Thesis · AI Financing, Credit & Capital Structure
Will a US-listed GPU cloud operator record a material GPU impairment or extend GPU useful life between 2026-11-15 and 2027-12-31?
At least one major listed neocloud will record a material impairment or accelerated depreciation charge on prior-generation GPU fleets, confirming that collateral value decays faster than the debt financing it.
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
GPU collateral loses value faster than the loan
Annual Nvidia generational cadence (Hopper to Blackwell to Rubin) compresses rental rates and utilization on older cohorts while six-year depreciation schedules and multi-year debt amortize slowly. A non-cash impairment or useful-life revision appears in audited filings, repricing GPU-backed collateral and tightening advance rates on new GPU loans.
The case for no
GPU collateral holds value through recontracting
Older accelerator generations like A100 remain commercially relevant beyond the four to five year useful-life assumptions embedded in depreciation schedules, so residual values and second-contract economics are better than models assume. Nvidia GPUs are broadly adopted, transferable between customers and operators, and life-extended by software, so third-party institutional capital is willing to underwrite AI compute projects at scale. This side holds if operators sign second or third contracts on A100/H100 fleets at rates implying residual value above book, with no material impairment or useful-life revision through 2027.
4 voices
GPU collateral loses value faster than the loan
When GPU collateral enters a liquidation cascade, retention rates fall sharply below underwriting assumptions. Nvidia's percentage-based backstop grows in absolute dollars but remains a fixed fraction of the widening shortfall, leaving bond holders…
CoreWeave depreciates technology equipment over six years, but Nvidia releases new generations yearly. Older GPUs decline in rental value and utilization as newer systems deliver more performance per watt and per dollar. Equipment losing value…
Alpha In Academia
NVIDIA funding OpenAI, OpenAI committing to Oracle capacity, and miners financing GPU colocation creates a circuit where each participant's funding depends on the next. A funding freeze blocks hardware refresh, the older machines become inefficient…
Asset-life mismatch in off-balance-sheet AI financing
3 voices
GPU collateral holds value through recontracting
CoreWeave disclosed an A100 contract running into 2029 on architecture from 2020. If hardware generations recontracts rather than being written off, and if this behaviour generalizes, then payback maths for neocloud fleets improve and the…
Because Nvidia compute retains rental value and can be redeployed to another operator if one fails, lenders and asset managers can treat it as an asset class with residual value, which unlocks large pools of institutional money for gigawatt-scale…
Whether Nebius' Meta deal lands closer to the bullish or bearish case depends primarily on GPU value at the end of the five-year contract; observed A100 resale and rental pricing suggests meaningful residual value.
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.
NVIDIA, long-term debt as a share of the quarter's revenue, quarter ended October 25, 2026: whether it comes in at or above 32%.
On the Docket Pendingth084-t04Pre-specified
- Question
- Did the long-term debt of Nvidia at the end of the quarter ended October 25, 2026 come to at or above 32.0% of its revenue for the quarter?
- Source
- SEC EDGAR, XBRL company facts of Nvidia (Forms 10-Q and 10-K)
- Rule
- Divide the long-term debt that Nvidia reports at the end of the quarter ended October 25, 2026 by the revenue it reports for the quarter ended October 25, 2026, in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 32.0%; no if it is below 32.0%.
- If the source is late, revised or silent
- If Nvidia has not filed by December 11, 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
- NVIDIA's long-term debt as a share of the quarter's revenue was 34.7% at the end of the quarter ended July 26, 2026, by its filing on SEC EDGAR.
- Outside view
- A year earlier, at the end of the quarter ended July 27, 2025, the same line read 18.1%, 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 62%.
0 · no100 · yes
The machinesClaude 14% · DeepSeek 3%
The ladder 3 rungs on one reading, Nov 18, 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. Above 32%Call 62% · Claude 14% · DeepSeek 3%
Above 36%Call 44% · Claude 14% · DeepSeek 3%
Above 40%Call 26% · Claude 6% · DeepSeek 3%
0 · no50100 · yes
Your numberSign in by email to forecast
NVIDIA, long-term debt as a share of the quarter's revenue, quarter ended October 25, 2026: whether it comes in at or above 36%.
On the Docket Pendingth084-t05Pre-specified
- Question
- Did the long-term debt of Nvidia at the end of the quarter ended October 25, 2026 come to at or above 36.0% of its revenue for the quarter?
- Source
- SEC EDGAR, XBRL company facts of Nvidia (Forms 10-Q and 10-K)
- Rule
- Divide the long-term debt that Nvidia reports at the end of the quarter ended October 25, 2026 by the revenue it reports for the quarter ended October 25, 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 Nvidia has not filed by December 11, 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
- NVIDIA's long-term debt as a share of the quarter's revenue was 34.7% at the end of the quarter ended July 26, 2026, by its filing on SEC EDGAR.
- Outside view
- A year earlier, at the end of the quarter ended July 27, 2025, the same line read 18.1%, 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 44%.
0 · no100 · yes
The machinesClaude 14% · DeepSeek 3%
Your numberSign in by email to forecast
NVIDIA, long-term debt as a share of the quarter's revenue, quarter ended October 25, 2026: whether it comes in at or above 40%.
