire cloud estate. A specialist can win a difficult, expensive slice of an application while leaving the surrounding machinery alone. But that narrow entry point creates a demanding test. Runpod has to make the chosen slice cheaper to operate after migration, support and interruptions, while retaining enough of the bill to build a durable company.\n\n## A larger business, with an incomplete financial picture\n\nThe company is already beyond the scale suggested by its developer-tool reputation. On October 13, Runpod said it had exceeded $325 million in annual recurring revenue, compared with $120 million in January. It also reported more than 1.2 million developers and appointed Brian Erickson, whose previous finance roles included DigitalOcean, as chief financial officer. These are company disclosures, not audited results presented with the announcement. [October announcement](https://www.runpod.io/press/runpod-325m-arr-new-cfo).\n\nThe reported ARR figure is roughly 2.7 times January's. That establishes the size of the claimed increase, not the quality of every dollar. A consumption-based infrastructure business needs a reconciliation between its annualized measure, recognized revenue, supplier payments and cash generation before readers can confidently compare it with a subscription software company. The release does not supply that bridge, a gross-margin figure or a customer-concentration table.\n\nIn June, Runpod announced a $100 million investment led by Summit Partners at a stated $1 billion valuation. It described a platform spanning experimentation, training, fine-tuning, inference and multi-node work. Summit's Michael Medici framed the investment around serving developers throughout that process, rather than one isolated stage. The investor has a financial interest in that argument. [Financing announcement](https://www.runpod.io/press/runpod-raises-100m-summit-partners); [Summit investment thesis](https://www.summitpartners.com/news/backing-the-ai-builders-our-investment-in-runpod).\n\nThe developer total needs similar care. A registered developer, an active workload and a paying corporate customer are different units. The announcement does not reconcile them. A platform can have a broad community while a small number of customers generate a large share of its spending. The next useful adoption figure would connect participation with repeat paid use, rather than leave readers to treat every developer as an equivalent buyer.\n\nThe expansion logic is straightforward. A developer who uses the service to test a model may later bring a production application, a larger team and a recurring procurement budget. Each transition could increase spending without requiring the company to find an entirely new buyer. The risk is equally concrete: a platform broad enough to serve all those stages must support more failure modes, purchasing requirements and operating habits than a simple GPU rental page.\n\n## What one supplier reveals about the money\n\nThe most illuminating document is not Runpod's own pitch. In its July 14 prospectus, GPU supplier QumulusAI described Runpod as its largest GPU-as-a-service channel partner. The disclosed arrangement allocated 80% of the transaction value to QumulusAI and 20% to Runpod, with no minimum purchase requirement. Runpod handled customer contracting, pricing, onboarding, billing and collections. Runpod contributed 71.9% of QumulusAI's GPU-as-a-service revenue in the first quarter of 2026. [QumulusAI prospectus](https://www.sec.gov/Archives/edgar/data/2084026/000143774926023622/quma20260714_424b4.htm).\n\nThat is a concrete division of labor. A supplier brings usable computing capacity; Runpod brings customers and the machinery for selling access to them. It also helps explain why developer distribution matters. If a supplier can reach a populated buying platform rather than recruit every customer itself, the platform has something to sell upstream as well as downstream. QumulusAI's concentration disclosure shows the importance of that channel to one supplier. It does not establish Runpod's dependence on that supplier.\n\nOn the disclosed split, a $100 transaction leaves $20 with Runpod before Runpod's own costs. That arithmetic is not a company-wide gross-margin estimate. The filing describes one relationship, and it does not resolve how Runpod recognizes revenue across different products or supplier contracts. Applying that percentage to the $325 million ARR headline would create a confident-looking answer to a question the evidence does not settle.\n\nCEO Zhen Lu calls Runpod a software company in a strategy essay updated September 13. The company also openly recruits data centers, cloud providers and GPU operators as supply partners. Together, those statements make its intended position clear: a customer-facing software layer over contributed computing capacity. They do not establish that every machine uses the same commercial terms. [Lu’s strategy essay](https://www.runpod.io/blog/the-chips-got-faster-the-stack-didnt); [Partner program](https://www.runpod.io/partners).\n\nAggregating demand can make a supplier’s machines easier to sell, but only when the available hardware suits the work. An idle GPU with insufficient memory, in the wrong location or without access to the required data is not interchangeable with the machine a customer needs. Runpod’s matching and deployment layer therefore has to do more than display a large inventory. It must turn a particular customer request into a working allocation.\n\nThe advantage of this position is that growth need not require Runpod to finance every underlying facility itself. The trade-off is limited direct control over parts of the service on which customers depend. Supplier economics and customer promises still have to meet. A contract can allocate the cost of a failure; it cannot make an unavailable machine finish an image.\n\n## The price of a completed photograph\n\nFor a roughly 20GB model, Nodalview reported first inference in four to five minutes on Runpod versus six to seven on AWS and Google Cloud, including container retrieval and model loading. Processing was under a minute per image once workers ran. These are different clocks, reported in a vendor case study rather than an independently reproduced test. [Test scope and workload](https://www.runpod.io/case-studies/how-nodalview-chose-what-to-move).\n\nA business buyer should keep both clocks. Steady-state processing measures how quickly an available worker completes work. Time to first useful output captures the wait before that worker can help. If demand arrives together, a good average processing time can coexist with an unacceptable queue. Conversely, keeping workers ready can reduce that wait while consuming some of the savings that made the specialist attractive.\n\nRunpod's documentation makes the purchasing choice explicit. Flex workers can scale to zero; active workers remain running. Billing follows worker runtime, rounded up to the second, and includes startup, execution and the waiting period before shutdown. Storage is additional. The documentation distinguishes loading a model into GPU memory from container retrieval and cached-model downloads, which normally happen before billed worker runtime. [Serverless pricing documentation, modified October 15](https://docs.runpod.io/serverless/pricing).\n\nThere is another important distinction between paying for resources and paying for an outcome. Runpod's October 16 billing explanation says an unsuccessful job can still incur charges while its worker runs. Repeated initialization failures, hung handlers and retries can therefore consume a balance without delivering successful requests. That is a disclosed billing model, not evidence that such charges are necessarily incorrect. [Failed-job billing explanation](https://www.runpod.io/blog/failed-requests-real-charges-how-runpod-serverless-billing-works-when-workers-fail).\n\nFor a photo-processing company, the useful denominator is consequently accepted images delivered within the promised time. The numerator should include compute, storage, retries, standby capacity and the engineering needed to operate the pipeline. A lower hourly GPU price can help, but it is only one input. A pilot should replay the actual afternoon arrival pattern, include a new-worker start, and record failed attempts as well as successes. Those measurements turn a cloud comparison into a product-cost comparison.\n\n## The contract has to catch up with the developer\n\nRunpod's September 24 enterprise announcement addresses the next purchasing obstacle. The company introduced organizational controls, SSO/SAML and cost-center billing, alongside a claimed ISO/IEC 27001 certification. It described reserved capacity, contractual service-level agreements and named technical account managers. Those features matter when an engineer's useful service becomes something a company must govern and keep available. Certification and feature availability still require checking against the customer's specific deployment and contract. [Enterprise announcement](https://www.runpod.io/press/runpod-expands-enterprise-platform).\n\nFor procurement, the questions are ordinary but consequential. Which GPU and region are reserved? How much capacity is actually protected? Who answers an urgent support request? What happens to queued work during an interruption, and which costs remain payable? A service-level agreement's financial remedy may be much smaller than the customer's lost business. The buyer needs the operational answer as well as the credit schedule. For Runpod, a capacity promise can also change the risk of the business. A supplier arrangement without purchase minimums says little about the cost of reserving a reliable substitute for an enterprise customer. The financial attraction of flexible supply has to be evaluated alongside the commitments made on the other side of the platform.\n\nA recent incident makes those questions timely without justifying a verdict on the whole platform. Runpod's official status log opened an incident affecting CA-MTL-1 and EU-SE-1 on October 31 at 20:30 UTC. It reported EU-SE-1 restored on November 3 and closed the remaining CA-MTL-1 issue on November 6 at 04:22 UTC. The final updates do not give a customer count or a completed root-cause account. [Official incident timeline](https://uptime.runpod.io/incident/1109748?mp=true).\n\nThe incident window should not be described as continuous failure for every user in both regions. The updates distinguish their restoration times and later refer to remaining machines in Montreal. Nor does one incident prove a comparative reliability ranking. It does give prospective customers a concrete event to discuss: which workloads were affected, what recovery required, and whether an equivalent interruption would be tolerable in their own application. A useful recovery exercise would move a representative job, confirm that its model and input data are available, prevent duplicate processing, and return the result through the normal application. Merely showing that another GPU starts does not test the whole customer experience.\n\n## The next useful numbers\n\nRunpod's opportunity is to collect a share of spending that developers would otherwise place through a different cloud interface, and to earn that share by reducing the work around the GPU. Its danger is that the software layer becomes responsible for a production promise while supplier costs, hardware availability and customer support consume the apparent advantage. Neither a large signup count nor a low catalog price settles that contest.\n\nFor the company, the revealing next disclosures would be retained revenue after supplier payments, customer concentration, renewal behavior and the cost of meeting enterprise commitments. For buyers, the useful evidence is smaller and more immediate: an agreed workload, a measured peak, a complete bill and a tested recovery path. Nodalview's selective migration offers a sensible way to ask the question. Runpod need not own the rest of the application to earn an important place in it. It must make the part it runs worth depending on.",
      "sources": [
        {
          "id": "S1",
          "title": "How Nodalview chose what to move",
          "publisher": "Runpod",
          "url": "https://www.runpod.io/case-studies/how-nodalview-chose-what-to-move",
          "published_at": "2026-09-09",
          "updated_at": null,
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Vendor-published customer case study; named customer and engineer, not an independent benchmark.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": "2026-09-09",
          "observation_date": "2026-11-08"
        },
        {
          "id": "S2",
          "title": "Runpod Surpasses $325 Million in ARR, Appoints Brian Erickson as Chief Financial Officer",
          "publisher": "Runpod",
          "url": "https://www.runpod.io/press/runpod-325m-arr-new-cfo",
          "published_at": "2026-10-13",
          "updated_at": null,
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Company press release; financial and adoption figures are self-reported.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": "2026-10-13",
          "observation_date": "2026-11-08"
        },
        {
          "id": "S3",
          "title": "Runpod Raises $100M Led by Summit Partners",
          "publisher": "Runpod",
          "url": "https://www.runpod.io/press/runpod-raises-100m-summit-partners",
          "published_at": "2026-06-24",
          "updated_at": null,
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Company financing announcement; valuation basis is not specified as pre- or post-money.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": "2026-06-24",
          "observation_date": "2026-11-08"
        },
        {
          "id": "S4",
          "title": "The Chips Got Faster. The Stack Didn’t.",
          "publisher": "Runpod; Zhen Lu",
          "url": "https://www.runpod.io/blog/the-chips-got-faster-the-stack-didnt",
          "published_at": null,
          "updated_at": "2026-09-13",
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "CEO strategy essay; updated date is not necessarily original publication date.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "S5",
          "title": "QumulusAI Form 424B4 prospectus",
          "publisher": "QumulusAI / U.S. SEC EDGAR",
          "url": "https://www.sec.gov/Archives/edgar/data/2084026/000143774926023622/quma20260714_424b4.htm",
          "published_at": "2026-07-14",
          "updated_at": null,
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Supplier securities filing; Business, Our Customers and revenue-recognition disclosures. SEC hosting does not mean SEC verification or endorsement.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": "2026-07-14",
          "observation_date": "2026-11-08"
        },
        {
          "id": "S6",
          "title": "Serverless pricing",
          "publisher": "Runpod documentation",
          "url": "https://docs.runpod.io/serverless/pricing",
          "published_at": null,
          "updated_at": "2026-10-15",
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Live product documentation; last-modified date shown on page.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "S7",
          "title": "How Runpod Serverless billing works when a job fails to complete",
          "publisher": "Runpod",
          "url": "https://www.runpod.io/blog/failed-requests-real-charges-how-runpod-serverless-billing-works-when-workers-fail",
          "published_at": null,
          "updated_at": "2026-10-16",
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Vendor billing explanation; distinguishes worker runtime from job outcome.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "S8",
          "title": "Runpod Expands Enterprise Platform for Mission-Critical AI Workloads",
          "publisher": "Runpod",
          "url": "https://www.runpod.io/press/runpod-expands-enterprise-platform",
          "published_at": "2026-09-24",
          "updated_at": null,
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Company release; announced features and contractual availability, not independent certification-scope verification.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": "2026-09-24",
          "observation_date": "2026-11-08"
        },
        {
          "id": "S9",
          "title": "CA-MTL-1 and EU-SE-1 DC is Unavailable",
          "publisher": "Runpod status",
          "url": "https://uptime.runpod.io/incident/1109748?mp=true",
          "published_at": "2026-10-31",
          "updated_at": "2026-11-06",
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Official incident timeline; does not identify affected customer count or give a final root cause.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": "2026-10-31",
          "observation_date": "2026-11-08"
        },
        {
          "id": "S10",
          "title": "Runpod Partner Program",
          "publisher": "Runpod",
          "url": "https://www.runpod.io/partners",
          "published_at": null,
          "updated_at": null,
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Live partner-program page; contributor categories, not a complete supplier inventory.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "S11",
          "title": "Backing the AI Builders: Our Investment in Runpod",
          "publisher": "Summit Partners; Michael Medici",
          "url": "https://www.summitpartners.com/news/backing-the-ai-builders-our-investment-in-runpod",
          "published_at": "2026-06-24",
          "updated_at": null,
          "retrieved_at": "2026-11-08",
          "evidence_type": "primary",
          "verification": "Investor thesis with a financial interest; not independent validation of platform economics.",
          "license_status": "Source rights retained by original publisher; public availability is not a republication license. Facts paraphrased; no source imagery reproduced.",
          "publication_date": "2026-06-24",
          "observation_date": "2026-11-08"
        }
      ],
      "facts": [
        {
          "id": "F1",
          "source_ids": [
            "S1"
          ],
          "fields": {
            "customer": "Nodalview",
            "monthly_photos_approx": 1500000,
            "peak_hour_share_up_to": 0.7,
            "engineering_team": 12,
            "retained_cloud": "AWS"
          },
          "qualifier": "Customer report carried by vendor.",
          "claim": "Structured observed values are recorded in fields; see the cited source and linked narrative paragraph."
