Who is on each side · Part 16
How much of the money buying AI chips was lent, guaranteed or invested by the companies selling them?
20 voices who disagree about The Seller Holds the Paper. Each voice is attributed and paraphrased, dated where the date is known, with a link where one exists.
The DeepStack callSkeptical
Nvidia's receivables ran 9% to 16% of sales above its own collection pace by July 26, 2026; five accounts owed 70% of them.
Our Call takes the side “The seller funds its demand”.
Three voices per side
Where the argument stands.
DeepStack’s pick of the voices that carry each side, in its own words, and the case against each side, chosen by hand with the reason given. Every voice the record holds is listed below; the grid and the CSV carry them all.
3 of 10 voices
The loan is a bridge
Third-party money from Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR was underwritten independently and replaces Nvidia's balance sheet rather than stretching it.
Nvidia acts as a central bank for AI, supplying credit where private lenders will not, because small buyers can pay for GPUs but cannot show a lender a credit rating.
Customers outside the traditional hyperscalers, from neoclouds to sovereign clouds, produced 59% of Nvidia's data center growth in the quarter to July 26, 2026.
3 of 9 voices
The seller funds its demand
Nvidia's revenue growth is mostly, not wholly, real, because circular financing holds marginal demand for compute above market rates.
CoreWeave exists mainly to create demand for Nvidia's GPUs: Nvidia is its only GPU supplier and also an investor, customer and sponsor.
Distillate Capital Partners LLC
Hyperscalers and Nvidia committed about $180 billion to OpenAI and Anthropic, which pledged about $1.1 trillion of compute spending back.
3 of 10 voices
Two labs carry the chain
OpenAI and Anthropic supplied an estimated 60% to 80% of the AI revenue at Amazon, Google and Microsoft and, citing The Information, about half of the hyperscalers' revenue backlog.
Two customers supplied nearly a quarter of Anthropic's 2025 revenue and 47% of its sales ran through Amazon and Google, a concentration a company seeking a $2 trillion value cannot afford.
The two labs' combined run-rate reached $105 billion in July 2026, $65 billion of it Anthropic's, nowhere near what the buildout modeled for 2027 requires.
- ARR
- Annualized recurring revenue: a month's or a quarter's revenue scaled to a year, not audited revenue.
- TPU
- Tensor Processing Unit, the AI accelerator Google designs with Broadcom.
- neocloud
- A cloud company that rents out AI computing, such as CoreWeave or Nebius, as distinct from the large hyperscalers.
Every voice, as the record holds it · 37 claims · 20 voices · 1 question
4 claims set aside by the editor stay in the grid and the CSV.
10 voices
The loan is a bridge
On circularity fears: recent third-party capital from Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR was independently underwritten apart from one another and replaces Nvidia's balance sheet rather than merely extending it, which…
Hyperscalers retained one structural advantage over neoclouds: investment-grade balance sheets that let them fund speculative capacity ahead of demand and rent it out at higher spot rates, whereas neoclouds could only get financing for hardware…
They characterize Nvidia as acting like the central bank of AI, supplying support where the private lending system will not, because small buyers can pay for GPUs but cannot offer creditors a credit rating, and Nvidia does not want demand…
NVIDIA's non-traditional hyperscaler customers (neoclouds, sovereign AI clouds, enterprises, and industrial customers) generated 59% of data center growth this quarter and are expected to drive Q3 growth primarily, because they rely more heavily on…
NVIDIA has invested approximately 50 billion dollars in AI labs and is working with firms like Blackstone and BlackRock to establish financing platforms targeting over 500 billion dollars of third-party infrastructure capital, with roughly…
Tae Kim worried about the sequence of information releases: had news of a slowdown in Anthropic's ARR growth from share losses to OpenAI come out before OpenAI's higher ARR figure, it could have spooked markets; now that the $70 billion number is…
Morgan Stanley's investor meetings with Nvidia leadership show accelerating growth rates even at 100 billion dollars revenue per quarter, strong demand for the Vera Rubin product cycle over the next 12 months, memory shortages lasting several…
Two examples of prior-generation demand: an AI-native scientific discovery company on NVIDIA A100s with a renewal term extending into 2029 and a 3-year contract signed in line with typical 1-year terms, and a leading early-stage research lab on…
Reported revenue acceleration at the two leading model companies — an annualised run-rate said to go from USD 1 billion at end-2024 to USD 9 billion at end-2025 to almost USD 50 billion in May this year at Anthropic, with a similar trajectory at…
The latest US AI lab models, GPT-6 Astra and Claude Fable 5.1, are priced at $10/$50 per million tokens, a similar band to GPT-4 ($30/$60) and Opus 3 ($15/$75) at their respective releases, leading Market Sentiment to conclude there is a separate…
7 voices
The seller funds its demand
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…
Michael Burry characterises Nvidia's November 2025 defence that its strategic investments in customers were a minimal portion of revenue as evasive, because it referred only to revenue already booked rather than to future revenue.
CoreWeave exists primarily to generate demand for Nvidia's GPUs, as evidenced by the fact that Nvidia is the sole GPU supplier and also an investor, customer, strategic sponsor, and that cash equipment purchases alone equaled approximately 84…
Nvidia, acting as supplier, investor, financier, and deal broker, introduced Nebius to four major customers worth over $1 billion each in Q1, demonstrating Nebius's dependence on Nvidia for customer acquisition.
