Thesis · Foundation Models & AI Intelligence

Will Microsoft report a company-wide gross margin percentage for the quarter ending June 30, 2027 that is at least 100 basis points above the same quarter of fiscal 2025, as disclosed in its SEC filings?

Falling cost per served token outpaces token price declines, so hyperscaler-reported gross margins expand despite surging AI inference volume and depreciation from the capex buildout.

6 dated tests · 0 resolved · next on

The tests More theses in Foundation Models & AI Intelligence

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

Cost per token falls faster than price

Hardware generation gains, inference software optimization (batching, caching, routing, distillation) and volume leverage cut cost per token faster than list prices fall, while agentic workloads inflect token volume. Reported cloud and company-level gross margins widen, the AI capex ROI overhang narrows, and reinvestment is funded from operating cash flow rather than margin sacrifice.

The case for no

The case against it

Tight compute supply combined with declining token pricing, so compressed margins or reduced investment returns for AI model developers. Introduction of real inference costs that scale with customer usage instead of marginal cost approaching zero, so traditional J-curve where customer lifetime value rises with retention no longer holds; retained customers become more expensive to serve, compressing margins. It would show first if hyperscaler gross margin percentage declines year over year while AI revenue grows, with management citing inference cost or depreciation.

2 voices

Cost per token falls faster than price

  1. Michael Spencer

    As inference costs decline through hardware improvements, software optimization, or volume leverage, a vendor can either cut prices or expand margins. Anthropic is expanding margins: compute spend per revenue dollar falling from 0.71 to 0.56 while…

    Source

  2. Market Sentiment

    If agents become profitable to run, token volumes inflect while unit costs fall faster than prices, so hyperscaler margins widen; once that margin improvement shows up in reported results, the market stops discounting the data center spend.

    Source

5 voices

The case against it

  1. DeepValue Capital

    While compute supply remains constrained, the price AI providers can charge per token is falling, squeezing the spread between input costs and revenue and potentially undermining the profitability case for AI infrastructure buildout.

    Source

  2. Saasholic

    Traditional SaaS had nearly zero marginal costs so the customer J-curve bent upward automatically. AI inverts this: cost to serve scales with usage because every query triggers inference and agent runs consume tokens. As AI improves and token costs…

    Source

  3. Prof G Markets

    Inference cost outrunning the labor it replaces

    Source

  4. Ruben Dominguez

    Traditional software has high fixed costs and low marginal costs per user. Usage-based AI inference reverses this: every user interaction incurs a direct variable cost. Profitability depends on controlling per-request costs through routing and…

    Source

  5. a16z

    Even as engineers achieve 10x and 100x efficiency improvements in inference, each new model generation offers dramatically better capabilities, and users immediately adopt the new model rather than capturing cost savings. The result is perpetually…

    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. Microsoft, research and development expense as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 10.8%.

    On the Docket Pendingth087-t01Pre-specified

    Question
    Did the research and development expense of Microsoft for the quarter ended September 30, 2026 come to at or above 10.8% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Microsoft (Forms 10-Q and 10-K)
    Rule
    Divide the research and development expense by the revenue that Microsoft 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 10.8%; no if it is below 10.8%.
    If the source is late, revised or silent
    If Microsoft 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
    Microsoft's research and development expense as a share of revenue was 11.1% 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 11.6%, 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 78%.

    78%

    0 · no100 · yes

    The machinesClaude 87% · DeepSeek 80%

    Claude: 87%DeepSeek: 80%
    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 10.8%Call 78% · Claude 87% · DeepSeek 80%

      78%

    2. Above 11.3%Call 48% · Claude 80% · DeepSeek 30%

      48%

    3. Above 11.9%Call 20% · Claude 39% · DeepSeek 14%

      20%

    0 · no50100 · yes

    Your number

  2. Microsoft, research and development expense as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 11.3%.

    On the Docket Pendingth087-t02Pre-specified

    Question
    Did the research and development expense of Microsoft for the quarter ended September 30, 2026 come to at or above 11.3% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Microsoft (Forms 10-Q and 10-K)
    Rule
    Divide the research and development expense by the revenue that Microsoft 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 11.3%; no if it is below 11.3%.
    If the source is late, revised or silent
    If Microsoft 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
    Microsoft's research and development expense as a share of revenue was 11.1% 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 11.6%, 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 dateEither answer fits the Call. Gives the yes answer 48%.

    48%

    0 · no100 · yes

    The machinesClaude 80% · DeepSeek 30%

    Claude: 80%DeepSeek: 30%

    Your number

  3. Microsoft, research and development expense as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 11.9%.

    On the Docket Pendingth087-t03Pre-specified

    Question
    Did the research and development expense of Microsoft for the quarter ended September 30, 2026 come to at or above 11.9% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Microsoft (Forms 10-Q and 10-K)
    Rule
    Divide the research and development expense by the revenue that Microsoft 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 11.9%; no if it is below 11.9%.
    If the source is late, revised or silent
    If Microsoft 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
    Microsoft's research and development expense as a share of revenue was 11.1% 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 11.6%, 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 20%.

    20%

    0 · no100 · yes

    The machinesClaude 39% · DeepSeek 14%

    Claude: 39%DeepSeek: 14%

    Your number

  4. Alphabet, research and development expense as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 14.8%.

    On the Docket Pendingth087-t04Pre-specified

    Question
    Did the research and development expense of Alphabet for the quarter ended September 30, 2026 come to at or above 14.8% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Alphabet (Forms 10-Q and 10-K)
    Rule
    Divide the research and development expense 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 14.8%; no if it is below 14.8%.
    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 research and development expense as a share of revenue was 15.2% 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 14.3%, 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 72%.

    72%

    0 · no100 · yes

    The machinesClaude 72% · DeepSeek 68%

    Claude: 72%DeepSeek: 68%
    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 14.8%Call 72% · Claude 72% · DeepSeek 68%

      72%

    2. Above 15.3%Call 46% · Claude 28% · DeepSeek 45%

      46%

    3. Above 15.9%Call 23% · Claude 33% · DeepSeek 15%

      23%

    0 · no50100 · yes

    Your number

  5. Alphabet, research and development expense as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 15.3%.

    On the Docket Pendingth087-t05Pre-specified

    Question
    Did the research and development expense of Alphabet for the quarter ended September 30, 2026 come to at or above 15.3% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Alphabet (Forms 10-Q and 10-K)
    Rule
    Divide the research and development expense 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 15.3%; no if it is below 15.3%.
    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 research and development expense as a share of revenue was 15.2% 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 14.3%, 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 dateEither answer fits the Call. Gives the yes answer 46%.

    46%

    0 · no100 · yes

    The machinesClaude 28% · DeepSeek 45%

    Claude: 28%DeepSeek: 45%

    Your number

  6. Alphabet, research and development expense as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 15.9%.

    On the Docket Pendingth087-t06Pre-specified

    Question
    Did the research and development expense of Alphabet for the quarter ended September 30, 2026 come to at or above 15.9% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Alphabet (Forms 10-Q and 10-K)
    Rule
    Divide the research and development expense 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 15.9%; no if it is below 15.9%.
    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 research and development expense as a share of revenue was 15.2% 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 14.3%, 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 23%.

    23%

    0 · no100 · yes

    The machinesClaude 33% · DeepSeek 15%

    Claude: 33%DeepSeek: 15%

    Your number

In The Buildout

The stories that take up this thesis, with their own Calls and tests.

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.