        },
        {
          "id": "F2",
          "source_ids": [
            "S2"
          ],
          "fields": {
            "reported_arr_usd_october": 325000000,
            "reported_arr_usd_january": 120000000,
            "developer_count_lower_bound": 1200000,
            "cfo": "Brian Erickson"
          },
          "qualifier": "Company-reported, unaudited announcement.",
          "claim": "Structured observed values are recorded in fields; see the cited source and linked narrative paragraph."
        },
        {
          "id": "F3",
          "source_ids": [
            "S3"
          ],
          "fields": {
            "investment_usd": 100000000,
            "stated_valuation_usd": 1000000000,
            "lead_investor": "Summit Partners"
          },
          "claim": "Structured observed values are recorded in fields; see the cited source and linked narrative paragraph."
        },
        {
          "id": "F4",
          "source_ids": [
            "S11"
          ],
          "claim": "Investor supports full-lifecycle platform thesis."
        },
        {
          "id": "F5",
          "source_ids": [
            "S5"
          ],
          "fields": {
            "supplier": "QumulusAI",
            "supplier_share": 0.8,
            "runpod_share": 0.2,
            "minimum_purchase_requirement": false,
            "supplier_gpu_revenue_share_via_runpod_q1_2026": 0.719
          },
          "qualifier": "One disclosed relationship, not universal terms.",
          "claim": "Structured observed values are recorded in fields; see the cited source and linked narrative paragraph."
        },
        {
          "id": "F6",
          "source_ids": [
            "S4",
            "S10"
          ],
          "claim": "Software-led positioning and third-party supply program."
        },
        {
          "id": "F7",
          "source_ids": [
            "S1"
          ],
          "fields": {
            "tested_model_gb_approx": 20,
            "runpod_first_inference_minutes": [
              4,
              5
            ],
            "comparison_minutes": [
              6,
              7
            ],
            "warm_processing_minutes_upper_bound": 1
          },
          "qualifier": "Customer test; startup and processing are separate.",
          "claim": "Structured observed values are recorded in fields; see the cited source and linked narrative paragraph."
        },
        {
          "id": "F8",
          "source_ids": [
            "S6"
          ],
          "claim": "Worker-runtime billing; storage additional; pre-runtime retrieval normally unbilled."
        },
        {
          "id": "F9",
          "source_ids": [
            "S7"
          ],
          "claim": "Unsuccessful jobs may consume billable runtime."
        },
        {
          "id": "F10",
          "source_ids": [
            "S8"
          ],
          "claim": "Enterprise controls, reserved capacity and contractual service levels announced."
        },
        {
          "id": "F11",
          "source_ids": [
            "S9"
          ],
          "fields": {
            "opened_at": "2026-10-31T20:30:00Z",
            "eu_se_1_restoration_reported_at": "2026-11-03T03:04:00Z",
            "closed_at": "2026-11-06T04:22:00Z"
          },
          "qualifier": "Timeline is not uniform customer outage duration.",
          "claim": "Structured observed values are recorded in fields; see the cited source and linked narrative paragraph."
        }
      ],
      "inferences": [
        {
          "id": "I1",
          "claim": "A specialist can enter an application through one expensive workload rather than whole-estate migration.",
          "based_on_fact_ids": [
            "F1"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I2",
          "claim": "The economic test includes operating effort and failure costs, not only GPU price.",
          "based_on_fact_ids": [
            "F1",
            "F8",
            "F9"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I3",
          "claim": "325/120 is about 2.71; annualized consumption and recognized revenue require separate analysis.",
          "based_on_fact_ids": [
            "F2"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I4",
          "claim": "Serving multiple lifecycle stages creates expansion opportunities and operating complexity.",
          "based_on_fact_ids": [
            "F3",
            "F4"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I5",
          "claim": "Supplier distribution has value; supplier dependence does not establish reverse dependence.",
          "based_on_fact_ids": [
            "F5"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I6",
          "claim": "The disclosed split yields $20 of $100 before Runpod costs, not a company-wide margin.",
          "based_on_fact_ids": [
            "F5"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I7",
          "claim": "Partner supply may limit capital requirements while leaving operational dependencies.",
          "based_on_fact_ids": [
            "F5",
            "F6"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I8",
          "claim": "Cold-worker latency and steady-state throughput measure different business constraints.",
          "based_on_fact_ids": [
            "F7",
            "F8"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I9",
          "claim": "Cost per timely accepted output is a more complete buyer denominator.",
          "based_on_fact_ids": [
            "F8",
            "F9"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I10",
          "claim": "Governance features help bridge individual adoption and organizational procurement.",
          "based_on_fact_ids": [
            "F10"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I11",
          "claim": "Contract remedies do not automatically replace lost operating value.",
          "based_on_fact_ids": [
            "F10"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I12",
          "claim": "The status event supports recovery diligence, not a platform-wide reliability ranking.",
          "based_on_fact_ids": [
            "F11"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I13",
          "claim": "Software value must exceed supplier, reliability and service costs.",
          "based_on_fact_ids": [
            "F5",
            "F10",
            "F11"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I15",
          "claim": "Developer totals do not establish paying-customer count or exposure to large accounts.",
          "based_on_fact_ids": [
            "F2"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I16",
          "claim": "Demand aggregation is useful only when compatible capacity can actually serve each workload.",
          "based_on_fact_ids": [
            "F5",
            "F6"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I17",
          "claim": "A realistic failover test must cover dependencies and output recovery, not merely launching a replacement worker.",
          "based_on_fact_ids": [
            "F8",
            "F11"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I18",
          "claim": "Reserved customer commitments can change a platform’s risk even when supplier purchase minimums are absent.",
          "based_on_fact_ids": [
            "F5",
            "F10"
          ],
          "status": "editorial_inference_not_reported_fact"
        },
        {
          "id": "I14",
          "claim": "Buyer pilots and more complete financial disclosures answer different parts of the business question.",
          "based_on_fact_ids": [
            "F1",
            "F2",
            "F5"
          ],
          "status": "editorial_inference_not_reported_fact"
        }
      ],
      "unknowns": [
        {
          "id": "U1",
          "text": "Runpod company-wide gross/net revenue presentation, ARR calculation methodology, audited recognized revenue, gross margins, cash generation and customer concentration were not established by the reviewed announcements."
        },
        {
          "id": "U2",
          "text": "The scope and economics of supplier contracts beyond QumulusAI are not disclosed in the reviewed material. Do not apply its 20% split to Runpod-wide ARR."
        },
        {
          "id": "U3",
          "text": "Affected customer count, continuous customer-specific outage duration, financial impact and final root cause of incident 1109748 remain unspecified in the reviewed updates."
        },
        {
          "id": "U4",
          "text": "Nodalview test methodology, hardware matching, full cost accounting and performance were not independently reproduced; no realized cost-saving percentage is asserted."
        },
        {
          "id": "U5",
          "text": "Enterprise contract pricing, certification scope and reserved capacity for any particular customer were not reviewed."
        }
      ],
      "correction_history": [],
      "license_status": "Original PCT editorial text; public license not selected. Source rights remain separate.",
      "public_url": null,
      "schema_version": "pct-research-item-1.0",
      "desk_extensions": {
        "schema_version": "pct.article.v1",
        "dek": "A property-photo workload and a supplier’s SEC filing show what the AI developer cloud sells, and where its business still needs to prove itself.",
        "status": "ready_for_editorial_review",
        "body_derivation": {
          "canonical_field": "sections",
          "method": "Join section headings and paragraph text in order; body is a derived rendering, not an independent editorial source."
        },
        "sections": [
          {
            "id": "pct-2026-11-08-company-runpod-opening",
            "heading": null,
            "paragraphs": [
              {
                "id": "pct-2026-11-08-company-runpod-opening-p01",
                "text": "A property photograph is a useful place to start understanding Runpod. The image has to be processed, returned and paid for. The customer does not gain much from knowing that a GPU was available somewhere if the finished picture arrives too late for the listing. That distance between renting a chip and completing a useful job is the business Runpod is trying to occupy.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I1"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-opening-p02",
                "text": "Nodalview illustrates the opportunity. In a September 9 case study published by Runpod, the property-media company reported processing almost 1.5 million photos a month. Up to 70% arrived within one afternoon hour. It moved the GPU layer to Runpod but retained AWS for its web platform, storage, authentication and Kubernetes cluster. Production inference uses Serverless; development and training use Pods. [Nodalview case study](https://www.runpod.io/case-studies/how-nodalview-chose-what-to-move).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S1"
                ],
                "claim_ids": [
                  "F1"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-opening-p03",
                "text": "This is a more plausible sales proposition than replacing an entire cloud estate. A specialist can win a difficult, expensive slice of an application while leaving the surrounding machinery alone. But that narrow entry point creates a demanding test. Runpod has to make the chosen slice cheaper to operate after migration, support and interruptions, while retaining enough of the bill to build a durable company.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I1",
                  "I2"
                ]
              }
            ]
          },
          {
            "id": "pct-2026-11-08-company-runpod-growth",
            "heading": "A larger business, with an incomplete financial picture",
            "paragraphs": [
              {
                "id": "pct-2026-11-08-company-runpod-growth-p01",
                "text": "The company is already beyond the scale suggested by its developer-tool reputation. On October 13, Runpod said it had exceeded $325 million in annual recurring revenue, compared with $120 million in January. It also reported more than 1.2 million developers and appointed Brian Erickson, whose previous finance roles included DigitalOcean, as chief financial officer. These are company disclosures, not audited results presented with the announcement. [October announcement](https://www.runpod.io/press/runpod-325m-arr-new-cfo).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S2"
                ],
                "claim_ids": [
                  "F2"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-growth-p02",
                "text": "The reported ARR figure is roughly 2.7 times January's. That establishes the size of the claimed increase, not the quality of every dollar. A consumption-based infrastructure business needs a reconciliation between its annualized measure, recognized revenue, supplier payments and cash generation before readers can confidently compare it with a subscription software company. The release does not supply that bridge, a gross-margin figure or a customer-concentration table.",
                "evidence_type": "calculation_and_analysis",
                "source_ids": [
                  "S2"
                ],
                "claim_ids": [
                  "I3",
                  "U1"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-growth-p03",
                "text": "In June, Runpod announced a $100 million investment led by Summit Partners at a stated $1 billion valuation. It described a platform spanning experimentation, training, fine-tuning, inference and multi-node work. Summit's Michael Medici framed the investment around serving developers throughout that process, rather than one isolated stage. The investor has a financial interest in that argument. [Financing announcement](https://www.runpod.io/press/runpod-raises-100m-summit-partners); [Summit investment thesis](https://www.summitpartners.com/news/backing-the-ai-builders-our-investment-in-runpod).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S3",
                  "S11"
                ],
                "claim_ids": [
                  "F3",
                  "F4"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-growth-p04",
                "text": "The developer total needs similar care. A registered developer, an active workload and a paying corporate customer are different units. The announcement does not reconcile them. A platform can have a broad community while a small number of customers generate a large share of its spending. The next useful adoption figure would connect participation with repeat paid use, rather than leave readers to treat every developer as an equivalent buyer.\n\nThe expansion logic is straightforward. A developer who uses the service to test a model may later bring a production application, a larger team and a recurring procurement budget. Each transition could increase spending without requiring the company to find an entirely new buyer. The risk is equally concrete: a platform broad enough to serve all those stages must support more failure modes, purchasing requirements and operating habits than a simple GPU rental page.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I4",
                  "I15"
                ]
              }
            ]
          },
          {
            "id": "pct-2026-11-08-company-runpod-supplier",
            "heading": "What one supplier reveals about the money",
            "paragraphs": [
              {
                "id": "pct-2026-11-08-company-runpod-supplier-p01",
                "text": "The most illuminating document is not Runpod's own pitch. In its July 14 prospectus, GPU supplier QumulusAI described Runpod as its largest GPU-as-a-service channel partner. The disclosed arrangement allocated 80% of the transaction value to QumulusAI and 20% to Runpod, with no minimum purchase requirement. Runpod handled customer contracting, pricing, onboarding, billing and collections. Runpod contributed 71.9% of QumulusAI's GPU-as-a-service revenue in the first quarter of 2026. [QumulusAI prospectus](https://www.sec.gov/Archives/edgar/data/2084026/000143774926023622/quma20260714_424b4.htm).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S5"
                ],
                "claim_ids": [
                  "F5"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-supplier-p02",
                "text": "That is a concrete division of labor. A supplier brings usable computing capacity; Runpod brings customers and the machinery for selling access to them. It also helps explain why developer distribution matters. If a supplier can reach a populated buying platform rather than recruit every customer itself, the platform has something to sell upstream as well as downstream. QumulusAI's concentration disclosure shows the importance of that channel to one supplier. It does not establish Runpod's dependence on that supplier.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I5"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-supplier-p03",
                "text": "On the disclosed split, a $100 transaction leaves $20 with Runpod before Runpod's own costs. That arithmetic is not a company-wide gross-margin estimate. The filing describes one relationship, and it does not resolve how Runpod recognizes revenue across different products or supplier contracts. Applying that percentage to the $325 million ARR headline would create a confident-looking answer to a question the evidence does not settle.",
