Thesis · Memory & Storage

Will memory makers keep earning scarcity profits from AI demand through 2027?

Micron's revenue for the quarter ended September 3, 2026 was $54.2 billion, against $11.3 billion a year earlier. The tests read Micron's next quarter and Korea's chip exports, most of which are memory.

8 dated tests · 0 resolved · next on

The tests More theses in Memory & Storage

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

Scarcity holds through 2027

AI accelerator demand requires HBM bits that only three suppliers can produce, with fab and stacking capacity locked by multi-year lead times and sold out through 2026-2027. HBM and conventional DRAM ASPs stay elevated, and memory makers capture scarcity rent in the form of oligopoly-level operating margins.

The case for no

Supply catches up

Memory margins staying high while the replacement cost of a memory business is small relative to hyperscaler annual capex invites new entrants, less memory-dependent designs, and eventual price wars that crash memory profits. Immature HBM front-end and stacking yields sit against commodity DRAM, which pays more per wafer and yields over 3x the bits, so minimal HBM wafer allocation and a near-total commodity DRAM revenue mix persist. It would show first if DRAM contract prices decline sequentially for two or more consecutive quarters while bit shipments rise.

6 voices

Scarcity holds through 2027

  1. Tae Kim

    The demand-supply imbalance in HBM memory creates a classical Economics 101 scenario where constrained supply meets accelerating demand. Customers requesting 5x–6x more capacity than current expansion plans can supply creates a pricing lever that…

    Source

  2. Harris Associates Oakmark

    AI-driven demand for high-bandwidth memory tightens supply relative to customer orders, supporting pricing power and operating profit for leading suppliers like Samsung.

    Source

  3. Aristotle Capital Management

    Artificial intelligence workloads have fundamentally shifted memory consumption patterns upward, and the prioritization of high-bandwidth memory for AI accelerators constrains conventional DRAM and NAND supply relative to demand, creating…

    Source

  4. a16z New Media

    When demand for memory accelerates faster than capacity can be added (taking years to build), and manufacturers prioritize higher-margin HBM while hyperscalers lock in long-term contracts to secure supply, prices rise dramatically and historically…

    Source

  5. Sands Capital

    Growing AI demand, particularly for high-bandwidth memory in AI servers and for conventional memory in AI inference workloads, strains current supply. Memory industry consolidation into an oligopoly limits capacity expansion speed, allowing…

    Source

  6. Harding Loevner

    Heavy AI server and data center buildout requires customized, high-performance memory products in complex 3D architectures, increasing barriers to entry and pricing power. Simultaneously, the memory wall constraint necessitates high-bandwidth…

    Source

2 voices

Supply catches up

  1. Andrew Walker

    When memory and semiconductor prices rise significantly above cost due to supply constraints, economic incentives drive multiple paths to increased supply, eventually breaking the cycle and reducing margins.

    Source

  2. Daniel Romero

    Contracts trade peak pricing for duration: the premium per bit compresses but earnings visibility lengthens, which lowers cyclicality risk and justifies a higher multiple on memory names like Micron, further helped by Micron regaining ~10% share in…

    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. Sandisk, operating cash flow as a share of revenue, quarter ended October 2, 2026: whether it comes in at or above 70%.

    On the Docket Pendingth002-t06Pre-specified

    Question
    Did the net cash from operating activities of Sandisk for the quarter ended October 2, 2026 come to at or above 70.0% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Sandisk (Forms 10-Q and 10-K)
    Rule
    Divide the net cash from operating activities by the revenue that Sandisk reports for the quarter ended October 2, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 70.0%; no if it is below 70.0%.
    If the source is late, revised or silent
    If Sandisk has not filed by November 18, 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
    Sandisk's operating cash flow as a share of revenue was 79.5% for the quarter ended July 3, 2026, by its filing on SEC EDGAR.
    Outside view
    A year earlier, for the quarter ended June 27, 2025, the same line read 4.9%, 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 68%.

    68%

    0 · no100 · yes

    The machinesClaude 40% · DeepSeek 68% · GPT 18%

    Claude: 40%DeepSeek: 68%GPT: 18%
    The ladder 3 rungs on one reading, Oct 29, 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 70%Call 68% · Claude 40% · DeepSeek 68% · GPT 18%

      68%

    2. Above 77%Call 49% · Claude 22% · DeepSeek 42% · GPT 8%

      49%

    3. Above 84%Call 22% · Claude 19% · DeepSeek 7% · GPT 22%

      22%

    0 · no50100 · yes

    Your number

  2. Sandisk, operating cash flow as a share of revenue, quarter ended October 2, 2026: whether it comes in at or above 77%.

