Nvidia Faces Scrutiny Over Circular Deal Structures Amid AI Investment Boom

Nvidia has found itself in the unusual position of reassuring investors that it bears no resemblance to Enron, the Houston energy conglomerate that collapsed spectacularly in 2001 through accounting fraud. The chipmaker, now valued at more than $4 trillion, has also distanced itself from dotcom-era casualties such as Lucent Technologies and WorldCom. The very need for such clarifications signals growing unease about the company’s complex financing arrangements.

The Santa Clara-based semiconductor manufacturer has emerged as the dominant supplier of specialised chips and software systems that power the global artificial intelligence revolution. Its products form the technological backbone of data centres from Scandinavia to North America, enabling the training and operation of large language models including ChatGPT. This year alone, Nvidia has executed at least $125 billion in agreements, ranging from a $5 billion stake in Intel to facilitate PC market access, through to a staggering $100 billion commitment to OpenAI.

These transactions have propelled Nvidia’s share price to record heights and enabled chief executive Jensen Huang to conduct an energetic series of international engagements. However, concerns have intensified regarding the circular nature of many arrangements, which critics suggest resemble vendor financing structures where the supplier effectively funds customer purchases of its own products.

The most significant of these deals involves Nvidia investing $10 billion annually into OpenAI over the next decade, with the majority of those funds earmarked for purchasing Nvidia’s own chip systems. A similar arrangement exists with CoreWeave, a cloud computing provider that leases Nvidia’s processing capacity to major AI firms. Such structures have drawn uncomfortable parallels with Lucent Technologies, the telecommunications company that extended aggressive customer financing during the dotcom bubble before overreaching itself and unravelling in the early 2000s.

Nvidia has vigorously contested these comparisons. A leaked internal memorandum stated explicitly that the company “does not rely on vendor financing arrangements to grow revenue”. Nonetheless, prominent technology investor James Anderson, whilst describing himself as a “huge admirer” of Nvidia, has expressed reservations about the OpenAI transaction. Anderson noted that the term “vendor financing” carries troubling historical connotations, drawing uncomfortable parallels with practices employed by telecommunications suppliers during 1999-2000.

The web of interconnected agreements extends beyond Nvidia. Oracle has committed $300 billion to construct data centres for OpenAI across the United States, with the ChatGPT developer subsequently paying approximately equivalent sums to utilise that infrastructure. In October, OpenAI signed a multibillion-dollar chip supply agreement with AMD, Nvidia’s competitor, which includes an option for OpenAI to acquire an equity stake in the chipmaker. CoreWeave has similarly agreed to provide OpenAI with $22 billion in data centre capacity, whilst OpenAI receives $350 million in CoreWeave equity.

Michael Intrator, CoreWeave’s chief executive, defended these arrangements when questioned about circularity within the AI sector, stating that companies are addressing “a violent change in supply and demand” through collaborative approaches. OpenAI maintains that whilst its Nvidia and AMD agreements contain investment components, these only activate after chip purchases and deployments are completed, creating aligned incentives for large-scale AI infrastructure development.

These transactions form part of OpenAI’s extraordinary $1.4 trillion wager on computing capacity to develop models that, the company argues, will fundamentally transform global economies and generate sufficient returns to justify the expenditure. Nvidia has additionally employed special-purpose vehicles in certain financing arrangements, most notably a $2 billion investment into an SPV connected to Elon Musk’s xAI venture, with proceeds designated for Nvidia chip acquisitions.

This structure inevitably triggered comparisons with Enron, which notoriously utilised SPVs to conceal debts and toxic assets from its balance sheets, misleading investors and creditors about its true financial condition whilst liabilities escalated. Nvidia has categorically rejected this parallel, emphasising in the same leaked memorandum that its reporting remains “complete and transparent” and that, “unlike Enron”, it “does not use special-purpose entities to hide debt and inflate revenue”.

Technology journalist Ed Zitron, a prominent sceptic of the current AI enthusiasm, concurs that Nvidia differs fundamentally from both precedents. Unlike Lucent, Nvidia does not appear to be accumulating substantial debt to finance circular arrangements, and most customers it supports present lower risk profiles than Lucent’s dotcom-era partners. Zitron argues Nvidia also differs from Enron through its relative transparency regarding complex, off-balance sheet transactions.

Charlie Dai, an analyst at research firm Forrester, characterises the concerns differently. Nvidia “is not hiding debt, but it is leaning heavily on vendor-financed demand, which creates exposure if AI growth slows,” he observes. “The concern is about sustainability, not legality.” The fundamental question centres on whether artificial intelligence will generate the anticipated billions in value for corporate users, enabling companies such as OpenAI, Anthropic, and CoreWeave to achieve profitability and maintain purchasing capacity for Nvidia’s systems.

Should this materialise, Nvidia’s strategy appears sound. However, if the AI revolution fails to deliver expected returns, Dai warns that Nvidia “could face write-downs on equity stakes and unpaid receivables”, potentially triggering substantial losses and share price deterioration. When approached for comment, an Nvidia spokesperson directed enquiries to remarks made by chief financial officer Colette Kress to investors in early December. Kress dismissed suggestions of an AI bubble, highlighting trillions of dollars in prospective business over the coming decade.

Kress argued that Nvidia’s recent substantial agreements represent merely the initial phase, with genuine profitability materialising through the replacement of existing data centre chips with Nvidia products. This calculation introduces additional complexity, as Nvidia’s financial health, and by extension significant portions of the global economy, depends upon AI adoption occurring rapidly enough for Nvidia and its customers to service debts from massive infrastructure investments and capital expenditures.

A final category of concern involves recent high-value agreements with sovereign nations including South Korea and Saudi Arabia, involving billions of dollars under opaque terms. In October, Nvidia announced it would supply 260,000 Blackwell chips to South Korea’s government and domestic companies. The transaction value remains undisclosed, though estimates suggest billions of dollars are involved.

Saudi Arabia presents a similar situation. Its government-owned AI venture, Humain, has committed to deploying up to 600,000 Nvidia chips, though precise deployment schedules, purchase timings, and pricing remain unannounced. Nvidia maintains numerous comparable strategic partnerships with Italy, French AI champion Mistral, and Deutsche Telekom, each involving thousands of chips and undisclosed financial commitments.

Whilst sovereign counterparties presumably possess payment capacity, these agreements introduce substantial uncertainties within an already strained network of commitments requiring enormous capital outlays, predicated upon ambitious assumptions regarding imminent economic transformation. Dai notes that such arrangements “concentrate risk in a few big customers” and warns that “if execution delays occur, Nvidia’s revenue recognition and cashflow could be affected”.

The situation presents investors with a complex assessment. Nvidia’s business model appears transparent and fundamentally different from historical corporate failures. However, the company’s extraordinary growth trajectory and the global economy’s increasing dependence upon its success rest heavily upon the artificial intelligence sector delivering transformational value within a compressed timeframe. Whether this gamble succeeds will determine not merely Nvidia’s fate, but potentially reshape substantial portions of the technology sector and broader financial markets.