                "evidence_type": "calculation_and_analysis",
                "source_ids": [
                  "S5",
                  "S2"
                ],
                "claim_ids": [
                  "I6",
                  "U1",
                  "U2"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-supplier-p04",
                "text": "CEO Zhen Lu calls Runpod a software company in a strategy essay updated September 13. The company also openly recruits data centers, cloud providers and GPU operators as supply partners. Together, those statements make its intended position clear: a customer-facing software layer over contributed computing capacity. They do not establish that every machine uses the same commercial terms. [Lu’s strategy essay](https://www.runpod.io/blog/the-chips-got-faster-the-stack-didnt); [Partner program](https://www.runpod.io/partners).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S4",
                  "S10"
                ],
                "claim_ids": [
                  "F6"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-supplier-p05",
                "text": "Aggregating demand can make a supplier’s machines easier to sell, but only when the available hardware suits the work. An idle GPU with insufficient memory, in the wrong location or without access to the required data is not interchangeable with the machine a customer needs. Runpod’s matching and deployment layer therefore has to do more than display a large inventory. It must turn a particular customer request into a working allocation.\n\nThe advantage of this position is that growth need not require Runpod to finance every underlying facility itself. The trade-off is limited direct control over parts of the service on which customers depend. Supplier economics and customer promises still have to meet. A contract can allocate the cost of a failure; it cannot make an unavailable machine finish an image.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I7",
                  "I16"
                ]
              }
            ]
          },
          {
            "id": "pct-2026-11-08-company-runpod-workload",
            "heading": "The price of a completed photograph",
            "paragraphs": [
              {
                "id": "pct-2026-11-08-company-runpod-workload-p01",
                "text": "For a roughly 20GB model, Nodalview reported first inference in four to five minutes on Runpod versus six to seven on AWS and Google Cloud, including container retrieval and model loading. Processing was under a minute per image once workers ran. These are different clocks, reported in a vendor case study rather than an independently reproduced test. [Test scope and workload](https://www.runpod.io/case-studies/how-nodalview-chose-what-to-move).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S1"
                ],
                "claim_ids": [
                  "F7"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-workload-p02",
                "text": "A business buyer should keep both clocks. Steady-state processing measures how quickly an available worker completes work. Time to first useful output captures the wait before that worker can help. If demand arrives together, a good average processing time can coexist with an unacceptable queue. Conversely, keeping workers ready can reduce that wait while consuming some of the savings that made the specialist attractive.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I8"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-workload-p03",
                "text": "Runpod's documentation makes the purchasing choice explicit. Flex workers can scale to zero; active workers remain running. Billing follows worker runtime, rounded up to the second, and includes startup, execution and the waiting period before shutdown. Storage is additional. The documentation distinguishes loading a model into GPU memory from container retrieval and cached-model downloads, which normally happen before billed worker runtime. [Serverless pricing documentation, modified October 15](https://docs.runpod.io/serverless/pricing).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S6"
                ],
                "claim_ids": [
                  "F8"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-workload-p04",
                "text": "There is another important distinction between paying for resources and paying for an outcome. Runpod's October 16 billing explanation says an unsuccessful job can still incur charges while its worker runs. Repeated initialization failures, hung handlers and retries can therefore consume a balance without delivering successful requests. That is a disclosed billing model, not evidence that such charges are necessarily incorrect. [Failed-job billing explanation](https://www.runpod.io/blog/failed-requests-real-charges-how-runpod-serverless-billing-works-when-workers-fail).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S7"
                ],
                "claim_ids": [
                  "F9"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-workload-p05",
                "text": "For a photo-processing company, the useful denominator is consequently accepted images delivered within the promised time. The numerator should include compute, storage, retries, standby capacity and the engineering needed to operate the pipeline. A lower hourly GPU price can help, but it is only one input. A pilot should replay the actual afternoon arrival pattern, include a new-worker start, and record failed attempts as well as successes. Those measurements turn a cloud comparison into a product-cost comparison.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I9"
                ]
              }
            ]
          },
          {
            "id": "pct-2026-11-08-company-runpod-enterprise",
            "heading": "The contract has to catch up with the developer",
            "paragraphs": [
              {
                "id": "pct-2026-11-08-company-runpod-enterprise-p01",
                "text": "Runpod's September 24 enterprise announcement addresses the next purchasing obstacle. The company introduced organizational controls, SSO/SAML and cost-center billing, alongside a claimed ISO/IEC 27001 certification. It described reserved capacity, contractual service-level agreements and named technical account managers. Those features matter when an engineer's useful service becomes something a company must govern and keep available. Certification and feature availability still require checking against the customer's specific deployment and contract. [Enterprise announcement](https://www.runpod.io/press/runpod-expands-enterprise-platform).",
                "evidence_type": "attributed_fact_and_analysis",
                "source_ids": [
                  "S8"
                ],
                "claim_ids": [
                  "F10",
                  "I10"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-enterprise-p02",
                "text": "For procurement, the questions are ordinary but consequential. Which GPU and region are reserved? How much capacity is actually protected? Who answers an urgent support request? What happens to queued work during an interruption, and which costs remain payable? A service-level agreement's financial remedy may be much smaller than the customer's lost business. The buyer needs the operational answer as well as the credit schedule. For Runpod, a capacity promise can also change the risk of the business. A supplier arrangement without purchase minimums says little about the cost of reserving a reliable substitute for an enterprise customer. The financial attraction of flexible supply has to be evaluated alongside the commitments made on the other side of the platform.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I11",
                  "I18"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-enterprise-p03",
                "text": "A recent incident makes those questions timely without justifying a verdict on the whole platform. Runpod's official status log opened an incident affecting CA-MTL-1 and EU-SE-1 on October 31 at 20:30 UTC. It reported EU-SE-1 restored on November 3 and closed the remaining CA-MTL-1 issue on November 6 at 04:22 UTC. The final updates do not give a customer count or a completed root-cause account. [Official incident timeline](https://uptime.runpod.io/incident/1109748?mp=true).",
                "evidence_type": "attributed_fact",
                "source_ids": [
                  "S9"
                ],
                "claim_ids": [
                  "F11"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-enterprise-p04",
                "text": "The incident window should not be described as continuous failure for every user in both regions. The updates distinguish their restoration times and later refer to remaining machines in Montreal. Nor does one incident prove a comparative reliability ranking. It does give prospective customers a concrete event to discuss: which workloads were affected, what recovery required, and whether an equivalent interruption would be tolerable in their own application. A useful recovery exercise would move a representative job, confirm that its model and input data are available, prevent duplicate processing, and return the result through the normal application. Merely showing that another GPU starts does not test the whole customer experience.",
                "evidence_type": "source_limitation_and_analysis",
                "source_ids": [
                  "S9"
                ],
                "claim_ids": [
                  "I12",
                  "U3",
                  "I17"
                ]
              }
            ]
          },
          {
            "id": "pct-2026-11-08-company-runpod-closing",
            "heading": "The next useful numbers",
            "paragraphs": [
              {
                "id": "pct-2026-11-08-company-runpod-closing-p01",
                "text": "Runpod's opportunity is to collect a share of spending that developers would otherwise place through a different cloud interface, and to earn that share by reducing the work around the GPU. Its danger is that the software layer becomes responsible for a production promise while supplier costs, hardware availability and customer support consume the apparent advantage. Neither a large signup count nor a low catalog price settles that contest.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I13"
                ]
              },
              {
                "id": "pct-2026-11-08-company-runpod-closing-p02",
                "text": "For the company, the revealing next disclosures would be retained revenue after supplier payments, customer concentration, renewal behavior and the cost of meeting enterprise commitments. For buyers, the useful evidence is smaller and more immediate: an agreed workload, a measured peak, a complete bill and a tested recovery path. Nodalview's selective migration offers a sensible way to ask the question. Runpod need not own the rest of the application to earn an important place in it. It must make the part it runs worth depending on.",
                "evidence_type": "editorial_analysis",
                "source_ids": [],
                "claim_ids": [
                  "I14"
                ]
              }
            ]
          }
        ],
        "word_count": 1834,
        "estimated_reading_minutes": 8.3,
        "reading_rate_words_per_minute": 220,
        "editorial_checks": {
          "prepublication_checks": [
            "AWS retained by Nodalview, not Google Cloud.",
            "Use October $325m ARR disclosure rather than older $120m or $240m figures as latest claimed level.",
            "Do not equate 20% supplier split to Runpod corporate margin.",
            "Do not confuse total incident window with uniform regional/customer downtime.",
            "Do not compare warm processing time with full cold startup."
          ],
          "excluded_unverified_material": "Uncorroborated social-media security allegations were not used.",
          "source_date_caution": "Live pages may be revised. The CEO essay and documentation use visible modification dates, not invented publication dates.",
          "independent_interviews_conducted": false
        },
        "provenance": {
          "research_basis": "Public primary sources, including company disclosures, customer case study, product documentation, official incident log and supplier SEC filing. No private interviews or hands-on benchmark claimed.",
          "factual_cutoff_date": "2026-11-08",
          "entity_focus": "Runpod",
          "previous_company_subjects_excluded": [
            "SF Compute",
            "CoreWeave",
            "Crusoe",
            "Nebius",
            "Modal"
          ]
        }
      },
      "public_projection": {
        "version": "1.0.0",
        "date": "2026-11-10",
        "changes": [
          "Internal workspace artifact paths omitted where present. Historical license/status fields retain original provenance; current public reuse policy is stated separately."
        ],
        "rights": "PCT reserves rights in original editorial content; third-party rights retained. No general source-data reuse license."
      }
    },
    {
      "id": "PCT-20261108-DD-A01",
      "content_type": "article",
      "desk": "deals",
      "title": "Akamai’s $11.6 billion Anthropic deal puts the spending before the CPU revenue",
      "publication_date": "2026-11-08",
      "observation_date": "2026-11-08",
      "editorial_voice": "AI-assisted Deals Desk",
      "body": "# Akamai’s $11.6 billion Anthropic deal puts the spending before the CPU revenue\n\n*Deals Desk | November 8, 2026 | Analysis of a September agreement and November disclosures*\n\nAkamai’s big Anthropic contract is a CPU story with a capital bill. The September agreement calls for $11.6 billion of services; the buildout was estimated to require $5.5 billion of capital expenditure. Fresh November disclosures make the gap between winning that business and earning it easier to inspect. [September announcement](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154dex991.htm); [November 6 filing](https://www.sec.gov/Archives/edgar/data/1086222/000108622226000100/akam-20260930.htm).\n\nThe deal was signed September 18 and announced September 24, not this weekend. It covers two project plans for dedicated cloud capacity and support. Each has a seven-year initial term starting at its own service commencement, with payments dependent on delivery and availability requirements. That clock matters: dividing the headline by seven does not produce a forecast for the next calendar year. [Transaction 8-K](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154d8k.htm).\n\n## The memory order comes first\n\nAkamai’s September schedule put about $1.7 billion of capital expenditure in the fourth quarter of 2026, $3.1 billion in 2027 and $700 million in 2028. It expected this expansion’s service initiation in late Q2 2027, $150 million to $300 million of revenue that year, and full contracted revenue run-rate by the end of 2028. Those are dated management estimates, not completed deliveries. [September presentation, page 4](https://www.ir.akamai.com/static-files/59654fd2-bd7c-4090-9ded-737b72ab2bb7).\n\nThe procurement mechanism is unusually concrete. A September 24 build request authorized Jabil to acquire roughly $1.7 billion of memory components. Akamai would pay the corresponding supplier invoices when Jabil received them; Jabil would hold the parts for Akamai and repurchase them at cost as they were used. The filing establishes an order and payment structure, not that every component has arrived. [Jabil arrangement in the 8-K](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154d8k.htm).\n\nEconomically, this brings supply-chain exposure onto the cloud provider’s books before the associated service revenue ramps. Buying early can reduce the chance of missing a build window. It also ties money to inventory and a deployment schedule. A large signed customer can justify that decision without eliminating the cost of carrying it.\n\nThere is a more recent execution update. On the November 5 earnings call, CFO Ed McGowan said Akamai had contracted all the colocation space and power needed for the expansion and secured about 70% of required hardware, memory and other components. He also described price adjustments protecting margins against increases on unordered components. These are management statements in a third-party transcript; the disclosed mechanism cannot be fully reconstructed from the redacted contract. [Published call transcript](https://www.fool.com/earnings/call-transcripts/2026/11/06/akamai-akam-q3-2026-earnings-call-transcript/).\n\nThe distinction is practical. Contracted space can still have a future availability date. Secured components still need installation and testing. Price protection may change who bears inflation; it does not necessarily remove schedule risk. None of those milestones alone demonstrates a customer-ready service.\n\n## The customer also gets an equity option\n\nAnthropic received a warrant, not an already-completed 5% ownership purchase. Forty percent of the warrant vests on the first payment under Project Plan 3. Three further tranches unlock as additional commitments accumulate in $3 billion steps, up to $9 billion. The common-share-equivalent exercise price is $111.33. The extra business requires expansion agreements; it should not be added to the signed $11.6 billion. [Warrant agreement](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154dex41.htm); [Announcement](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154dex991.htm).