Thesis · Hyperscalers, Cloud & Compute Services

Will at least two of Microsoft, Alphabet, Amazon and Meta state on a 2027 earnings call that they remain capacity- or power-constrained in AI compute?

Through 2027 the largest cloud operators continue to describe themselves as short of AI capacity rather than long, confirming installed accelerators stay fully utilized, though whether any will use that precise phrasing on an earnings call is itself uncertain.

8 dated tests · 0 resolved · next on

The tests More theses in Hyperscalers, Cloud & Compute Services

Who is on each side

The two sides.

Each side in DeepStack’s words, then the voices who argue it, attributed and paraphrased, dated where the date is known. A voice links to its page on DeepStack where one exists.

The case for yes

At least two hyperscalers call themselves capacity-constrained

Finite deployed GPU/TPU pools shared across search, ads, internal training and external cloud customers saturate, forcing rationing of marginal workloads. Management continues to guide capex higher and to say demand exceeds supply, while marginal or edge-case workloads get throttled rather than capacity sitting idle. On this view, at least two of Microsoft, Alphabet, Amazon and Meta describe themselves as capacity- or power-constrained on a 2027 earnings call.

The case for no

Fewer than two hyperscalers describe themselves as constrained

This side holds that fewer than two of Microsoft, Alphabet, Amazon and Meta describe themselves as capacity- or power-constrained on a 2027 earnings call. A precedent for slack appearing quickly is the compute freed up at xAI after its research effort stalled, which became available to market to external customers rather than staying locked to one user. Strong utilization of even aging infrastructure can coexist with freed-up surplus elsewhere, so a confirming sign would be two or more hyperscalers stating ample or excess AI capacity alongside reduced capex guidance.

6 voices

At least two hyperscalers call themselves capacity-constrained

  1. Moses Sternstein

    As demand for AI infrastructure continues to grow faster than production capacity can scale, orders remain unfilled for extended periods, driving up prices for scarce components and rental capacity while manufacturers of cooling and power systems…

    Source

  2. Kakashii

    Competitive pressure produces more products and more token-hungry models, while live capacity cannot expand because datacenters take about three years to build, so the excess demand is expressed as price rather than volume.

    Source

  3. Openclaw Unboxed

    Google runs multiple GPU-intensive operations. When that finite pool becomes constrained, something has to give. Machine-speed OpenClaw usage is an edge case that can be easily shed without affecting core products. Enforcement arrives when…

    Source

  4. a16z

    Baker argues the defining feature of the 2000 telecom bubble was that laid fiber optic capacity went unused (dark fiber), whereas today GPU capacity is fully saturated to the point of overheating, so the same overbuild dynamic that signaled the…

    Source

  5. AI Mafia

    When GPU capacity is insufficient, companies must allocate scarce compute resources to their highest-priority use cases, deferring feature rollouts and service tier expansions. OpenAI prioritizes GPT-4 training but cannot simultaneously deploy all…

    Source

  6. Daniel Romero

    Meta having spare capacity while Google runs a $460B backlog means compute is scarce industry-wide but unevenly distributed; Zuckerberg would not enter the neocloud business expecting GPUs to sell for pennies - he sees the shortage and wants to…

    Source

2 voices

Fewer than two hyperscalers describe themselves as constrained

  1. Tae Kim

    When xAI's AI research program faltered, most of the original team left, freeing up the compute infrastructure they had been using. This surplus compute became available to be marketed to external customers like Google and Anthropic.

    Source

  2. David George

    Strong compute utilization across even aging infrastructure, combined with Jevons' Paradox effects (reduced token costs driving increased usage), indicates genuine underlying demand for AI rather than speculative excess.

    Source

What would settle it

The dated tests.

Each test names the public source that settles it, the rule and the day. The Call’s number, the readers’ median and the machines’ numbers stand beside it, as on the Docket.

  1. Alphabet, purchases of property and equipment as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 31%.

    On the Docket Pendingth008-t03Pre-specified

    Question
    Did the purchases of property and equipment of Alphabet for the quarter ended September 30, 2026 come to at or above 31.0% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Alphabet (Forms 10-Q and 10-K)
    Rule
    Divide the purchases of property and equipment by the revenue that Alphabet reports for the quarter ended September 30, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 31.0%; no if it is below 31.0%.
    If the source is late, revised or silent
    If Alphabet has not filed by November 16, 2026, or a filing restates a period the rule reads, the test is inconclusive, and the Result says what was read and what was missing.
    Base rate
    Alphabet's purchases of property and equipment as a share of revenue was 37.5% for the quarter ended June 30, 2026, by its filing on SEC EDGAR.
    Outside view
    A year earlier, for the quarter ended June 30, 2025, the same line read 23.3%, so the rungs sit on the last print.
    What would change the number
    A change in product mix, a one-time charge, or a reclassified line would move the reading.

    The Call, before the dateExpects yes. Gives the yes answer 82%.

