This article summarizes publicly available information and expert opinions about GPU-backed lending and the AI infrastructure market. It is not investment advice. The views expressed by analysts and commentators are their own. Readers should verify all claims independently before making financial decisions.
That someone is CoreWeave, the publicly traded "neocloud" company that helped establish GPU-backed lending as a financing model in 2023. The model is straightforward: take out loans using Nvidia chips as collateral, buy more Nvidia chips, repeat. The first CoreWeave loan, led by hedge fund Magnetar Capital and Blackstone, carried about 14 percent interest. More private credit firms piled in. The GPU-backed debt market has since grown to tens of billions of dollars, and major banks including Goldman Sachs, JPMorgan Chase, and Wells Fargo have entered the space.
But Nvidia isn't just selling chips to these companies. It's also investing in them and signing contracts to rent capacity from them. According to Nvidia's most recent 10-Q, the company has committed $26 billion for "cloud compute services" through 2031. This obligation is disclosed in the filing's footnotes, as is standard for such commitments.
Jay Goldberg, an analyst at Seaport Research Partners (and one of the few on Wall Street with a sell rating on Nvidia), believes that number may represent backstop agreements with neoclouds. He points to timing that he views as notable: CoreWeave announced a $6.3 billion deal with Nvidia in September, and Nvidia's cloud compute commitments increased by $13 billion that same quarter. Nvidia has not confirmed any connection between these figures.
"Nvidia is buying demand here," Goldberg told The Verge, describing his interpretation of the arrangement.
CoreWeave's financials
CoreWeave reported revenue of $1.36 billion in Q3 2025, up 134 percent from a year earlier. Its revenue backlog reached $55.6 billion. The company went public in March at $40 per share and the stock has since climbed above $100.
The growth comes with costs. CoreWeave's Q3 SEC filing shows approximately $14 billion in total debt. Interest expense for the quarter was $310.6 million, more than triple what it was a year ago. The company posted a net loss of $110 million despite adjusted EBITDA margins above 60 percent.
The loans are secured by the GPUs themselves. Loan-to-value ratios vary widely—some deals reportedly exceed 100 percent—according to industry participants interviewed by The Verge. The terms depend on loan length, confidence in management teams, and other contract details. Ryan Little, senior managing director of equipment financing at Trinity Capital, has said that as long as companies are utilizing the compute, the chips will hold value.
Nvidia's involvement with CoreWeave extends beyond chip sales. In 2024, Nvidia was CoreWeave's second-largest customer. The chipmaker agreed to spend $1.3 billion over four years to rent capacity from CoreWeave, according to a report from The Information. Then in September came the $6.3 billion deal, which CoreWeave disclosed in an 8-K filing with the SEC.
Nvidia's 10-Q describes the cloud compute obligations as related to "R&D and DGX cloud offerings." Goldberg questions whether that fully explains the scale of the commitments. In his analysis, the numbers exceed what he would expect for typical R&D spending. He calculates that if those cloud commitments were reclassified as cost-of-goods-sold, Nvidia's margins would drop from 72 percent to 68 percent and earnings per share would fall from $6.28 to $5.97.
The depreciation question
All this debt is backed by chips that depreciate over time. Investors including Jim Chanos (the short-seller known for his early Enron call) and Michael Burry (known for his role in "The Big Short") have raised questions about how quickly these assets lose value.
CoreWeave depreciates its chips over approximately six years. Microsoft and Google use similar schedules. Chanos argues the real economic life may be closer to three years.
"If the chips last for three years, you have to depreciate a third of what you spend," Chanos explained. "That's the bet you have to make if you're a CoreWeave investor."
He has been direct about the potential implications: "Business models like the neoclouds, a lot of the AI companies themselves are just loss-making enterprises right now. You've gotta hope that changes, because if it doesn't, there's going to be debt defaults on these things."
Nvidia CEO Jensen Huang's public comments about chip generations have drawn attention. At the company's GTC conference in March, while promoting the newer Blackwell architecture, he remarked: "There are circumstances where Hopper is fine... That's the best thing I can say about Hopper." In earlier remarks, he had stated that once Blackwell ships in volume, "you couldn't give Hoppers away." Such comments, while typical of product launch messaging, have been cited by some analysts as relevant to depreciation discussions.
CoreWeave management has disputed bearish interpretations of chip longevity. On recent earnings calls, executives said the company successfully recontracts older chip capacity, including A100s and H100s, with contracts extending one to three years after initial terms. In Q3, CoreWeave reported that a customer with more than 10,000 H100s proactively recontracted at a price within 5 percent of the original agreement. The company's SEC filings do acknowledge that overestimating useful life represents a risk factor.