\n\nThe November 6 quarterly filing gives a useful checkpoint: as of September 30, no warrant shares had vested or been exercised. Akamai nevertheless expected the first tranche to vest and recorded its $149.1 million fair value as additional paid-in capital with a corresponding asset. That is accounting for an expected customer incentive, not evidence of exercise proceeds or cash collected. The September snapshot does not establish the warrant’s status today. [10-Q, Note 9](https://www.sec.gov/Archives/edgar/data/1086222/000108622226000100/akam-20260930.htm).\n\nThis adds a second price to the cloud sale. The service commitment brings prospective revenue, while the warrant gives the customer potential participation in the supplier’s upside. Existing shareholders need to weigh that incentive alongside equipment spending and operating costs. Treating the contract headline as the entire economic bargain leaves out something material.\n\n## A high operating cash margin still has hardware underneath\n\nAkamai supplied another helpful comparison on November 5. Its model for four large cloud deals signed this year projects about $2.2 billion of annual revenue once fully ramped, roughly 75% adjusted EBITDA margin and approximately 30% non-GAAP operating margin. The model includes about $1 billion of annual depreciation. These are aggregate forecasts across four deals, not realized results or a standalone margin disclosure for Anthropic. [Supplemental schedule, page 3](https://www.ir.akamai.com/static-files/aaf06a02-7f57-4b58-97a6-34442b6444af).\n\nDepreciation explains why quoting only the larger percentage can mislead. Equipment costs money before deployment, then its cost is recognized over time. An EBITDA measure strips out that expense. It can help describe operations, but it cannot by itself tell a reader whether the earlier capital investment earned an attractive return. The contractual revenue total is equally insufficient: timing, operating costs, taxes and financing all matter.\n\n## What the documents still cannot price\n\nThe master agreement filed in November assigns infrastructure allocation to Akamai and is generally non-exclusive, subject to separately agreed terms. But confidential provisions are redacted and the project-plan detail is not substantively disclosed. The public package therefore cannot supply a reliable per-core price, exact machine count or complete customer acceptance schedule. [Filed master agreement](https://www.sec.gov/Archives/edgar/data/1086222/000108622226000100/akam10q9302026ex101.htm).\n\nThis is also why the transaction belongs in a CPU-cloud category rather than a GPU-hour comparison. Converting dollars into accelerator equivalents would manufacture a quantity the deal does not disclose. The stronger market signal is narrower: an AI developer has committed substantial money to the surrounding compute platform, giving a provider reason to reserve components and expand.\n\nThe next evidence should be operational: capacity accepted under the September plans, revenue tied to that rollout, and spending measured against the original timetable. For now, the record supports a signed expansion with procurement and site contracting under way. Its economic test is whether those advance commitments turn into usable service on time, at a return that pays for both the hardware and the customer incentive.\n\n",
      "sources": [
        {
          "source_id": "PCT-20261108-DD-S01",
          "id": "PCT-20261108-DD-S01",
          "title": "Akamai September transaction 8-K",
          "url": "https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154d8k.htm",
          "publication_date": "2026-09-24",
          "retrieved_date": "2026-11-08",
          "type": "primary_regulatory_filing",
          "locator": "Item 1.01; Anthropic project plans and Jabil build request",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        },
        {
          "source_id": "PCT-20261108-DD-S02",
          "id": "PCT-20261108-DD-S02",
          "title": "Akamai announcement of Anthropic expansion",
          "url": "https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154dex991.htm",
          "publication_date": "2026-09-24",
          "retrieved_date": "2026-11-08",
          "type": "primary_issuer_release",
          "locator": "CPU workload scope, capital expenditure estimate and expansion option",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        },
        {
          "source_id": "PCT-20261108-DD-S03",
          "id": "PCT-20261108-DD-S03",
          "title": "Akamai September deal presentation",
          "url": "https://www.ir.akamai.com/static-files/59654fd2-bd7c-4090-9ded-737b72ab2bb7",
          "publication_date": "2026-09-24",
          "retrieved_date": "2026-11-08",
          "type": "primary_issuer_presentation",
          "locator": "Page 4, deal timing and cash generation phasing; estimates as of September 24",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        },
        {
          "source_id": "PCT-20261108-DD-S04",
          "id": "PCT-20261108-DD-S04",
          "title": "Akamai third-quarter 10-Q",
          "url": "https://www.sec.gov/Archives/edgar/data/1086222/000108622226000100/akam-20260930.htm",
          "publication_date": "2026-11-06",
          "retrieved_date": "2026-11-08",
          "type": "primary_regulatory_filing",
          "locator": "Note 9 warrant accounting; Anthropic concentration risk; quarter ended September 30",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        },
        {
          "source_id": "PCT-20261108-DD-S05",
          "id": "PCT-20261108-DD-S05",
          "title": "Akamai and Anthropic warrant agreement",
          "url": "https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154dex41.htm",
          "publication_date": "2026-09-24",
          "retrieved_date": "2026-11-08",
          "type": "primary_contract_exhibit",
          "locator": "Section 1.01 Vesting Event and warrant terms; issued September 18",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        },
        {
          "source_id": "PCT-20261108-DD-S06",
          "id": "PCT-20261108-DD-S06",
          "title": "Akamai and Anthropic Master Services Agreement",
          "url": "https://www.sec.gov/Archives/edgar/data/1086222/000108622226000100/akam10q9302026ex101.htm",
          "publication_date": "2026-11-06",
          "retrieved_date": "2026-11-08",
          "type": "primary_contract_exhibit",
          "locator": "May 5 agreement; Sections 3.2, 4.2, 16; project-plan schedules not substantively disclosed",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        },
        {
          "source_id": "PCT-20261108-DD-S07",
          "id": "PCT-20261108-DD-S07",
          "title": "Akamai Q3 supplemental earnings schedule",
          "url": "https://www.ir.akamai.com/static-files/aaf06a02-7f57-4b58-97a6-34442b6444af",
          "publication_date": "2026-11-05",
          "retrieved_date": "2026-11-08",
          "type": "primary_issuer_presentation",
          "locator": "Page 3, aggregate fully ramped economics of four large CIS deals signed year to date",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        },
        {
          "source_id": "PCT-20261108-DD-S08",
          "id": "PCT-20261108-DD-S08",
          "title": "Akamai Q3 earnings-call transcript",
          "url": "https://www.fool.com/earnings/call-transcripts/2026/11/06/akamai-akam-q3-2026-earnings-call-transcript/",
          "publication_date": "2026-11-06",
          "retrieved_date": "2026-11-08",
          "type": "secondary_transcript_of_primary_speech",
          "locator": "November 5 call, CFO Ed McGowan on Anthropic colo, components and price-adjustment mechanism; not independently verified deployment",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        },
        {
          "source_id": "PCT-20261108-DD-S09",
          "id": "PCT-20261108-DD-S09",
          "title": "Akamai Q3 earnings release",
          "url": "https://www.ir.akamai.com/news-releases/news-release-details/akamai-reports-third-quarter-2026-financial-results",
          "publication_date": "2026-11-05",
          "retrieved_date": "2026-11-08",
          "type": "primary_issuer_release",
          "locator": "Actual Q3 CIS revenue and company-level forward guidance",
          "access": "Public source read; relevant document passages checked where applicable.",
          "license": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          },
          "observation_date": "2026-11-08",
          "license_status": {
            "status": "no_open_republication_license_verified",
            "treatment": "Facts paraphrased and linked; source text, slides and images not republished; rights remain with respective owners"
          }
        }
      ],
      "facts": [
        {
          "id": "PCT-20261108-DD-F01",
          "statement": "Signing, disclosure, component procurement and customer service acceptance are separate milestones.",
          "source_ids": [
            "PCT-20261108-DD-S01",
            "PCT-20261108-DD-S03",
            "PCT-20261108-DD-S08"
          ],
          "claim": "Signing, disclosure, component procurement and customer service acceptance are separate milestones."
        },
        {
          "id": "PCT-20261108-DD-F02",
          "statement": "The November economics slide covers four large CIS deals; it is not an Anthropic-only profit forecast.",
          "source_ids": [
            "PCT-20261108-DD-S07"
          ],
          "claim": "The November economics slide covers four large CIS deals; it is not an Anthropic-only profit forecast."
        },
        {
          "id": "PCT-20261108-DD-F03",
          "statement": "The first warrant tranche had not vested as of September 30, according to the November 6 filing.",
          "source_ids": [
            "PCT-20261108-DD-S04"
          ],
          "claim": "The first warrant tranche had not vested as of September 30, according to the November 6 filing."
        }
      ],
      "inferences": [
        {
          "id": "PCT-20261108-DD-I01",
          "statement": "Early component purchasing reduces some supply risk while increasing the period of capital exposure before service revenue.",
          "claim": "Early component purchasing reduces some supply risk while increasing the period of capital exposure before service revenue."
        },
        {
          "id": "PCT-20261108-DD-I02",
          "statement": "The equity incentive must be considered alongside service revenue and capital spending when assessing shareholder economics.",
          "claim": "The equity incentive must be considered alongside service revenue and capital spending when assessing shareholder economics."
        },
        {
          "id": "PCT-20261108-DD-I03",
          "statement": "A CPU-cloud commitment cannot supply an executable GPU-hour price without invented capacity assumptions.",
          "claim": "A CPU-cloud commitment cannot supply an executable GPU-hour price without invented capacity assumptions."
        }
      ],
      "unknowns": [
        {
          "uncertainty_id": "PCT-20261108-DD-U01",
          "subject": "Delivery",
          "detail": "No full delivery or customer acceptance independently verified; contracted or secured is not installed."
        },
        {
          "uncertainty_id": "PCT-20261108-DD-U02",
          "subject": "Pricing adjustment",
          "detail": "Management description available; complete operative mechanics not publicly reconstructed from redacted/omitted project-plan detail."
        },
        {
          "uncertainty_id": "PCT-20261108-DD-U03",
          "subject": "Warrant",
          "detail": "September 30 snapshot only; no claim that it remained unvested through November 8."
        },
        {
          "uncertainty_id": "PCT-20261108-DD-U04",
          "subject": "Capacity and price",
          "detail": "Exact CPU count, site-specific capacity and per-core unit price not established."
        },
        {
          "uncertainty_id": "PCT-20261108-DD-U05",
          "subject": "Forecast timing",
          "detail": "Deal-specific schedule is the September 24 estimate; broader November CIS forecasts include earlier contracts and cannot be substituted for this deal’s actual revenue."
        }
      ],
      "correction_history": [],
      "license_status": "Original article; third-party facts paraphrased and linked. No open redistribution license verified for source documents; no source images or substantial source text reproduced.",
      "public_url": null,
      "schema_version": "pct-research-item-1.0",
      "desk_extensions": {
        "schema_version": "pct.deals-desk.v1",
        "edition_date": "2026-11-08",
        "section": "Deals Desk",
        "word_count": 985,
        "estimated_read_minutes": 4.9,
        "news_status": "Fresh analysis of November 5–6 disclosures about a September 18 agreement, not a new November 8 signing",
        "paragraphs": [
          {
            "id": "PCT-20261108-DD-P01",
            "classification": "analysis_with_facts",
            "source_ids": [
              "PCT-20261108-DD-S02",
              "PCT-20261108-DD-S04"
            ],
            "text": "Akamai’s big Anthropic contract is a CPU story with a capital bill. The September agreement calls for $11.6 billion of services; the buildout was estimated to require $5.5 billion of capital expenditure. Fresh November disclosures make the gap between winning that business and earning it easier to inspect. [September announcement](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154dex991.htm); [November 6 filing](https://www.sec.gov/Archives/edgar/data/1086222/000108622226000100/akam-20260930.htm)."
          },
          {
            "id": "PCT-20261108-DD-P02",
            "classification": "verified_contract_fact",
            "source_ids": [
              "PCT-20261108-DD-S01"
            ],
            "text": "The deal was signed September 18 and announced September 24, not this weekend. It covers two project plans for dedicated cloud capacity and support. Each has a seven-year initial term starting at its own service commencement, with payments dependent on delivery and availability requirements. That clock matters: dividing the headline by seven does not produce a forecast for the next calendar year. [Transaction 8-K](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154d8k.htm)."
          },
          {
            "id": "PCT-20261108-DD-SECTION01",
            "classification": "heading",
            "source_ids": [],
            "text": "The memory order comes first"
          },
          {
            "id": "PCT-20261108-DD-P03",
            "classification": "issuer_forecast",
            "source_ids": [
              "PCT-20261108-DD-S03"
            ],
            "text": "Akamai’s September schedule put about $1.7 billion of capital expenditure in the fourth quarter of 2026, $3.1 billion in 2027 and $700 million in 2028. It expected this expansion’s service initiation in late Q2 2027, $150 million to $300 million of revenue that year, and full contracted revenue run-rate by the end of 2028. Those are dated management estimates, not completed deliveries. [September presentation, page 4](https://www.ir.akamai.com/static-files/59654fd2-bd7c-4090-9ded-737b72ab2bb7)."
          },
          {
            "id": "PCT-20261108-DD-P04",
            "classification": "verified_contract_fact",
            "source_ids": [
              "PCT-20261108-DD-S01"
            ],
            "text": "The procurement mechanism is unusually concrete. A September 24 build request authorized Jabil to acquire roughly $1.7 billion of memory components. Akamai would pay the corresponding supplier invoices when Jabil received them; Jabil would hold the parts for Akamai and repurchase them at cost as they were used. The filing establishes an order and payment structure, not that every component has arrived. [Jabil arrangement in the 8-K](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154d8k.htm)."
          },
          {
            "id": "PCT-20261108-DD-P05",
            "classification": "analysis",
            "source_ids": [],
            "text": "Economically, this brings supply-chain exposure onto the cloud provider’s books before the associated service revenue ramps. Buying early can reduce the chance of missing a build window. It also ties money to inventory and a deployment schedule. A large signed customer can justify that decision without eliminating the cost of carrying it."
          },
          {
            "id": "PCT-20261108-DD-P06",
            "classification": "attributed_management_claim",
            "source_ids": [
              "PCT-20261108-DD-S08"
            ],
            "text": "There is a more recent execution update. On the November 5 earnings call, CFO Ed McGowan said Akamai had contracted all the colocation space and power needed for the expansion and secured about 70% of required hardware, memory and other components. He also described price adjustments protecting margins against increases on unordered components. These are management statements in a third-party transcript; the disclosed mechanism cannot be fully reconstructed from the redacted contract. [Published call transcript](https://www.fool.com/earnings/call-transcripts/2026/11/06/akamai-akam-q3-2026-earnings-call-transcript/)."