    On the Docket Pendingth002-t07Pre-specified

    Question
    Did the net cash from operating activities of Sandisk for the quarter ended October 2, 2026 come to at or above 77.0% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Sandisk (Forms 10-Q and 10-K)
    Rule
    Divide the net cash from operating activities by the revenue that Sandisk reports for the quarter ended October 2, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 77.0%; no if it is below 77.0%.
    If the source is late, revised or silent
    If Sandisk has not filed by November 18, 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
    Sandisk's operating cash flow as a share of revenue was 79.5% for the quarter ended July 3, 2026, by its filing on SEC EDGAR.
    Outside view
    A year earlier, for the quarter ended June 27, 2025, the same line read 4.9%, 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 49%.

    49%

    0 · no100 · yes

    The machinesClaude 22% · DeepSeek 42% · GPT 8%

    Claude: 22%DeepSeek: 42%GPT: 8%

    Your number

  3. Sandisk, operating cash flow as a share of revenue, quarter ended October 2, 2026: whether it comes in at or above 84%.

    On the Docket Pendingth002-t08Pre-specified

    Question
    Did the net cash from operating activities of Sandisk for the quarter ended October 2, 2026 come to at or above 84.0% of its revenue?
    Source
    SEC EDGAR, XBRL company facts of Sandisk (Forms 10-Q and 10-K)
    Rule
    Divide the net cash from operating activities by the revenue that Sandisk reports for the quarter ended October 2, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 84.0%; no if it is below 84.0%.
    If the source is late, revised or silent
    If Sandisk has not filed by November 18, 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
    Sandisk's operating cash flow as a share of revenue was 79.5% for the quarter ended July 3, 2026, by its filing on SEC EDGAR.
    Outside view
    A year earlier, for the quarter ended June 27, 2025, the same line read 4.9%, 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 22%.

    22%

    0 · no100 · yes

    The machinesClaude 19% · DeepSeek 7% · GPT 22%

    Claude: 19%DeepSeek: 7%GPT: 22%

    Your number

  4. Korea's semiconductor exports for the first 20 days of November 2026: whether they grow at or above 100.0% from a year earlier.

    On the Docket Pendingth002-t04Open question

    Specification pending

    The machinesClaude 55% · DeepSeek 78% · GPT 33%

    Claude: 55%DeepSeek: 78%GPT: 33%

    Your number

  5. Korea's semiconductor exports for November 2026: whether they grow at or above 100.0% from a year earlier.

    On the Docket Pendingth002-t05Pre-specified

    Question
    Did Korea's semiconductor exports in November 2026 grow at or above 100.0% from a year earlier?
    Source
    Ministry of Trade, Industry and Resources (MOTIR), monthly exports and imports release
    Rule
    Read the semiconductor line of the trade ministry's (MOTIR) release on November 2026 exports and imports, published December 1, 2026. The answer is yes if semiconductor exports are up at or above 100.0% from a year earlier; no if they are below 100.0%.
    If the source is late, revised or silent
    The figure as first published counts; if the release is not out within seven days of the date, the test is inconclusive, and the Result says what was read and what was missing.
    Base rate
    Micron, the largest memory maker that files with the SEC, reported revenue almost five times its year-earlier level for the quarter ended September 3, 2026.
    Outside view
    Korea's chip exports are mostly memory, so they track memory prices more than unit volumes.
    What would change the number
    The customs release for the first 20 days of November, due November 23, would move the number either way.

    The Call, before the dateExpects no. Gives the yes answer 42%.

    42%

    0 · no100 · yes

    The machinesClaude 42% · DeepSeek 80% · GPT 38%

    Claude: 42%DeepSeek: 80%GPT: 38%

    Your number

  6. Micron, quarter ended December 3, 2026: whether revenue grows at or above 250.0% from a year earlier.

    On the Docket Pendingth002-t01Open question

    Specification pending

    The machinesClaude 63% · DeepSeek 90% · GPT 82%

    Claude: 63%DeepSeek: 90%GPT: 82%

    Your number

  7. Micron, quarter ended December 3, 2026: whether revenue grows at or above 300.0% from a year earlier.

    On the Docket Pendingth002-t02Open question

    Specification pending

    The machinesClaude 72% · DeepSeek 88% · GPT 20%

    Claude: 72%DeepSeek: 88%GPT: 20%

    Your number

  8. Micron, quarter ended December 3, 2026: whether revenue grows at or above 350.0% from a year earlier.

    On the Docket Pendingth002-t03Open question

    Specification pending

    The machinesClaude 10% · DeepSeek 45% · GPT 44%

    Claude: 10%DeepSeek: 45%GPT: 44%

    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.