Risk perspectives
Sarah Bloom Raskin, former deputy secretary of the US Treasury, has commented on structural features of GPU-related debt. In an interview with The Verge, she reportedly noted that data centers are creating asset-backed securities and that GPU debt is spawning derivatives.
"They're like the derivatives we saw with the mortgages" before 2008, she told The Verge, according to the report. Raskin did not predict an imminent crisis but pointed to structural parallels worth monitoring.
Nvidia's own history illustrates how quickly GPU markets can shift. In 2022, crypto mining demand collapsed. Nvidia was left with more than $1 billion in chip inventory. Net income dropped 55 percent. Crypto lenders repossessed so many mining rigs that some reportedly began mining themselves just to liquidate the assets.
That represented roughly $4 billion in total crypto mining debt. The current GPU-backed debt market is substantially larger, and it involves traditional financial institutions in ways crypto lending did not.
Goldberg outlined what he describes as a hypothetical stress scenario. A smaller player takes out loans to build a data center. Construction gets delayed due to weather, power issues, or other factors. The company can't make payments. While a larger player like CoreWeave might survive a setback, a smaller company could fail. Given the interconnected nature of these financing arrangements, Goldberg suggests one collapse could potentially affect other parties up the chain.
The competitive landscape
Some analysts have noted that Nvidia's customer base is concentrated. In Q2 of fiscal 2026, two unnamed customers accounted for 23 percent and 16 percent of sales. According to Data Center Dynamics, analysts believe these are likely hyperscalers. Several of these large cloud providers are developing proprietary chips: Google has TPUs, Amazon has Trainium, Microsoft has Maia, and Meta has MTIA.
When Google released Gemini 3 in November, reportedly trained entirely on its own TPUs rather than Nvidia hardware, the market took notice. Nvidia's stock dropped 3 percent on reports that Meta was in talks to use Google TPUs. The company posted a statement on X: "We're delighted by Google's success—they've made great advances in AI and we continue to supply to Google. NVIDIA is a generation ahead of the industry—it's the only platform that runs every AI model and does it everywhere computing is done."
Neoclouds, by contrast, are focused on providing compute infrastructure rather than developing custom silicon. Some analysts, including Goldberg, have characterized the neocloud ecosystem as an alternative revenue stream for Nvidia. Supporting that ecosystem, however, requires capital commitments.
Goldberg believes the current trajectory faces constraints. "We're bumping up against the limit of what it's possible for them to support and finance," he said. "It can't go on forever. I don't know if it stops next year or the year after, but it can't go on at this pace. Something's gotta give."
Key developments to monitor
CoreWeave has approximately $7.5 billion in debt that requires refinancing in the near term. In December, the company completed a $2.25 billion convertible note offering at 1.75 percent interest, significantly cheaper than its earlier high-yield debt. Capital markets remain willing to fund the expansion.
But execution risk remains. According to media reports, construction delays at a Texas data center pushed completion back several months. That site is reportedly intended to serve OpenAI, which, according to the same reports, has contract terms allowing it to exit if CoreWeave cannot deliver on schedule. CoreWeave's SEC filings indicate that OpenAI-related contracts represent a significant portion of its backlog.
D.A. Davidson analyst Gil Luria has quantified the margin challenge: CoreWeave operates at roughly 4 percent margins, less than half what it pays in interest.
"The bull case is that they'll scale into it, and that a lot of companies have low margins to start, but this is a company at scale," Luria told the Wall Street Journal.
Mark Zandi, chief economist at Moody's Analytics, sees broader considerations. The tech sector has taken on more debt than it did during the dot-com bubble, he noted. Bond issuance by top AI companies is projected to reach $120 billion this year. Unlike the dot-com era, when internet companies were "funded by stocks and venture capital," current AI buildout relies heavily on debt financing.
Analyst perspectives
According to public filings, Nvidia sells chips to companies in which it also invests and with which it signs capacity contracts. How to interpret this structure is a matter of debate.
Bulls argue that if AI adoption continues, neoclouds will generate sustainable profits, refinance their debt on favorable terms, and reward investors. Bears, including Goldberg and Chanos, suggest that if growth slows or if competition from custom silicon intensifies, the financing arrangements could face stress. In a downside scenario, private credit funds could absorb losses that would flow through to their limited partners, including pension funds, endowments, and family offices.
When asked about the parallels to previous credit cycles, Raskin observed: "The parallels to the financial crisis are interesting. It's rhyming in a number of ways."
The outcome will depend on several factors that remain uncertain: the pace at which AI generates returns for customers, competitive dynamics in the chip market, capital market appetite for refinancing, and whether current depreciation schedules reflect actual asset life.