          },
          {
            "id": "PCT-20261108-DD-P07",
            "classification": "analysis",
            "source_ids": [],
            "text": "The distinction is practical. Contracted space can still have a future availability date. Secured components still need installation and testing. Price protection may change who bears inflation; it does not necessarily remove schedule risk. None of those milestones alone demonstrates a customer-ready service."
          },
          {
            "id": "PCT-20261108-DD-SECTION02",
            "classification": "heading",
            "source_ids": [],
            "text": "The customer also gets an equity option"
          },
          {
            "id": "PCT-20261108-DD-P08",
            "classification": "verified_contract_fact",
            "source_ids": [
              "PCT-20261108-DD-S05",
              "PCT-20261108-DD-S02"
            ],
            "text": "Anthropic received a warrant, not an already-completed 5% ownership purchase. Forty percent of the warrant vests on the first payment under Project Plan 3. Three further tranches unlock as additional commitments accumulate in $3 billion steps, up to $9 billion. The common-share-equivalent exercise price is $111.33. The extra business requires expansion agreements; it should not be added to the signed $11.6 billion. [Warrant agreement](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154dex41.htm); [Announcement](https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154dex991.htm)."
          },
          {
            "id": "PCT-20261108-DD-P09",
            "classification": "verified_as_of_fact",
            "source_ids": [
              "PCT-20261108-DD-S04"
            ],
            "text": "The November 6 quarterly filing gives a useful checkpoint: as of September 30, no warrant shares had vested or been exercised. Akamai nevertheless expected the first tranche to vest and recorded its $149.1 million fair value as additional paid-in capital with a corresponding asset. That is accounting for an expected customer incentive, not evidence of exercise proceeds or cash collected. The September snapshot does not establish the warrant’s status today. [10-Q, Note 9](https://www.sec.gov/Archives/edgar/data/1086222/000108622226000100/akam-20260930.htm)."
          },
          {
            "id": "PCT-20261108-DD-P10",
            "classification": "analysis",
            "source_ids": [],
            "text": "This adds a second price to the cloud sale. The service commitment brings prospective revenue, while the warrant gives the customer potential participation in the supplier’s upside. Existing shareholders need to weigh that incentive alongside equipment spending and operating costs. Treating the contract headline as the entire economic bargain leaves out something material."
          },
          {
            "id": "PCT-20261108-DD-SECTION03",
            "classification": "heading",
            "source_ids": [],
            "text": "A high operating cash margin still has hardware underneath"
          },
          {
            "id": "PCT-20261108-DD-P11",
            "classification": "issuer_forecast",
            "source_ids": [
              "PCT-20261108-DD-S07"
            ],
            "text": "Akamai supplied another helpful comparison on November 5. Its model for four large cloud deals signed this year projects about $2.2 billion of annual revenue once fully ramped, roughly 75% adjusted EBITDA margin and approximately 30% non-GAAP operating margin. The model includes about $1 billion of annual depreciation. These are aggregate forecasts across four deals, not realized results or a standalone margin disclosure for Anthropic. [Supplemental schedule, page 3](https://www.ir.akamai.com/static-files/aaf06a02-7f57-4b58-97a6-34442b6444af)."
          },
          {
            "id": "PCT-20261108-DD-P12",
            "classification": "analysis",
            "source_ids": [],
            "text": "Depreciation explains why quoting only the larger percentage can mislead. Equipment costs money before deployment, then its cost is recognized over time. An EBITDA measure strips out that expense. It can help describe operations, but it cannot by itself tell a reader whether the earlier capital investment earned an attractive return. The contractual revenue total is equally insufficient: timing, operating costs, taxes and financing all matter."
          },
          {
            "id": "PCT-20261108-DD-SECTION04",
            "classification": "heading",
            "source_ids": [],
            "text": "What the documents still cannot price"
          },
          {
            "id": "PCT-20261108-DD-P13",
            "classification": "verified_document_observation",
            "source_ids": [
              "PCT-20261108-DD-S06"
            ],
            "text": "The master agreement filed in November assigns infrastructure allocation to Akamai and is generally non-exclusive, subject to separately agreed terms. But confidential provisions are redacted and the project-plan detail is not substantively disclosed. The public package therefore cannot supply a reliable per-core price, exact machine count or complete customer acceptance schedule. [Filed master agreement](https://www.sec.gov/Archives/edgar/data/1086222/000108622226000100/akam10q9302026ex101.htm)."
          },
          {
            "id": "PCT-20261108-DD-P14",
            "classification": "analysis",
            "source_ids": [],
            "text": "This is also why the transaction belongs in a CPU-cloud category rather than a GPU-hour comparison. Converting dollars into accelerator equivalents would manufacture a quantity the deal does not disclose. The stronger market signal is narrower: an AI developer has committed substantial money to the surrounding compute platform, giving a provider reason to reserve components and expand."
          },
          {
            "id": "PCT-20261108-DD-P15",
            "classification": "analysis",
            "source_ids": [],
            "text": "The next evidence should be operational: capacity accepted under the September plans, revenue tied to that rollout, and spending measured against the original timetable. For now, the record supports a signed expansion with procurement and site contracting under way. Its economic test is whether those advance commitments turn into usable service on time, at a return that pays for both the hardware and the customer incentive."
          }
        ],
        "structured_record": {
          "record_id": "PCT-20261108-DD-D01",
          "seller": "Akamai Technologies, Inc.",
          "buyer": "Anthropic PBC",
          "deal_type": "dedicated CPU cloud capacity and managed support",
          "signed_date": "2026-09-18",
          "announced_date": "2026-09-24",
          "latest_primary_disclosure_date": "2026-11-06",
          "contract_value_usd": 11600000000,
          "currency": "USD",
          "value_basis": "Approximate aggregate commitment under Project Plans 2 and 3; subject to performance and termination provisions",
          "term_years": 7,
          "term_start": "Separate service start date for each project plan",
          "optional_expansion_usd": 9000000000,
          "optional_expansion_status": "Not counted as committed value; requires additional agreements",
          "capital_expenditure_estimate_usd": 5500000000,
          "capital_expenditure_estimate_date": "2026-09-24",
          "hardware_category": "CPU workloads; GPU quantities not disclosed",
          "gpu_count": null,
          "cpu_count": null,
          "price_per_core_hour": null,
          "executable_price_per_gpu_hour": null,
          "delivered_capacity_mw": null,
          "execution_status": "Signed and in procurement/site contracting; no full customer acceptance verified",
          "execution_evidence": {
            "as_of": "2026-11-05",
            "source_id": "PCT-20261108-DD-S08",
            "label": "Management claim reported in third-party transcript",
            "space_power": "All required colocation space and power contracted, with availability during future ramp",
            "components": "Approximately 70% secured; not described here as installed or accepted"
          },
          "warrant_status": {
            "as_of": "2026-09-30",
            "vested": false,
            "exercised": false,
            "current_status_after_as_of": "Not established by reviewed dated disclosure",
            "source_id": "PCT-20261108-DD-S04"
          },
          "id": "PCT-20261108-DD-D01",
          "content_type": "transaction_record",
          "schema_version": "pct-transaction-record-1.0",
          "publication_date": "2026-11-08",
          "observation_date": "2026-11-08",
          "parent_article_id": "PCT-20261108-DD-A01",
          "source_ids": [
            "PCT-20261108-DD-S01",
            "PCT-20261108-DD-S02",
            "PCT-20261108-DD-S03",
            "PCT-20261108-DD-S04",
            "PCT-20261108-DD-S05",
            "PCT-20261108-DD-S06",
            "PCT-20261108-DD-S07",
            "PCT-20261108-DD-S08",
            "PCT-20261108-DD-S09"
          ],
          "factual_claims": "See parent article facts array and linked source IDs; management claims and forecasts retain their labels.",
          "inference": "See parent article inferences array; no transaction price or delivered capacity is inferred from the contract value.",
          "unknowns": [
            {
              "uncertainty_id": "PCT-20261108-DD-U01",
              "subject": "Delivery",
              "detail": "No full delivery or customer acceptance independently verified; contracted or secured is not installed."
            },
            {
              "uncertainty_id": "PCT-20261108-DD-U02",
              "subject": "Pricing adjustment",
              "detail": "Management description available; complete operative mechanics not publicly reconstructed from redacted/omitted project-plan detail."
            },
            {
              "uncertainty_id": "PCT-20261108-DD-U03",
              "subject": "Warrant",
              "detail": "September 30 snapshot only; no claim that it remained unvested through November 8."
            },
            {
              "uncertainty_id": "PCT-20261108-DD-U04",
              "subject": "Capacity and price",
              "detail": "Exact CPU count, site-specific capacity and per-core unit price not established."
            },
            {
              "uncertainty_id": "PCT-20261108-DD-U05",
              "subject": "Forecast timing",
              "detail": "Deal-specific schedule is the September 24 estimate; broader November CIS forecasts include earlier contracts and cannot be substituted for this deal’s actual revenue."
            }
          ],
          "correction_history": [],
          "license_status": "Original article; third-party facts paraphrased and linked. No open redistribution license verified for source documents; no source images or substantial source text reproduced."
        },
        "uncertainty": [
          {
            "id": "PCT-20261108-DD-U01",
            "subject": "Delivery",
            "status": "No full delivery or customer acceptance independently verified; contracted or secured is not installed."
          },
          {
            "id": "PCT-20261108-DD-U02",
            "subject": "Pricing adjustment",
            "status": "Management description available; complete operative mechanics not publicly reconstructed from redacted/omitted project-plan detail."
          },
          {
            "id": "PCT-20261108-DD-U03",
            "subject": "Warrant",
            "status": "September 30 snapshot only; no claim that it remained unvested through November 8."
          },
          {
            "id": "PCT-20261108-DD-U04",
            "subject": "Capacity and price",
            "status": "Exact CPU count, site-specific capacity and per-core unit price not established."
          },
          {
            "id": "PCT-20261108-DD-U05",
            "subject": "Forecast timing",
            "status": "Deal-specific schedule is the September 24 estimate; broader November CIS forecasts include earlier contracts and cannot be substituted for this deal’s actual revenue."
          }
        ],
        "history": [
          {
            "id": "PCT-20261108-DD-H01",
            "event_date": "2026-05-05",
            "event": "Master services agreement executed",
            "source_ids": [
              "PCT-20261108-DD-S06"
            ]
          },
          {
            "id": "PCT-20261108-DD-H02",
            "event_date": "2026-09-18",
            "event": "Project Plans 2 and 3 and warrant entered",
            "source_ids": [
              "PCT-20261108-DD-S01",
              "PCT-20261108-DD-S05"
            ]
          },
          {
            "id": "PCT-20261108-DD-H03",
            "event_date": "2026-09-24",
            "event": "Expansion announced; Jabil memory build request reported",
            "source_ids": [
              "PCT-20261108-DD-S01",
              "PCT-20261108-DD-S02"
            ]
          },
          {
            "id": "PCT-20261108-DD-H04",
            "event_date": "2026-09-30",
            "disclosed_date": "2026-11-06",
            "event": "Quarter-end warrant snapshot: none vested or exercised",
            "source_ids": [
              "PCT-20261108-DD-S04"
            ]
          },
          {
            "id": "PCT-20261108-DD-H05",
            "event_date": "2026-11-05",
            "event": "Fresh aggregate deal economics and management procurement update",
            "source_ids": [
              "PCT-20261108-DD-S07",
              "PCT-20261108-DD-S08"
            ]
          },
          {
            "id": "PCT-20261108-DD-H06",
            "event_date": "2026-11-06",
            "event": "10-Q and master agreement filed",
            "source_ids": [
              "PCT-20261108-DD-S04",
              "PCT-20261108-DD-S06"
            ]
          }
        ],
        "licensing": {
          "article": "Original analysis composed for this edition; no third-party images, charts or substantial quotations reproduced",
          "sources": "Public accessibility is not treated as an open license; retained attribution and links",
          "direct_quoted_words": 0
        },
        "dedup": {
          "checked_date": "2026-11-08",
          "scope": "Canonical PCT editions dated November 3–7, 2026",
          "result": "No earlier article on this lead in the checked pilot editions."
        },
        "editorial_cautions": [
          "Do not label September signing new today.",
          "Do not combine $9B expansion potential with the signed $11.6B.",
          "Do not call warrant issuance an exercised ownership purchase.",
          "Keep as-of September 30 attached to warrant status.",
          "Keep portfolio scope attached to margin and depreciation figures.",
          "Do not turn components secured or sites contracted into operational capacity.",
          "Do not claim reported management statements were independently field-verified."
        ],
        "provenance": {
          "reporting_method": "Public-document review; no interviews or private-source claims",
          "written_date": "2026-11-08",
          "body_sha256": "f82a041a113f410b5dbb20d0976eec675045ef3875baaf13738d26af79dd2b1d",
          "last_reviewed": "2026-11-08"
        },
        "dates": {
          "edition": "2026-11-08",
          "agreement": "2026-09-18",
          "announcement": "2026-09-24",
          "latest_primary_disclosure": "2026-11-06",
          "warrant_observation_as_of": "2026-09-30",
          "execution_management_update": "2026-11-05"
        }
      },
      "public_projection": {
        "version": "1.0.0",
        "date": "2026-11-10",
        "changes": [
          "Internal workspace artifact paths omitted where present. Historical license/status fields retain original provenance; current public reuse policy is stated separately."
        ],
        "rights": "PCT reserves rights in original editorial content; third-party rights retained. No general source-data reuse license."