    82%

    0 · no100 · yes

    The machinesClaude 80% · DeepSeek 70%

    Claude: 80%DeepSeek: 70%
    The ladder 3 rungs on one reading, Oct 28, 2026: each asks whether it lands above its threshold, so the Call’s numbers down the rungs sketch where it expects the reading to land.

    1. Above 31%Call 82% · Claude 80% · DeepSeek 70%

      82%

    2. Above 36%Call 58% · Claude 80% · DeepSeek 57%

      58%

    3. Above 41%Call 33% · Claude 48% · DeepSeek 2%

      33%

    0 · no50100 · yes

    Your number

  2. Alphabet, purchases of property and equipment as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 36%.

    On the Docket Pendingth008-t04Pre-specified

    Question
    Did the purchases of property and equipment of Alphabet for the quarter ended September 30, 2026 come to at or above 36.0% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Alphabet (Forms 10-Q and 10-K)
    Rule
    Divide the purchases of property and equipment by the revenue that Alphabet reports for the quarter ended September 30, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 36.0%; no if it is below 36.0%.
    If the source is late, revised or silent
    If Alphabet has not filed by November 16, 2026, or a filing restates a period the rule reads, the test is inconclusive, and the Result says what was read and what was missing.
    Base rate
    Alphabet's purchases of property and equipment as a share of revenue was 37.5% for the quarter ended June 30, 2026, by its filing on SEC EDGAR.
    Outside view
    A year earlier, for the quarter ended June 30, 2025, the same line read 23.3%, so the rungs sit on the last print.
    What would change the number
    A change in product mix, a one-time charge, or a reclassified line would move the reading.

    The Call, before the dateExpects yes. Gives the yes answer 58%.

    58%

    0 · no100 · yes

    The machinesClaude 80% · DeepSeek 57%

    Claude: 80%DeepSeek: 57%

    Your number

  3. Alphabet, purchases of property and equipment as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 41%.

    On the Docket Pendingth008-t05Pre-specified

    Question
    Did the purchases of property and equipment of Alphabet for the quarter ended September 30, 2026 come to at or above 41.0% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Alphabet (Forms 10-Q and 10-K)
    Rule
    Divide the purchases of property and equipment by the revenue that Alphabet reports for the quarter ended September 30, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 41.0%; no if it is below 41.0%.
    If the source is late, revised or silent
    If Alphabet has not filed by November 16, 2026, or a filing restates a period the rule reads, the test is inconclusive, and the Result says what was read and what was missing.
    Base rate
    Alphabet's purchases of property and equipment as a share of revenue was 37.5% for the quarter ended June 30, 2026, by its filing on SEC EDGAR.
    Outside view
    A year earlier, for the quarter ended June 30, 2025, the same line read 23.3%, so the rungs sit on the last print.
    What would change the number
    A change in product mix, a one-time charge, or a reclassified line would move the reading.

    The Call, before the dateExpects no. Gives the yes answer 33%.

    33%

    0 · no100 · yes

    The machinesClaude 48% · DeepSeek 2%

    Claude: 48%DeepSeek: 2%

    Your number

  4. CoreWeave, purchases of property and equipment as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 190%.

    On the Docket Pendingth008-t01Open question

    Specification pending

    The machinesClaude 62% · DeepSeek 70%

    Claude: 62%DeepSeek: 70%

    Your number

  5. CoreWeave, operating cash flow as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 23%.

    On the Docket Pendingth008-t06Open question

    Specification pending

    The machinesClaude 73% · DeepSeek 74%

    Claude: 73%DeepSeek: 74%

    Your number

  6. CoreWeave, operating cash flow as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 26%.

    On the Docket Pendingth008-t07Open question

    Specification pending

    The machinesClaude 61% · DeepSeek 72%

    Claude: 61%DeepSeek: 72%

    Your number

  7. CoreWeave, operating cash flow as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 29%.

    On the Docket Pendingth008-t08Open question

    Specification pending

    The machinesClaude 63% · DeepSeek 40%

    Claude: 63%DeepSeek: 40%

    Your number

  8. Oracle, purchases of property and equipment as a share of revenue, quarter ended November 30, 2026: whether it comes in at or above 105%.

    On the Docket Pendingth008-t02Open question

    Specification pending

    The machinesClaude 55% · DeepSeek 40%

    Claude: 55%DeepSeek: 40%

    Your number

A thesis page states one question and both sides of it; DeepStack takes no side in words, and its view is the number it gives each test before the date. Analysis and opinion, not investment advice. Voices are paraphrased from public writing and attributed by name; DeepStack shows no score, ranking or accuracy for any person. Pre-specified means the words were approved on or before the test’s date. Specified after the date means they were written afterwards: such a test is counted and shown, and never rated. An open question has no words yet. The Call’s number is the chance, from 0 to 100, that the Call gives the yes answer to a test’s question, recorded before the test’s date; it is DeepStack’s own reading, not a market price and not advice. The readers’ mark on the same line is the readers’ median, the middle of the numbers signed-in readers gave the test, shown from 25 readers. The machines: Each model read this page and nothing else. The number that counts is the latest one each model gave before the test closed.