On the Docket Pendingth084-t06Pre-specified
- Question
- Did the long-term debt of Nvidia at the end of the quarter ended October 25, 2026 come to at or above 40.0% of its revenue for the quarter?
- Source
- SEC EDGAR, XBRL company facts of Nvidia (Forms 10-Q and 10-K)
- Rule
- Divide the long-term debt that Nvidia reports at the end of the quarter ended October 25, 2026 by the revenue it reports for the quarter ended October 25, 2026, in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 40.0%; no if it is below 40.0%.
- If the source is late, revised or silent
- If Nvidia has not filed by December 11, 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
- NVIDIA's long-term debt as a share of the quarter's revenue was 34.7% at the end of the quarter ended July 26, 2026, by its filing on SEC EDGAR.
- Outside view
- A year earlier, at the end of the quarter ended July 27, 2025, the same line read 18.1%, 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 26%.
0 · no100 · yes
The machinesClaude 6% · DeepSeek 3%
Your numberSign in by email to forecast
Oracle, interest expense as a share of revenue, quarter ended November 30, 2026: whether it comes in at or above 7.6%.
On the Docket Pendingth084-t01Pre-specified
- Question
- Did the interest expense of Oracle for the quarter ended November 30, 2026 come to at or above 7.6% of its revenue?
- Source
- SEC EDGAR, XBRL company facts of Oracle (Forms 10-Q and 10-K)
- Rule
- Divide the interest expense by the revenue that Oracle reports for the quarter ended November 30, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 7.6%; no if it is below 7.6%.
- If the source is late, revised or silent
- If Oracle has not filed by January 18, 2027, 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
- Oracle's interest expense as a share of revenue was 7.4% for the quarter ended August 31, 2026, by its filing on SEC EDGAR.
- Outside view
- A year earlier, for the quarter ended August 31, 2025, the same line read 6.2%, 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%.
0 · no100 · yes
The machinesClaude 72% · DeepSeek 64%
The ladder 3 rungs on one reading, Dec 14, 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. Above 7.6%Call 72% · Claude 72% · DeepSeek 64%
Above 8.1%Call 48% · Claude 63% · DeepSeek 56%
Above 8.6%Call 25% · Claude 48% · DeepSeek 15%
0 · no50100 · yes
Your numberSign in by email to forecast
Oracle, interest expense as a share of revenue, quarter ended November 30, 2026: whether it comes in at or above 8.1%.
On the Docket Pendingth084-t02Pre-specified
- Question
- Did the interest expense of Oracle for the quarter ended November 30, 2026 come to at or above 8.1% of its revenue?
- Source
- SEC EDGAR, XBRL company facts of Oracle (Forms 10-Q and 10-K)
- Rule
- Divide the interest expense by the revenue that Oracle reports for the quarter ended November 30, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 8.1%; no if it is below 8.1%.
- If the source is late, revised or silent
- If Oracle has not filed by January 18, 2027, 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
- Oracle's interest expense as a share of revenue was 7.4% for the quarter ended August 31, 2026, by its filing on SEC EDGAR.
- Outside view
- A year earlier, for the quarter ended August 31, 2025, the same line read 6.2%, 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%.
0 · no100 · yes
The machinesClaude 63% · DeepSeek 56%
Your numberSign in by email to forecast
Oracle, interest expense as a share of revenue, quarter ended November 30, 2026: whether it comes in at or above 8.6%.
On the Docket Pendingth084-t03Pre-specified
- Question
- Did the interest expense of Oracle for the quarter ended November 30, 2026 come to at or above 8.6% of its revenue?
- Source
- SEC EDGAR, XBRL company facts of Oracle (Forms 10-Q and 10-K)
- Rule
- Divide the interest expense by the revenue that Oracle reports for the quarter ended November 30, 2026 in its 10-Q or 10-K filing on SEC EDGAR. The answer is yes if the ratio is at or above 8.6%; no if it is below 8.6%.
- If the source is late, revised or silent
- If Oracle has not filed by January 18, 2027, 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
- Oracle's interest expense as a share of revenue was 7.4% for the quarter ended August 31, 2026, by its filing on SEC EDGAR.
- Outside view
- A year earlier, for the quarter ended August 31, 2025, the same line read 6.2%, 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 25%.
0 · no100 · yes
The machinesClaude 48% · DeepSeek 15%
Your numberSign in by email to forecast
In The Buildout
The Parts on the same question.
The stories that take up this thesis, with their own Calls and tests.
Part 01: Nvidia’s Demand Is Real. Its Financing Is Fragile.
Nvidia’s demand looks real today, but private credit, slow-to-build working compute and fast obsolescence make the revenue recurrence less secure than the backlog suggests.
Part 03: The Machines Will Work. The Debt Might Not.
AI demand is holding up, but vendor guarantees, short-term premiums and widening credit concessions expose a fragile financing chain.
Part 08: AI Lenders Are Financing the End of Scarcity
AI capacity is scarce, but lenders are financing the supply that could erase that scarcity before hyperscaler cash flow catches up.
Part 17: The Guarantee Has No Date
The leases the AI builders signed reach their balance sheets on dates their filings print. The guarantees behind the projects carry little or no recorded liability, and four filings between December 2026 and February 2027 test whether the risk off the books is still rising.
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.