      }
    },
    {
      "id": "pct-2026-11-08-formation-spectrum-repacking",
      "content_type": "article",
      "desk": "formation",
      "title": "Assembling compute capacity before selling it",
      "publication_date": "2026-11-08",
      "observation_date": "2026-11-08",
      "editorial_voice": "AI-assisted Formation Desk | accessible academic analysis",
      "body": "A cluster can have sixteen idle GPUs without being able to accept a sixteen-GPU job. Eight may sit in one network domain and eight in another, while the workload needs a single approved fabric. A price list can describe those fragments accurately and still offer nothing the buyer can use. The missing service is sometimes an agreed rearrangement of existing allocations.\n\nThe U.S. broadcast incentive auction offers a useful historical case. Its distinctive institution combined payments for relinquishing rights with an engineering process that made the resulting spectrum usable. For compute, the relevant question is whether an operator can assemble a deliverable block by compensating customers willing to move, while preserving every commitment that remains.\n\n## The market included a rearrangement\n\nCongress authorized the broadcast auction in 2012. Broadcasters could voluntarily relinquish spectrum usage rights, including through channel sharing or a move from UHF to VHF. The FCC could also reassign television channels, subject to statutory protections, to reorganize the remaining broadcast band. Voluntary sale and channel reassignment were different parts of the institution. [47 U.S.C. 1452, especially subsections (a) and (b)](https://uscode.house.gov/us/usc/119-118/t47/s1452/).\n\nThe auction's results were announced on April 13, 2017. Gross proceeds were approximately $19.8 billion. The process repurposed 84 MHz, of which 70 MHz went to licensed use; the two quantities should not be treated as interchangeable. [FCC results announcement, April 13, 2017](https://docs.fcc.gov/public/attachments/DOC-344397A1.pdf); [federal spectrum report, May 2019, pages 5 and 55](https://www.nitrd.gov/nitrdgroups/images/f/f0/Emerging-Technologies-and-Impact-on-Non-Federal-Spectrum-Demand-Report-May-2019.pdf).\n\nThe essential coordination problem was that a station's departure affected the possible locations of other stations. Acquiring a cheap set of relinquishments was useful only if the stations that remained could be assigned channels under the interference constraints. Those dependencies made the identity of the sellers matter alongside their prices.\n\nThe FCC's reverse auction used descending offers. Feasibility checks determined whether stations could be accommodated in their previous bands as bids were processed; an unsuccessful check could freeze a station's offer rather than reduce it. Forward bidding tested demand for the newly available licenses. If the final-stage conditions were not met, the design provided for another stage with a lower spectrum-clearing target. [Final bidding procedures, October 14, 2015, sections V and VII](https://www.govinfo.gov/content/pkg/FR-2015-10-14/pdf/2015-25579.pdf).\n\nThe computational component was concrete. The FCC published SATFC, software for checking whether channel reassignments were feasible. It combined a logical representation of the constraints with specialized solving methods. [FCC SATFC repository, release notes through March 29, 2016](https://github.com/FCC/SATFC). This was infrastructure for deciding which combinations could be offered, not merely a way to display bids.\n\n## A hypothetical compute rearrangement\n\nConsider one operator with two sixteen-GPU fabric domains, A and B, using compatible hardware. Each has eight idle GPUs. A contains an eight-GPU job whose customer is willing to checkpoint and restart elsewhere. B contains another eight-GPU job that must remain in place. A new customer wants sixteen GPUs together for eight hours.\n\nThese are invented conditions, not an available provider offer. Assume the flexible job can run in B, its customer approves the move, and the operator can preserve tenant isolation, data-location requirements and software permissions. Moving that job into B's idle half would leave all sixteen GPUs in A available for the new customer. The initial inventory contains enough idle devices, but the rearrangement makes them usable as one block.\n\nThe operator should first test the move: checkpoint duration, restart success, storage access and performance in the destination domain. The new eight-hour service window begins only after migration and acceptance checks. If the customer needs an immediate start, or the existing job cannot tolerate the pause, the proposed trade is infeasible regardless of the offered price.\n\nSuppose the new customer offers $400 for the block. The incumbent accepts an $80 credit for moving. Migration and validation cost $40. Assume a further $240 covers delivery costs and all foregone alternative revenue, including use of the otherwise-idle space in B. The remaining contribution is $40. Every dollar figure is hypothetical; this arithmetic is a way to expose required inputs, not a compute price estimate.\n\nA second move costing $60 would make the same arrangement unattractive under those assumptions. So would discovering that the $240 allowance omitted valuable demand for B. The test concerns the whole rearrangement, including displaced opportunities. Counting the new buyer's payment as the entire benefit would overstate the gain.\n\n## What the analogy can support\n\nThe inference for compute is that a marketplace may need a joint allocation plan before it can quote useful supply. An operator could collect conditional offers to shift jobs, combine them with a buyer's requirements, and commit only when the complete placement is technically feasible and commercially acceptable. Such a service would sell access to a particular workable configuration rather than sum nominal idle hours.\n\nA small trial need not reproduce the FCC's auction. With one movable job, direct negotiation and a scheduler may be sufficient. Competitive bidding becomes more interesting when several feasible rearrangements exist and their costs are privately known. That possibility does not establish that an exchange would outperform an operator's existing scheduling process.\n\nThe authority differs sharply. The FCC operated under a special statute; a compute intermediary gains no power to relocate tenants from this analogy. Its permissions must come from actual contracts and consent. A solver also cannot manufacture checkpoint support, acceptable application performance or customer willingness to move. Those are inputs that require evidence.\n\nThe timescale matters as well. Spectrum reorganization involved a 39-month transition period ending in July 2020, with some extensions completed by September. [FCC Order 21-111, released October 25, 2021, paragraph 6 and footnote 20](https://docs.fcc.gov/public/attachments/FCC-21-111A1_Rcd.pdf). Compute availability can expire while a migration is still being prepared. A rearrangement that is eventually feasible may be worthless for a buyer's start deadline.\n\nThe practical experiment is therefore narrow: one operator, explicitly movable workloads and tested destination placements. Record incremental contribution, displaced revenue, migration failures, missed deadlines and useful work delivered. Include proposals rejected as infeasible. The useful lesson from spectrum repacking is that assembly costs and compatibility constraints belong inside the market's allocation decision. Whether paying for that assembly creates worthwhile compute supply remains an empirical question.",
      "sources": [
        {
          "id": "pct-source-47usc1452",
          "title": "47 U.S.C. 1452: Special requirements for incentive auction of broadcast TV spectrum",
          "publisher": "Office of the Law Revision Counsel, U.S. House of Representatives",
          "url": "https://uscode.house.gov/us/usc/119-118/t47/s1452/",
          "published_at": "2012-02-22",
          "version_date": "2026-10-07",
          "retrieved_at": "2026-11-08",
          "date_note": "Original enactment February 22, 2012. Retrieved consolidated text says laws in effect through Public Law 119-120, October 7, 2026; 2018 additions are separately identified. Historical claims use subsections (a) and (b), not later amendments.",
          "evidence_type": "Primary statutory text",
          "supports": "Voluntary relinquishment options and FCC reassignment authority subject to protections.",
          "verification": "Read official HTML, subsections (a), (b), (c), source credit and amendments.",
          "license_status": "U.S. federal statutory text; no source imagery or substantial verbatim text reproduced.",
          "publication_date": "2012-02-22",
          "observation_date": "2026-11-08"
        },
        {
          "id": "pct-source-fcc-doc344397-20170413",
          "title": "FCC Announces Results of World's First Broadcast Incentive Auction",
          "publisher": "Federal Communications Commission",
          "url": "https://docs.fcc.gov/public/attachments/DOC-344397A1.pdf",
          "published_at": "2017-04-13",
          "retrieved_at": "2026-11-08",
          "date_note": "Document dateline verified; search crawler dates were not used as publication dates.",
          "evidence_type": "Primary regulator results announcement",
          "supports": "April 13 results announcement and approximately $19.8 billion in gross revenue for 70 MHz.",
          "verification": "Official PDF indexed extracts read; direct PDF open returned HTTP 403. Amount corroborated by the fully accessible 2019 federal report. No claim to full-PDF inspection.",
          "license_status": "U.S. government-authored notice, generally public domain in the United States under 17 U.S.C. 105; factual paraphrase only.",
          "publication_date": "2017-04-13",
          "observation_date": "2026-11-08"
        },
        {
          "id": "pct-source-ostp-spectrum-201905",
          "title": "Emerging Technologies and Their Expected Impact on Non-Federal Spectrum Demand",
          "publisher": "White House Office of Science and Technology Policy / NITRD",
          "url": "https://www.nitrd.gov/nitrdgroups/images/f/f0/Emerging-Technologies-and-Impact-on-Non-Federal-Spectrum-Demand-Report-May-2019.pdf",
          "published_at": "2019-05",
          "retrieved_at": "2026-11-08",
          "date_note": "Month and year printed on cover. No exact publication day asserted.",
          "evidence_type": "Primary federal retrospective report, incorporating FCC auction results",
          "supports": "84 MHz repurposed, including 70 MHz licensed, and approximately $19.8 billion gross proceeds.",
          "verification": "Read parsed official PDF, printed pages 5 and 55, and Appendix B footnote 39 on printed page 74. Bidding end March 30 distinguished from official April 13 closing notice.",
          "license_status": "Federal report factual paraphrase. No third-party charts, tables or images reproduced; no blanket license inferred for embedded third-party material.",
          "publication_date": "2019-05",
          "observation_date": "2026-11-08"
        },
        {
          "id": "pct-source-fcc15-78-fr2015-25579",
          "title": "Broadcast Incentive Auction Scheduled To Begin on March 29, 2016; Procedures for Competitive Bidding in Auction 1000",
          "publisher": "Federal Communications Commission / Federal Register",
          "url": "https://www.govinfo.gov/content/pkg/FR-2015-10-14/pdf/2015-25579.pdf",
          "published_at": "2015-10-14",
          "adopted_at": "2015-08-06",
          "released_at": "2015-08-11",
          "retrieved_at": "2026-11-08",
          "date_note": "Federal Register publication/effective date October 14; underlying FCC 15-78 adopted August 6 and released August 11.",
          "evidence_type": "Primary final bidding procedures",
          "supports": "Descending offers, feasibility checks and frozen bids; smaller clearing target at a later stage if final-stage conditions fail.",
          "verification": "Read official PDF text, section V paragraphs 85, 102, 104, 108–110; lower-target discussion paragraph 34; section VII cost and final-stage provisions. Detailed freeze rules are more nuanced than simplified narrative.",
          "license_status": "U.S. government final rule, generally public domain in the United States under 17 U.S.C. 105; paraphrase only.",
          "publication_date": "2015-10-14",
          "observation_date": "2026-11-08"
        },
        {
          "id": "pct-source-fcc-satfc-release",
          "title": "FCC SATFC public repository",
          "publisher": "Federal Communications Commission; software authors credited to Auctionomics collaborators",
          "url": "https://github.com/FCC/SATFC",
          "published_at": null,
          "version_date": "2016-03-29",
          "retrieved_at": "2026-11-08",
          "date_note": "README release history identifies SATFC 2.3 dated March 29, 2016. No claim that this is the repository's last commit date or exact deployed version.",
          "evidence_type": "Primary implementation documentation",
          "supports": "Public feasibility-checking software and its constraint-based computational approach.",
          "verification": "Read repository README introduction and release history; no software run or benchmark performed.",
          "license_status": "Repository has a separate LICENSE file; its terms were not verified. No code or documentation passages are reproduced, and no software-use permission is asserted.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "pct-source-fcc21-111",
          "title": "Expanding the Economic and Innovation Opportunities of Spectrum Through Incentive Auction, Order FCC 21-111",
          "publisher": "Federal Communications Commission",
          "url": "https://docs.fcc.gov/public/attachments/FCC-21-111A1_Rcd.pdf",
          "published_at": "2021-10-25",
          "adopted_at": "2021-10-22",
          "retrieved_at": "2026-11-08",
          "date_note": "Adoption/release dates verified from official FCC-21-111A1.pdf cover extract; cited FCC Record version is the same order.",
          "evidence_type": "Primary regulator retrospective order",
          "supports": "39-month transition ended July 13, 2020; a few extensions completed by September 30, 2020.",
          "verification": "Official indexed text read for paragraph 6 and footnote 20. Footnote qualifies broad completion language; no full PDF rendering claimed.",
          "license_status": "U.S. government-authored order, generally public domain in the United States under 17 U.S.C. 105; factual paraphrase only.",
          "publication_date": "2021-10-25",
          "observation_date": "2026-11-08"
        }
      ],
      "facts": [
        {
          "id": "pct-20261108-formation-f01",
          "claim": "The 2012 statute provided voluntary relinquishment options and separate channel-reassignment authority with protections.",
          "source_ids": [
            "pct-source-47usc1452"
          ]
        },
        {
          "id": "pct-20261108-formation-f02",
          "claim": "The FCC announced results April 13, 2017; gross proceeds were about $19.8 billion. Of 84 MHz repurposed, 70 MHz was licensed.",
          "source_ids": [
            "pct-source-fcc-doc344397-20170413",
            "pct-source-ostp-spectrum-201905"
          ]
        },
        {
          "id": "pct-20261108-formation-f03",
          "claim": "Final procedures coupled descending reverse offers with feasibility checking, freezing and a provision for stages at lower clearing targets.",
          "source_ids": [
            "pct-source-fcc15-78-fr2015-25579"
          ]
        },
        {
          "id": "pct-20261108-formation-f04",
          "claim": "The FCC publicly documented SATFC as a spectrum-repacking feasibility checker.",
          "source_ids": [
            "pct-source-fcc-satfc-release"
          ]
        },
        {
          "id": "pct-20261108-formation-f05",
          "claim": "The 39-month transition ended July 13, 2020; the FCC later recorded some extensions completed September 30.",
          "source_ids": [
            "pct-source-fcc21-111"
          ]
        }
      ],
      "inferences": [
        {
          "id": "pct-2026-11-08-formation-spectrum-repacking-I001",
          "claim": "A compute allocation can require a jointly feasible rearrangement of incumbent jobs before idle inventory becomes usable supply.",
          "label": "PCT editorial inference"
        },
        {
          "id": "pct-2026-11-08-formation-spectrum-repacking-I002",
          "claim": "Conditional relocation offers and a placement plan could support such a service; this is an institutional proposal, not an observed PCT or provider capability.",
          "label": "PCT editorial inference"
        },
        {
          "id": "pct-2026-11-08-formation-spectrum-repacking-I003",
          "claim": "Incremental contribution must include relocation compensation, transition costs and displaced opportunities across the complete plan.",
          "label": "PCT editorial inference"
        },
        {
          "id": "pct-2026-11-08-formation-spectrum-repacking-I004",
          "claim": "A small operator pilot can compare assembly against existing scheduling; a full auction or independent exchange is not presumed necessary.",
          "label": "PCT editorial inference"
        }
      ],
      "unknowns": [
        "No current compute provider, buyer or incumbent has agreed to the illustrative arrangement.",
        "All job sizes, placement permissions, durations and dollar values in the example are hypothetical. Migration capabilities are assumptions requiring tests.",
        "No estimate of market-wide fragmented inventory, realized efficiency, auction causal welfare effect or compute transaction demand is made.",
        "A feasibility solver's output depends on accurate constraints and does not certify unmeasured performance, security, consent or software rights.",
        "FCC release and order excerpts were publicly indexed but direct FCC PDF retrieval was unavailable. Readable official federal sources supported key mechanics and totals.",
        "The analogy supplies no statutory or contractual relocation powers to a compute intermediary."