Distillate Capital Partners LLC
Circular financing distorts the picture: hyperscalers and NVIDIA have committed roughly $180 billion to OpenAI and Anthropic, which pledge about $1.1 trillion of compute spending back to them against far smaller revenues, and $53 billion of…
Steve raises the concern that Nvidia had an insane quarter yet seventy percent of its receivables come from five accounts, and that much of the food chain depends on the health of Anthropic and OpenAI.
The AI trade's foundations are shaky because roughly seventy percent of the hyperscalers' AI revenue traces back to just two customers, OpenAI and Anthropic, both of which are burning enormous amounts of cash, so if either stumbles the entire…
8 voices
Two labs carry the chain
NVIDIA's revenue growth is concentrated among a small number of customers, with one customer accounting for 16% of quarterly revenue and five customers making up 70% of accounts receivable.
The visible, Nvidia-funded circular financing in AI (Nvidia investing in or lending to companies that then buy Nvidia chips) is out in the open and funded from Nvidia's own cash flow, and is not what disturbs the speaker; what disturbs him is the…
Back-end AI can only look profitable while unusually rich and irrational customers burn billions with no return; Ed Elson estimates OpenAI and Anthropic alone account for 60-80% of AI revenues at Amazon, Google and Microsoft and, per The…
Anthropic's expected ~$2 trillion IPO valuation is not justified by its numbers, given that it would need $1.2 trillion in annual revenue within 10 years (per Aswath Damodaran) to justify that valuation while the entire current AI products and…
Anthropic's customer concentration is a risk for its $2 trillion valuation: nearly a quarter of its 2025 revenue came from two customers and 47% of sales ran through Amazon and Google, which is an acceptable problem for a small company but not for…
OpenAI's combined annualized revenue with Anthropic reached $105 billion (OpenAI $40 billion, Anthropic $65 billion) in July 2026, which is nowhere near the figure needed to support the modeled 2027 buildout, and this understates the true gap since…
With Anthropic and OpenAI delaying their IPO dates, there are increasing signs that the public equity market is running out of marginal buyers.
Neoclouds resemble aircraft lessors stripped of the two features that made lessors safe: their GPU fleet depreciates far faster than an airframe as the next generation arrives, and their customer book is often concentrated in a handful of names.
8 voices
Voices that cross sides
NVIDIA has extended payment terms for investment-grade customers from 90 days to a full year, lengthening days sales outstanding by 14 days.
Anthropic and OpenAI together account for roughly 70% of hyperscaler AI revenue and 25-35% of total cloud revenue, with Microsoft, Amazon, and Google's growth substantially dependent on these two companies.
The bull argument for the financing structure holds that Nvidia is not lending money but providing a revenue guarantee on compute capacity with fundamentally different risk, the compute retains value because it is redeployable, and the customers…
The core criticism of Nvidia's financing structure is that revenue guarantees on compute capacity becoming cheaper every year means the floor Nvidia is guaranteeing could end up above market rates, and if token prices keep falling, the neoclouds'…
An Anthropic IPO might be a safer bet than semiconductor or neo-cloud stocks in this climate, because the U.S. needs its frontier model maker and enterprise AI leader, so if Anthropic were ever in serious financial trouble, Nvidia, Amazon and…
Anthropic is far ahead of OpenAI pre-IPO, both in model quality and ARR, and that lead should only accelerate with IPO proceeds heading into its November IPO.
Nvidia's 70% growth guide is restricted by memory chip supply rather than demand, so Nvidia's revenue line is set to stop being a read on AI demand and instead become a signal for component availability.
Nvidia is financing neoclouds into existence because that is what the market wants and what benefits Nvidia the most.
Notes
Not a disagreement.
Where two voices only seem to differ, or where the thread runs back to an earlier Part.
The thread
Part 1 called Nvidia's demand real and its financing fragile; Part 11 found the longer terms. Part 16 measures where the risk went.
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.
Mia Silverio contradicts Michael Spencer, on the evidence
Steve Eisman qualifies Tae Kim, on the mechanism
Michael Spencer qualifies Daniel Romero, on the evidence
Hugh Vuillier contradicts Tae Kim, on the evidence
Hugh Vuillier qualifies Mia Silverio, on the mechanism
Steve Eisman qualifies Daniel Romero, on the evidence
Michael Burry qualifies Steve Eisman, on the mechanism
Michael Spencer contradicts Steve Eisman, on the evidence
Steve Eisman qualifies Market Sentiment, on the evidence
Steve Eisman qualifies Tae Kim, on the evidence
Michael Spencer contradicts Tae Kim, on the evidence
Michael Spencer qualifies Michael Burry, on the evidence
Michael Spencer contradicts Steve Eisman, on the mechanism
What would settle it
The dated tests.
The same tests the story set, on the Docket; results land on the Results page.
Nvidia reports its quarter to October 25, 2026: whether days sales outstanding, receivables over revenue times 91, hold at or above 51.1, the average before July, after 59.6 in the quarter to July 26.
Nvidia reports: whether the guarantees note in its 10-Q puts total maximum exposure above $108.5 billion, the figure in the table for July 26, 2026.
Nvidia reports: whether its 10-Q puts AI cloud agreements, cloud services it commits to buy from its own customers, above the $36 billion of July 26, 2026.
Nvidia reports: whether the direct customers listed at 10% or more of receivables hold more than 70% together, after five held 70% on July 26, 2026.
- When filed
Anthropic's first public registration statement: its first audited revenue, and how much of it runs through investors that also sell it compute.
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.