How DeepStack settles a question · Part 01

Is AI demand real?

Is the marginal buyer of AI compute real?

From Part 01: Nvidia’s Demand Is Real. Its Financing Is Fragile.. 5 voices on 2 sides, attributed and dated; the DeepStack Call with its history; 3 dated tests that settle it; the results as they come in.

The mechanism

Demand for Nvidia's chips is real today: Google paid roughly double the market rate for Nvidia capacity while cheaper TPUs were available, a costly revealed preference rather than paper demand. What makes the revenue less secure than the order book implies is how it is financed. Private credit funded the compute platform, a $38 billion Apollo transaction in late May and a $35 billion first tranche from Apollo and Blackstone, so part of the marginal buyer's spending rests on lenders rather than on end customers. Between the ordered chips and working compute sits a power bottleneck, and depreciation schedules may lag obsolescence. The Call therefore prices the revenue as financed, not recurring: once the bottleneck clears, the question is whether end customers keep absorbing capacity at market-rate financing.

Approved by Marco Cauduro on October 10, 2026.

The DeepStack callSplit

Nvidia’s orders are real. Price its revenue as financed, not recurring.

Bull on demand. Bear on the financing. · Conviction Moderate (3/5)

Our Call takes neither side alone.

Read the full call

Call history2 versions · last changed
  1. v2Bull on demand. Bear on the financing.Conviction Moderate (3/5)SynthesisSide named by the owner.
  2. v1Bull on demand. Bear on the financing.Conviction Moderate (3/5)First call, as published with the story. The conviction, the actions and the flips were set down for the site on October 7, 2026.

2 voices

Manufactured

  1. Michael Burry

    Much and possibly most of current and future AI compute demand is not end-customer demand; the majority of future revenues are financed through a circular off-balance-sheet arrangement.

    Source

  2. Michael Burry

    NVIDIA's revenue growth is only mostly real because circular financing sustains marginal AI compute demand above market-rate levels, and resolution of the power bottleneck will reduce revenue recurrence and compress the scarcity premium in margins…

    Source

2 voices

Insatiable

  1. Kris

    The Google SpaceX GPU rental is bullish for Nvidia because there are still not enough GPUs to quench demand; Nvidia ramps production continuously but cannot follow demand.

    Source

  2. Invest in Assets

    A visibility window that extends by a full year across the customer base is much harder to manufacture than a single-customer beat and signals that AI demand is structural rather than lumpy.

    Source

3 voices

Voices that cross sides

  1. Longriver

    Long since 2025, and author of the bear case's most honest sentence.

  2. Kris

    Bullish on demand, and endorses the GFC earnings-quality parallel.

    Source

  3. NZS Capital

    Exited on the bear case; re-entered when the price paid for the risk.

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.

Alternative explanations

Three other readings.

What would have to be true for the Call to be wrong, as the other side argues it.

  1. The demand is structural, not financed: a visibility window that extends a full year across the customer base is harder to manufacture than one customer's beat (the case Invest in Assets makes).
  2. Most of the marginal demand is not end-customer demand at all: future revenue is financed through a circular off-balance-sheet arrangement, so the orders overstate what customers will pay (the case Michael Burry makes).
  3. The shortage is real and the financing is beside the point: Nvidia ramps production continuously and still cannot follow demand, so recurrence holds as long as supply lags (the case Kris makes).

What settles it

The dated tests.

The tests the story set, each with its specification where the owner has written one and the piece of the argument it verifies; on the Docket, resolved on the Results page.

  1. When the bottleneck clears

    If end demand survives market-rate financing, the financed-demand warning weakens; if recurrence falls, private-credit risk moves to the front.

    Part 01: Nvidia’s Demand Is Real. Its Financing Is Fragile. Pendingp01-t01Open questionChecks the financed-demand claim

    Specification pending

  2. As capacity ramps

    Order cancellations would erode year-long customer visibility; continued bookings would support the scarcity case.

    Part 01: Nvidia’s Demand Is Real. Its Financing Is Fragile. Pendingp01-t02Open questionChecks the customer-visibility claim

    Specification pending

  3. Next depreciation tests

    Schedules that lag obsolescence flatter earnings; schedules that catch up would strengthen the bears.

    Part 01: Nvidia’s Demand Is Real. Its Financing Is Fragile. Pendingp01-t03Open questionChecks the depreciation-lag claim

    Specification pending

What changed

The record of Part 01.

The Part’s publication, every change to the Call, every result that came in and the next dated test, oldest first.

  1. PublishedPart 01 published; the Call: Bull on demand. Bear on the financing..
  2. Call updatedThe Call: Bull on demand. Bear on the financing.. Nvidia’s orders are real. Price its revenue as financed, not recurring. Our Call takes neither side alone. by Marco Cauduro
Sources (4)
  1. Invest In Assets - Broadcom AI financing and orders
  2. Potential Multibaggers - Google rents Nvidia GPUs
  3. Fiscal.ai - Longriver Q2 2026 letter
  4. Fiscal.ai - NZS Capital Q2 2026 letter

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.