      ],
      "correction_history": [],
      "license_status": "Original PCT editorial prose; public license not selected. Source rights remain separate. No source code, images, datasets or substantial quotations reproduced.",
      "public_url": null,
      "schema_version": "pct-research-item-1.0",
      "desk_extensions": {
        "word_count": 1008,
        "estimated_reading_minutes": 4.6,
        "hypothetical_example": {
          "status": "Invented design assumptions; not a supplier quote or verified operational capability",
          "fabric_domains": 2,
          "gpus_per_domain": 16,
          "idle_gpus_per_domain_before_move": 8,
          "movable_job_gpus": 8,
          "new_job_gpus": 16,
          "new_job_service_hours": 8,
          "currency": "USD",
          "new_customer_payment": 400,
          "incumbent_credit": 80,
          "migration_validation_cost": 40,
          "delivery_and_all_opportunity_costs": 240,
          "incremental_contribution": 40,
          "formula": "400 - 80 - 40 - 240 = 40",
          "timing": "Eight-hour window begins after migration and acceptance, not during them.",
          "boundary": "Costs include all displaced opportunities in both domains. Compatible hardware, consent, tenant isolation, licensing, data location, checkpoint support and successful restart are assumed and must be tested."
        },
        "analogy_limits": [
          "Special public-law reassignment authority does not transfer to compute.",
          "Compute hours and start windows expire; the historical transition took years.",
          "A logical placement plan cannot itself establish application performance or consent.",
          "Small cases may be handled better by ordinary scheduling and negotiation."
        ],
        "distinct_from_previous_issue": "Nov7 gas asset managers retained a service obligation while trading surplus. This piece concerns the coupled acquisition of flexibility and physical rearrangement needed to assemble a feasible block. It does not reuse the prior freight clearing, Nord Pool, grain inspection, airport-slot transfer-rights or FERC asset-manager institution.",
        "editorial_checks": [
          "No em dashes; links are verified public URLs.",
          "84 MHz total repurposed distinguished from 70 MHz licensed.",
          "Gross auction revenue is not treated as welfare or net Treasury receipts.",
          "July transition deadline qualified by September extension completion.",
          "Historical dates distinguished from crawl/retrieval dates.",
          "No claim of actual compute price, live inventory or implemented product.",
          "Prose explicitly identifies the compute conclusion as inference and example as hypothetical."
        ],
        "schema_version": "pct.article.v1",
        "dek": "The FCC's broadcast incentive auction joined bidding to a feasible rearrangement of incumbent users. A compute counterpart would need consent, tested migrations and enough value to pay for the move."
      },
      "public_projection": {
        "version": "1.0.0",
        "date": "2026-11-10",
        "changes": [
          "Internal workspace artifact paths omitted where present. Historical license/status fields retain original provenance; current public reuse policy is stated separately."
        ],
        "rights": "PCT reserves rights in original editorial content; third-party rights retained. No general source-data reuse license."
      }
    },
    {
      "id": "pct-2026-11-08-spotlight-fba4521180",
      "content_type": "article",
      "desk": "spotlight",
      "title": "A GPU has to earn its place in an audio transcription queue",
      "publication_date": "2026-11-08",
      "observation_date": "2026-11-08",
      "editorial_voice": "AI-assisted Spotlight Desk",
      "body": "# A GPU has to earn its place in an audio transcription queue\n\n*Product spotlight | Peterson Compute Trading | November 8, 2026*\n*AI-assisted public-source analysis. No transcription benchmark, cloud deployment or purchase was performed. All prices and calculations are in US dollars.*\n\nConsider a hypothetical archive of a thousand hours of recorded lectures that needs searchable transcripts by next week. The recordings are owned or appropriately licensed, the speakers have consented, and the collection contains no sensitive material. This is a useful inference workload: it can wait, it divides into independent jobs, and a failed attempt should not require starting the whole collection again.\n\nThe first purchasing question is whether a GPU saves money after its supporting CPU and memory are counted. Using Modal's public rates, the two configurations below put that hurdle at about 3.60 times the CPU worker's throughput. A hypothetical fourfold speedup clears it narrowly. That small margin is the interesting part.\n\n## Start with a transcript worth keeping\n\nThe proposed engine is faster-whisper, a CTranslate2 implementation of Whisper. Begin with the same small English checkpoint on both machines, the same beam size, language setting, timestamp requirements and silence filtering. CPU INT8 and GPU INT8 with FP16 arithmetic are candidate execution modes, not a guarantee of identical output. Choose acceptable transcription quality before selecting the cheapest configuration. [Project documentation](https://github.com/SYSTRAN/faster-whisper)\n\nCreate a representative evaluation set before processing the archive: quiet narration, unfamiliar technical names, poor microphones, accents, music and long pauses. Compare against human-checked reference transcripts. Record word error rate, missing passages, invented passages and timestamp failures. A cheap transcript that requires extensive correction can be more expensive than the compute it saved. Whisper's model card explicitly warns about hallucinated text and uneven performance across languages and speakers. [Whisper model card](https://github.com/openai/whisper/blob/main/model-card.md)\n\nThe project's published performance table is useful orientation, but it compares a small model on an Intel i7-12700K with large-v2 on an RTX 3070 Ti in separate sections. Those rows do not establish a speedup for this CPU-versus-L4 experiment. Timing must also consume faster-whisper's output iterator: transcription work happens during iteration, so timing only the initial call can produce a misleadingly tiny number. [Benchmark and usage notes](https://github.com/SYSTRAN/faster-whisper)\n\n## Price the whole worker\n\nModal's November 8 rate card lists CPU at $0.0000131 per physical-core-second, memory at $0.00000222 per GiB-second, and an L4 at $0.000222 per second. The planning configurations are four physical CPU cores plus 8 GiB, or one L4 plus one physical core and 8 GiB. Multiplying by 3,600 gives $0.252576 and $0.910296 per worker-hour respectively. These use default Function rates with automatic placement, not a regional premium or a Sandbox. [Published prices](https://modal.com/pricing)\n\nThe resulting GPU-to-CPU price ratio is 3.604. For equal accepted output and no extra overhead, GPU throughput must exceed that multiple to reduce compute spending. Actual CPU and memory usage can change the arithmetic: Modal bills the higher of requested resources and actual usage, and its CPU setting denotes physical cores rather than vCPUs. The proposed allocations are sizing assumptions to test, not measured requirements. [Resource accounting](https://modal.com/docs/guide/resources)\n\n## A fourfold speedup can save only five dollars\n\nFor illustration only, assume the CPU consumes 0.20 worker-hours per audio hour and the GPU consumes 0.05. These are invented planning inputs, not measured or vendor-promised speeds. Processing 1,000 audio hours would then consume 200 CPU worker-hours or 50 GPU worker-hours, costing $50.52 or $45.51 before other charges. The GPU saves about $5.00, or 9.9%, despite finishing the compute four times faster at equal worker counts.\n\nNow add ten billed worker-hours of overhead to each option across the collection. CPU compute becomes $53.04; GPU compute becomes $54.62. The ranking reverses. Across 6,000 ten-minute shards, ten hours is an average of six additional billed seconds per shard. This sensitivity test does not predict a cold start on every shard. Containers can be reused, and the mix of initialization, downloading, retries and processing will determine actual overhead. [Cold-start behavior](https://modal.com/docs/guide/cold-start)\n\nThese are gross compute scenarios. They exclude credits, taxes, storage, network transfer, transcription review, orchestration and engineering time. Queueing can affect the deadline without being equivalent to billed worker time. Modal documents no minimum usage-time increment; an hourly equivalent is a comparison unit, not a one-hour reservation. [Billing documentation](https://modal.com/docs/guide/billing)\n\n## Design for interruption\n\nModal Functions are preemptible by default and restart with the same input after an interruption. Its documentation recommends small tasks or checkpoints and retry-safe work. The nonpreemptible option adds a CPU-and-memory premium but is not supported for GPU Functions. An L4 worker therefore needs an interruption plan rather than a checkbox promising uninterrupted execution. [Preemption documentation](https://modal.com/docs/guide/preemption)\n\nA practical design begins with roughly ten-minute audio shards, adjusted around utterance boundaries. Any overlap must be reconciled when timestamps and text are joined, and the extra decoding belongs in the bill. Give each shard a stable identity derived from the source content hash, time range, model revision and decoding options. Store complete outputs durably under that identity and use conditional creation or an equivalent atomic commit. A retry checks for an accepted result before doing expensive work again.\n\nKeep a separate manifest for accepted, failed and pending shards. Do not mark a shard complete until its output is durable and validated. Quarantine corrupt inputs, cap retries and reconcile the manifest against outputs at the end. This proposed design limits repeated work to unfinished shards, subject to correct persistence and concurrency handling. It has not been deployed or tested here.\n\n## A small trial before the archive\n\nMeasure both configurations on the same sample, separating cold and warm runs. Record completed, quality-accepted audio hours per billed worker-hour, peak memory, retries and elapsed completion time. Cache the selected model revision, reuse initialized workers where appropriate, and repeat runs enough to expose variation. If batching improves throughput, rerun quality checks and retain its memory and startup costs.\n\nThe faster-whisper implementation and upstream Whisper code and weights carry MIT licenses. Preserve required notices when distributing software, and separately verify rights to the audio and any additional components. A software license does not grant permission to process someone else's recordings. [faster-whisper license](https://github.com/SYSTRAN/faster-whisper/blob/master/LICENSE), [Whisper licensing](https://github.com/openai/whisper#license)\n\nThe decision rule is simple: buy the configuration with the lowest total cost per accepted audio hour that can meet the deadline. This case gives the GPU a measurable hurdle, not an endorsement. If a representative trial barely clears 3.60 times the throughput, small overhead or quality differences deserve more attention than the GPU's headline speed.\n",
      "sources": [
        {
          "id": "SP-S1",
          "title": "Pricing: Pay per second for GPUs and CPUs",
          "publisher": "Modal",
          "published": null,
          "updated": null,
          "url": "https://modal.com/pricing",
          "accessed": "2026-11-08",
          "access_note": "Opened current public rate card. Checked Function CPU, memory and L4 rates, physical-core units, regional premiums, and separate Sandbox rates. No publication or update date displayed.",
          "license_status": "Public documentation; reuse license not established; linked and paraphrased, no screenshots reproduced.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "SP-S2",
          "title": "Faster Whisper transcription with CTranslate2",
          "publisher": "SYSTRAN",
          "published": null,
          "updated": null,
          "url": "https://github.com/SYSTRAN/faster-whisper",
          "accessed": "2026-11-08",
          "access_note": "Opened README, including benchmark environments, CPU/GPU modes, generator behavior, batching, and comparison cautions. Benchmark table identifies faster-whisper v1.1.0; it is not a new November 8 benchmark.",
          "license_status": "Repository MIT license verified separately at SP-S3; required notices apply to redistributed software.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "SP-S3",
          "title": "faster-whisper LICENSE",
          "publisher": "SYSTRAN",
          "published": null,
          "updated": null,
          "url": "https://github.com/SYSTRAN/faster-whisper/blob/master/LICENSE",
          "accessed": "2026-11-08",
          "access_note": "Opened MIT license including copyright and permission-notice condition.",
          "license_status": "MIT; copyright 2023 SYSTRAN. License applies to the covered software, not all linked projects or input audio.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "SP-S4",
          "title": "Configuring CPU, memory, and disk",
          "publisher": "Modal",
          "published": null,
          "updated": null,
          "url": "https://modal.com/docs/guide/resources",
          "accessed": "2026-11-08",
          "access_note": "Opened resource units, request parameters and billing section. CPU refers to physical cores; CPU/memory billing uses higher of request and actual usage.",
          "license_status": "Public documentation; reuse license not established; linked and paraphrased.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "SP-S5",
          "title": "Cold start performance",
          "publisher": "Modal",
          "published": null,
          "updated": null,
          "url": "https://modal.com/docs/guide/cold-start",
          "accessed": "2026-11-08",
          "access_note": "Opened container reuse, initialization and queueing explanations. No numerical startup assumption was borrowed as a measured result.",
          "license_status": "Public documentation; reuse license not established; linked and paraphrased.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "SP-S6",
          "title": "Preemption",
          "publisher": "Modal",
          "published": null,
          "updated": null,
          "url": "https://modal.com/docs/guide/preemption",
          "accessed": "2026-11-08",
          "access_note": "Opened defaults, same-input restart, checkpoint/idempotency advice, and GPU exclusion from nonpreemptible option.",
          "license_status": "Public documentation; reuse license not established; linked and paraphrased.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "SP-S7",
          "title": "Billing",
          "publisher": "Modal",
          "published": null,
          "updated": null,
          "url": "https://modal.com/docs/guide/billing",
          "accessed": "2026-11-08",
          "access_note": "Opened resource-use/request billing and no-minimum-usage-time-increment statement.",
          "license_status": "Public documentation; reuse license not established; linked and paraphrased.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "SP-S8",