The information in this article is derived from publicly available sources, including SEC filings, press releases, and media reports. Analyst opinions cited herein represent the views of the individuals quoted and not the views of this publication. This article is not investment advice and should not be relied upon for trading or investment decisions. Companies mentioned in this article may dispute the characterizations or interpretations presented. Readers are encouraged to review primary sources and consult qualified professionals before taking any action based on this information.
Sources
SEC Filings
CoreWeave Q3 2025 10-Q — Revenue, debt, interest expense, and depreciation schedules
https://www.sec.gov/Archives/edgar/data/1769628/000176962825000062/crwv-20250930.htm
CoreWeave 8-K (September 2025) — $6.3 billion Nvidia capacity agreement disclosure
https://www.sec.gov/Archives/edgar/data/1769628/000176962825000047/crwv-20250909.htm
Nvidia Q2 FY2026 10-Q — Customer concentration, cloud compute commitments ($26B), and revenue breakdown
https://www.sec.gov/Archives/edgar/data/1045810/000104581025000209/nvda-20250727.htm
News and Analysis
The Verge, "Nvidia's chip loans to AI companies are becoming a problem" — Jay Goldberg analysis, Sarah Bloom Raskin interview, neocloud debt structures
https://www.theverge.com/ai-artificial-intelligence/848988/nvidia-chip-loans-coreweave-gpu-debt-ai-neocloud
The Information, "Nvidia to Spend $1.3 Billion Renting Its Own Chips Back From CoreWeave" — Nvidia as CoreWeave's second-largest customer
https://www.theinformation.com/
CNBC, "Nvidia says its GPUs are a 'generation ahead' of Google's AI chips" — Nvidia's defensive statement on X, Meta/Google TPU talks
https://www.cnbc.com/2025/11/25/nvidia-says-its-gpus-are-a-generation-ahead-of-googles-ai-chips.html
Wall Street Journal, "How Wall Street Lenders Are Betting Big on the AI Boom" — Blackstone/Magnetar deal structure, GPU-backed lending mechanics
https://www.wsj.com/finance/investing/ai-chips-loans-wall-street-nvidia-c8f3f8e2
Benzinga, "Nvidia's Depreciation Time Bomb: Jim Chanos Warns Of 'Massive Financial Risk' For CoreWeave, Oracle" (December 2025) — Depreciation concerns, neocloud debt default warnings
https://www.benzinga.com/markets/equities/25/12/49413156/nvidias-depreciation-time-bomb-jim-chanos-warns-of-massive-financial-risk-for-coreweave-oracle
Fortune, "Borrowing by AI companies represents a 'mounting potential threat'" (December 2025) — Mark Zandi analysis on AI debt vs. dot-com era
https://fortune.com/2025/12/09/ai-debt-bond-issuance-potential-threat-financial-system-top-economist-investment/
Wall Street Journal, "CoreWeave's Thin Margins Raise Questions" — Gil Luria analysis on 4% margins vs. interest costs
https://www.wsj.com/
PitchBook, "As venture debt gambles on GPUs, not all are sold on silicon-backed loans" — Ryan Little, Trinity Capital on GPU financing
https://pitchbook.com/news/articles/ai-venture-debt-gpu-chip-backed-loans
Fortune, "Inside OpenAI's fragile lead in the AI race" — Sam Altman "rough vibes" memo, Google Gemini 3 competitive pressure
https://fortune.com/2025/12/17/sam-altman-chatgpt-openai-versus-google-gemini-code-red-strategy/
Seeking Alpha, CoreWeave Earnings Call Transcripts — Management statements on recontracting older GPU capacity
https://seekingalpha.com/symbol/CRWV/earnings/transcripts
Company Announcements
CoreWeave press release (August 2023) — $2.3 billion debt facility led by Magnetar Capital and Blackstone
https://www.coreweave.com/blog/coreweave-secures-2-3-billion-debt-financing-magnetar-capital-blackstone
CoreWeave press release (December 2025) — $2.25 billion convertible note offering
https://www.coreweave.com/
Background
Laptop Mag, "Nvidia's Jensen Huang says new Blackwell chips make previous-gen feel obsolete" — GTC 2025 keynote quotes on Hopper obsolescence
https://www.laptopmag.com/ai/jensen-huang-nvidia-gtc-blackwell-ultra-vera-rubin-chips
Data Center Dynamics, "Two unnamed customers accounted for almost 40% of Nvidia's Q2 2026 revenue" — Customer concentration analysis
https://www.datacenterdynamics.com/en/news/two-unnamed-customers-accounted-for-almost-40-of-nvidias-q2-2026-revenue/
