How DeepStack settles a question · Part 03

The debt behind the machines: 1999 capex, 2008 credit, or neither?

From Part 03: The Machines Will Work. The Debt Might Not.. 7 voices on 2 sides, attributed and dated; the DeepStack Call with its history; 5 dated tests that settle it; the results as they come in.

The DeepStack callSkeptical

The machines will earn. The credit stack is where this buildout breaks first.

Bearish on the credit · Conviction High (4/5)

Read the full call

Call history

  1. v1Bearish on the creditConviction High (4/5)First call, as published with the story. The conviction, the actions and the flips were set down for the site on October 7, 2026.

3 voices

1999 capex plus 2008 credit

  1. Michael Burry

    Nebius extended its server depreciation schedule from four years to five this year even though its own deal pricing implies roughly 50 percent annual decay; customers that would not quit depreciating are what extended the bust twenty-five years ago.

    Source

  2. Michael Burry

    Nebius's own disclosed terms show compute in backwardation: 20 to 25 million dollars per megawatt on one-to-three-year contracts against 40 to 50 million per megawatt for capacity of up to six months; buyers paying twice the rate for near-term use…

    Source

  3. Vitaliy Katsenelson

    AI profitability is elusive while open-source and Chinese models deliver comparable results at a fraction of the cost, so data-center pricing faces a race to the bottom exactly as fiber optics did after 1999; then the bankruptcies reveal who was…

    Source

4 voices

The borrowing is rational

  1. Market Sentiment

    Capacity is sold before it is built: AWS backlog reached 496 billion dollars, Microsoft closed June 2026 with 678 billion of commercial RPO up 84 percent year over year, Google Cloud's backlog hit 514 billion; CoreWeave's 2020-era A100s are fully…

    Source

  2. Market Sentiment

    Modeled bottom-up, 100 billion dollars of annual data-center capex sustains roughly 160 billion of steady-state annual revenue, a 1.6x revenue-to-capex multiple and about a 21 percent IRR, even assuming a five-year chip life and a one-third annual…

    Source

  3. Oguz Erkan

    Rental prices for six-year-old A100 GPUs are up more than 20 percent year-to-date, a direct read on AI model providers' willingness to pay and, by extension, end-customer demand.

    Source

  4. Variant Perception

    Still risk-on, semis capital-scarce, July was a tradeable low.

3 voices

Voices that cross sides

  1. JunkBondInvestor

    Three endings fit today's numbers: squeeze, triumph, or spreads just cheapen. Unsortable yet.

  2. Oguz Erkan

    His own tripwire: hyperscaler free cash flow may go negative in 2027; leverage on speculative returns repels.

    Source

  3. Market Sentiment

    His own bar: the spend needs about $960B of cloud revenue; the run rate is about $380B, growing 40% or more.

    Source

Where they cross

Who answers whom.

The specific points where one voice meets another: the same evidence read two ways, or the same mechanism with a different sign.

What settles it

The dated tests.

The tests the story set, each with its specification where the owner has written one; on the Docket, resolved on the Results page.

  1. Short-term premiums collapsing toward long-term rates would weaken the backwardation warning; persistence would make it harder to dismiss.

    Part 03: The Machines Will Work. The Debt Might Not. Pendingp03-t01

    Question
    Did Nebius's September-quarter results, in the release, the shareholder letter or on the call, still price compute capacity of six months or less at a premium to its one-to-three-year contracts?
    Source
    Nebius Group, quarterly results furnished to SEC EDGAR (Form 6-K)
    Rule
    Read the price per megawatt (or per GPU-hour) Nebius gives for terms of six months or less and for one-to-three-year terms; a short-term price at or above 1.5 times the long-term price resolves confirmed (the premium persists), a ratio below 1.5 resolves refuted (the premium collapsed toward long-term rates).
    If the source is late, revised or silent
    If the results carry no term pricing, or Nebius reports after November 30, 2026, the test is inconclusive, reviewed by the owner.
  2. Next H100 batches

    Re-lets near 95 percent support durable value; faster decay would expose a gap in the economics.

    Part 03: The Machines Will Work. The Debt Might Not. Pendingp03-t02

    Specification pending

  3. Resilient hyperscaler free cash flow would ease leverage fears; negative cash flow would validate Erkan’s tripwire.

    Part 03: The Machines Will Work. The Debt Might Not. Pendingp03-t03

    Specification pending

  4. The next $50B-$60B issuance will show whether new-issue concessions keep widening or start to normalize.

    Part 03: The Machines Will Work. The Debt Might Not. Inconclusivep03-t04

    Question
    Did investment-grade borrowers pay narrower new-issue concessions on the $50 billion to $60 billion of paper priced in the four weeks after Labor Day 2026 (September 8 to October 2) than on August's deals?
    Source
    SIFMA, US corporate bond issuance statistics (September 2026)
    Rule
    Yes if September's investment-grade issuance reached $50 billion and the average new-issue concession on those deals, as the dealer figures cited in JunkBondInvestor's Credit Weekly report it, was below the double-digit basis points paid in August; the test resolves confirmed on yes and refuted if concessions held or widened.
    If the source is late, revised or silent
    If no public concession figure covers the window, or issuance fell short of $50 billion, the test is inconclusive, reviewed by the owner.
  5. When the streak ends

    The 182-day run ending would test whether unusual market calm is breaking; repetition would extend the signal.

    Part 03: The Machines Will Work. The Debt Might Not. Pendingp03-t05

    Specification pending

The test came in

The test did not settle it.

Every result is listed on the Results page.

Inconclusive

· Part 03: The Machines Will Work. The Debt Might Not. · Test p03-t04

Inconclusive: no public figure covers the September concessions

The test. The next $50B-$60B issuance will show whether new-issue concessions keep widening or start to normalize.

The test asked whether investment-grade borrowers paid narrower new-issue concessions on the $50 billion to $60 billion of paper priced in the four weeks after Labor Day 2026, September 8 to October 2, than on August's deals. The specification names SIFMA's issuance statistics as the public source and the dealer figures cited in JunkBondInvestor's Credit Weekly for the concession reading.

No public statistic reports an average new-issue concession for the window, and the dealer figures are not published as a series. Under the specification's own ambiguity rule the test is inconclusive. The Call stands as written.

Average new-issue concession on investment-grade deals priced Sept. 8 to Oct. 2, 2026
No public figure; dealer estimates are not published as a series
SIFMA, US corporate bond issuance statistics (September 2026) · Oct 9, 2026

The DeepStack call. Conviction unchanged; the Call stands as written.

Sources (3)
  1. Capitalist Letters - Oguz Erkan’s compute framework
  2. Market Sentiment - capex sustainability model
  3. JunkBondInvestor - Credit Weekly

Market data are as of the dates cited. Voices are paraphrased from public writing and attributed by name.

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. Every question, one page each.