          "title": "Model Card: Whisper",
          "publisher": "OpenAI",
          "published": null,
          "updated": null,
          "url": "https://github.com/openai/whisper/blob/main/model-card.md",
          "accessed": "2026-11-08",
          "access_note": "Opened limitations and use sections covering hallucinations, unequal accuracy and consent/context evaluation cautions.",
          "license_status": "Public repository material; no verbatim passages reproduced.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        },
        {
          "id": "SP-S9",
          "title": "Whisper README licensing",
          "publisher": "OpenAI",
          "published": null,
          "updated": null,
          "url": "https://github.com/openai/whisper#license",
          "accessed": "2026-11-08",
          "access_note": "Opened README License section explicitly applying MIT License to code and model weights.",
          "license_status": "MIT for upstream Whisper code and weights; input recordings and other dependencies require separate review.",
          "publication_date": null,
          "observation_date": "2026-11-08"
        }
      ],
      "facts": [
        {
          "id": "SP-F1",
          "claim": "Default Function list rates are USD 0.0000131/physical-core-second, 0.00000222/GiB-second and 0.000222/L4-second.",
          "source_ids": [
            "SP-S1"
          ],
          "evidence_type": "current_published_price_not_transaction_quote"
        },
        {
          "id": "SP-F2",
          "claim": "CPU allocation means physical cores. CPU and memory charges use the higher of requested resources and actual usage.",
          "source_ids": [
            "SP-S4"
          ],
          "evidence_type": "current_documentation"
        },
        {
          "id": "SP-F3",
          "claim": "Functions are preemptible by default and restart on the same input after preemption. Nonpreemptible execution is not supported for GPU Functions.",
          "source_ids": [
            "SP-S6"
          ],
          "evidence_type": "current_documentation"
        },
        {
          "id": "SP-F4",
          "claim": "Containers can be reused; cold-start latency includes queueing and initialization.",
          "source_ids": [
            "SP-S5"
          ],
          "evidence_type": "current_documentation"
        },
        {
          "id": "SP-F5",
          "claim": "Modal states there are no minimum usage-time increments.",
          "source_ids": [
            "SP-S7"
          ],
          "evidence_type": "current_billing_terms"
        },
        {
          "id": "SP-F6",
          "claim": "faster-whisper supports CPU INT8 and GPU INT8/FP16 modes. Transcription runs while iterating its segments generator. Its published CPU small-model and GPU large-v2 benchmark sections use different models and hardware.",
          "source_ids": [
            "SP-S2"
          ],
          "evidence_type": "project_documentation_not_independent_benchmark"
        },
        {
          "id": "SP-F7",
          "claim": "Whisper documentation warns of hallucinations, uneven recognition accuracy and the need for context-specific evaluation.",
          "source_ids": [
            "SP-S8"
          ],
          "evidence_type": "model_developer_documentation"
        },
        {
          "id": "SP-F8",
          "claim": "faster-whisper is MIT licensed, and upstream Whisper states its code and model weights are MIT licensed.",
          "source_ids": [
            "SP-S3",
            "SP-S9"
          ],
          "evidence_type": "published_software_license"
        }
      ],
      "inferences": [
        {
          "id": "SP-I1",
          "claim": "The proposed worker configurations cost USD 0.252576/hour (CPU) and USD 0.910296/hour (L4 including CPU and memory), requiring GPU throughput greater than 3.604047891 times CPU throughput to lower baseline compute cost.",
          "basis_fact_ids": [
            "SP-F1",
            "SP-F2"
          ],
          "label": "PCT arithmetic from public prices, not a quote or measured workload",
          "assumptions": [
            "CPU: 4 physical cores and 8 GiB; GPU: one L4, 1 physical core and 8 GiB; actual usage does not exceed requested CPU/memory.",
            "Default preemptible Function rates, automatic placement, no regional premiums.",
            "Equal quality acceptance and same checkpoint/decoding requirements; no output quality equivalence has been established."
          ]
        },
        {
          "id": "SP-I2",
          "claim": "At assumed real-time factors 0.20 and 0.05, gross compute for 1,000 audio hours is USD 50.5152 and USD 45.5148. GPU savings are USD 5.0004, or approximately 9.90%.",
          "basis_fact_ids": [
            "SP-F1"
          ],
          "label": "hypothetical sensitivity calculation, not measured performance",
          "assumptions": [
            "1,000 unique audio hours with required recording rights and speaker consent.",
            "CPU: 4 physical cores and 8 GiB; GPU: one L4, 1 physical core and 8 GiB; actual usage does not exceed requested CPU/memory.",
            "Default preemptible Function rates, automatic placement, no regional premiums.",
            "Equal quality acceptance and same checkpoint/decoding requirements; no output quality equivalence has been established.",
            "Hypothetical CPU real-time factor 0.20 and GPU real-time factor 0.05, where real-time factor is worker processing hours divided by input audio hours.",
            "6,000 nominal ten-minute shards before any overlap.",
            "Sensitivity case adds 10 billed worker-hours to each configuration, not a prediction of cold starts or failures."
          ]
        },
        {
          "id": "SP-I3",
          "claim": "Adding ten billed worker-hours to both cases changes gross compute to USD 53.04096 (CPU) and USD 54.61776 (GPU), reversing the ranking.",
          "basis_fact_ids": [
            "SP-F1"
          ],
          "label": "hypothetical overhead sensitivity, not a predicted interruption rate",
          "assumptions": [
            "1,000 unique audio hours with required recording rights and speaker consent.",
            "CPU: 4 physical cores and 8 GiB; GPU: one L4, 1 physical core and 8 GiB; actual usage does not exceed requested CPU/memory.",
            "Default preemptible Function rates, automatic placement, no regional premiums.",
            "Equal quality acceptance and same checkpoint/decoding requirements; no output quality equivalence has been established.",
            "Hypothetical CPU real-time factor 0.20 and GPU real-time factor 0.05, where real-time factor is worker processing hours divided by input audio hours.",
            "6,000 nominal ten-minute shards before any overlap.",
            "Sensitivity case adds 10 billed worker-hours to each configuration, not a prediction of cold starts or failures."
          ]
        },
        {
          "id": "SP-I4",
          "claim": "Stable shard identities, durable accepted outputs and atomic commits can limit restart work to unfinished shards; boundaries and overlaps must preserve transcript quality.",
          "basis_fact_ids": [
            "SP-F3"
          ],
          "label": "PCT proposed implementation; no deployment or transcription test"
        },
        {
          "id": "SP-I5",
          "claim": "A purchase decision should compare total cost per quality-accepted audio hour and deadline performance, not GPU-only hourly price.",
          "basis_fact_ids": [
            "SP-F1",
            "SP-F2",
            "SP-F7"
          ],
          "label": "PCT editorial interpretation"
        }
      ],
      "unknowns": [
        "Measured throughput, cold-start/initialization time, quality, peak memory and retry behavior for this exact small-English transcription configuration.",
        "Live capacity, quotas, available regions and account-specific pricing or terms.",
        "Total invoice after allowances, taxes, storage, transfer, orchestration and any additional charges.",
        "Whether 8 GiB and the stated core allocations are sufficient for each actual recording and batching setting.",
        "End-to-end transcript quality across chunk boundaries and overlapped segments.",
        "Production implementation, authentication, atomic persistence, replay safety and queueing behavior; no executable pipeline was deployed."
      ],
      "correction_history": [],
      "license_status": "Original PCT editorial text; public license not selected. Software and source rights remain separate. No source images or long quotations reproduced.",
      "public_url": null,
      "schema_version": "pct-research-item-1.0",
      "desk_extensions": {
        "id": "pct-2026-11-08-spotlight-faster-whisper",
        "section": "Product spotlight",
        "publisher": "Peterson Compute Trading",
        "as_of_date": "2026-11-08",
        "status": "desk_research_with_audited_hypothetical_cost_model",
        "subject": "Restartable faster-whisper lecture transcription on CPU or Modal L4 Functions",
        "deck": "A thousand-hour archive shows why a fourfold GPU speedup can leave only five dollars of compute savings.",
        "editorial_word_count": 1037,
        "reading_minutes": 4.7,
        "reading_wpm": 220,
        "publication_note": "Evergreen technical analysis using documentation and prices observed November 8, 2026. No claim of a November 8 product release or new benchmark.",
        "cost_model": {
          "currency": "USD",
          "as_of": "2026-11-08",
          "pricing_mode": "Default preemptible Modal Functions with automatic placement; no regional premium",
          "rate_source_id": "SP-S1",
          "cpu_rate_per_physical_core_second": 1.31e-05,
          "memory_rate_per_gib_second": 2.22e-06,
          "l4_rate_per_second": 0.000222,
          "cpu_worker": {
            "physical_cores": 4,
            "memory_gib": 8,
            "cost_per_hour": 0.252576
          },
          "gpu_worker": {
            "gpu": "L4",
            "gpu_count": 1,
            "physical_cores": 1,
            "memory_gib": 8,
            "cost_per_hour": 0.910296
          },
          "break_even_throughput_ratio": 3.604047890535918,
          "input_audio_hours": 1000,
          "hypothetical_cpu_real_time_factor": 0.2,
          "hypothetical_gpu_real_time_factor": 0.05,
          "cpu_processing_worker_hours": 200,
          "gpu_processing_worker_hours": 50,
          "cpu_gross_compute_usd": 50.5152,
          "gpu_gross_compute_usd": 45.5148,
          "gpu_gross_compute_savings_usd": 5.0004,
          "gpu_gross_compute_savings_percent": 9.898802736602052,
          "sensitivity_added_billed_worker_hours_each": 10,
          "cpu_with_overhead_gross_compute_usd": 53.04096,
          "gpu_with_overhead_gross_compute_usd": 54.61776,
          "nominal_shard_audio_minutes": 10,
          "nominal_shard_count": 6000,
          "sensitivity_mean_extra_billed_seconds_per_shard": 6,
          "assumptions": [
            "1,000 unique audio hours with required recording rights and speaker consent.",
            "CPU: 4 physical cores and 8 GiB; GPU: one L4, 1 physical core and 8 GiB; actual usage does not exceed requested CPU/memory.",
            "Default preemptible Function rates, automatic placement, no regional premiums.",
            "Equal quality acceptance and same checkpoint/decoding requirements; no output quality equivalence has been established.",
            "Hypothetical CPU real-time factor 0.20 and GPU real-time factor 0.05, where real-time factor is worker processing hours divided by input audio hours.",
            "6,000 nominal ten-minute shards before any overlap.",
            "Sensitivity case adds 10 billed worker-hours to each configuration, not a prediction of cold starts or failures."
          ],
          "exclusions": [
            "Credits and allowances",
            "Taxes",
            "Storage and network transfer",
            "Orchestration and separate preparation resources",
            "Transcript review and correction",
            "Engineering labor",
            "Any unmodeled resource usage or overlap decoding"
          ]
        },
        "implementation_plan": {
          "status": "proposed_not_executed",
          "inputs": "Rights-cleared, consented, non-sensitive English lecture audio; exact checkpoint revision and dependency versions pinned for trial.",
          "steps": [
            "Construct a human-checked representative evaluation sample and set acceptance criteria.",
            "Compare the same checkpoint and decoding options on CPU and L4; candidate modes CPU int8 and GPU int8_float16 require quality checks.",
            "Split near utterance boundaries into roughly ten-minute shards, track original offsets, reconcile any overlap.",
            "Identify each shard by content hash, bounds, checkpoint revision and decoding options.",
            "Persist complete validated output with conditional creation or equivalent atomic commit before recording acceptance.",
            "Skip accepted outputs on retry, cap attempts, quarantine permanent failures and reconcile the manifest.",
            "Record billed resource time and accepted audio duration separately from elapsed queueing and processing latency."
          ],
          "no_transactions": true,
          "no_deployment": true,
          "no_performance_test": true
        },
        "verification": {
          "date": "2026-11-08",
          "public_primary_sources_opened": 9,
          "arithmetic": "Decimal-based local assertions passed for hourly prices and both scenario totals.",
          "benchmark": "No transcription benchmarks run. Existing project benchmark is attributed context only.",
          "cloud_actions": "None; no account changes, uploads, purchases or deployments.",
          "source_date_caution": "Access dates are not publication dates. No last-updated date was visible for the cited pages.",
          "editorial_checks": [
            "Separated published facts from calculations and implementation proposals.",
            "Avoided comparing CPU small-model and GPU large-v2 benchmark rows as equivalent.",
            "Included CPU and memory alongside GPU cost.",
            "Distinguished worker-hours from wall-clock completion time.",
            "Did not imply free credits make workload free.",
            "Did not claim source audio rights derive from MIT software licensing."
          ]
        }
      },
      "public_projection": {
        "version": "1.0.0",
        "date": "2026-11-10",
        "changes": [
          "Internal workspace artifact paths omitted where present. Historical license/status fields retain original provenance; current public reuse policy is stated separately."
        ],
        "rights": "PCT reserves rights in original editorial content; third-party rights retained. No general source-data reuse license."
      }
    }
  ],
  "corrections": [
    {
      "date": "2026-11-05",
      "scope": "PCT_2026-11-04.json",
      "description": "H100 spot and on-demand 06.ai records shared an identifier; prior file remains unchanged, this edition uses unique observation IDs. No price correction implied.",
      "prior_archive_id": "libfile_5a5661b01848819197ddf74ee9b1d16d"
    }
  ],
  "omitted": "Separate benchmark, provider-price and market-signal datasets are not republished in this article archive."
}