The artificial intelligence industry’s most ambitious infrastructure agreement has quietly disintegrated, revealing deep fissures in what was supposed to be a defining partnership between two of the sector’s most powerful players. The $100 billion memorandum of understanding between Nvidia and OpenAI, announced with considerable fanfare last September, has effectively stalled, according to multiple sources familiar with the negotiations. What was billed as “the largest computing project in history” by Nvidia CEO Jensen Huang now stands as a cautionary tale about the perils of announcing transformative deals before the ink has dried on binding contracts.

The original agreement, unveiled at Nvidia’s Santa Clara, California headquarters, called for the chip giant to build at least 10 gigawatts of computing power for OpenAI while investing up to $100 billion to help finance the infrastructure. OpenAI would lease the chips from Nvidia in an arrangement designed to provide the ChatGPT maker with the massive computational resources needed to train and deploy its next-generation AI models. At the time, The Wall Street Journal reported that OpenAI expected negotiations to conclude within weeks, but the talks never progressed beyond preliminary stages.

Behind Closed Doors: Huang’s Growing Skepticism

Jensen Huang has privately expressed significant reservations about the partnership to industry associates in recent months, according to people familiar with the conversations. The Nvidia chief executive has emphasized that the original $100 billion agreement was nonbinding and never reached finalization. More pointedly, Huang has criticized what he characterizes as a lack of business discipline at OpenAI and voiced concerns about the intensifying competition the company faces from rivals including Google and Anthropic, sources told The Wall Street Journal.

These private misgivings stand in stark contrast to the public optimism displayed when the deal was announced. Nvidia’s stock surged nearly 4% following the September announcement, propelling the company’s market valuation to approximately $4.5 trillion. The agreement included discussions about Nvidia guaranteeing loans OpenAI planned to secure for building its own data centers, representing an extraordinary level of financial commitment from the chip manufacturer to its customer.

The Shifting Terms of Engagement

Rather than abandoning their relationship entirely, both companies are now exploring alternative arrangements. Current discussions center on a potential equity investment of tens of billions of dollars as part of OpenAI’s ongoing funding round, according to sources familiar with the negotiations. This represents a significant scaling back from the original $100 billion infrastructure and investment package, though it would still constitute one of the technology sector’s largest financial commitments.

An OpenAI spokesman maintained a diplomatic stance, stating: “Our teams are actively working through details of our partnership. NVIDIA technology has underpinned our breakthroughs from the start, powers our systems today, and will remain central as we scale what comes next.” Nvidia echoed this sentiment, with a spokeswoman noting that the company has been OpenAI’s preferred partner for a decade and looks forward to continuing the collaboration.

Altman’s High-Stakes Dealmaking Strategy Under Scrutiny

The stalled agreement represents a significant setback for OpenAI CEO Sam Altman’s aggressive strategy of securing vast amounts of computing capacity to fuel the company’s growth. OpenAI is laying groundwork for a potential public offering by the end of 2026, making reliable access to computational infrastructure critical for demonstrating the company’s ability to scale its operations and maintain technological leadership. The collapse of the Nvidia megadeal raises questions about Altman’s approach of announcing major agreements before finalizing binding terms.

Following the September Nvidia announcement, OpenAI proceeded to sign multiple additional agreements with chip manufacturers and cloud computing providers, collectively contributing to a global technology stock rally. However, investor sentiment has since soured as concerns mounted about OpenAI’s ability to fulfill its financial obligations. Altman has publicly stated that these deals place the company on the hook for $1.4 trillion in computing commitments—a figure exceeding 100 times the revenue OpenAI was projected to generate last year, according to The Wall Street Journal.

The Reality Behind the Trillion-Dollar Commitments

OpenAI executives have attempted to temper concerns about these astronomical figures, explaining that the actual commitments are lower when accounting for overlapping provisions in various agreements. Additionally, these deals are structured to unfold over extended timeframes rather than requiring immediate payment. Nevertheless, the sheer scale of the financial obligations has contributed to volatility in technology stocks associated with OpenAI’s ecosystem, as investors reassess the startup’s financial sustainability.

The timing of the Nvidia deal’s deterioration is particularly problematic for OpenAI given mounting competitive pressures. Google’s Gemini application has demonstrated unexpected strength, slowing ChatGPT’s user growth and prompting OpenAI to declare an internal “code red” situation. Anthropic, another formidable competitor, has gained significant traction with its AI coding agent called Claude Code, further intensifying the competitive environment that Huang reportedly cited as a concern.

Regulatory Disclosures Foreshadowed Trouble

Signs of trouble emerged in regulatory filings months before the deal’s collapse became widely known. In a November filing, Nvidia disclosed that “there was no assurance” the company would “enter into definitive agreements with respect to the OpenAI opportunity or other potential investments, or that any investment will be completed on expected terms, if at all.” This cautious language provided early indication that the September announcement had overstated the certainty of the partnership’s future.

At a UBS conference in Scottsdale, Arizona, in December, Nvidia CFO Colette Kress publicly acknowledged that the company had not completed a definitive agreement with OpenAI. These statements marked a significant departure from the confident tone struck during the September announcement, when both companies presented the deal as essentially finalized pending only procedural details.

The Strategic Imperative Driving Nvidia’s Calculations

Despite his reservations, Huang has indicated to associates that providing OpenAI with substantial financial support remains strategically important for Nvidia, according to sources familiar with his thinking. The rationale is straightforward: OpenAI ranks among Nvidia’s largest customers, and any significant decline in the company’s market position could materially impact Nvidia’s revenue. This creates a delicate balancing act for Huang, who must weigh the risks of overcommitting to a single customer against the dangers of allowing that customer to falter.

The competitive dynamics in AI chip markets add another layer of complexity to Nvidia’s calculations. Anthropic relies heavily on Amazon Web Services’ Trainium chips and Google’s TPU processors for training its AI models, while Google predominantly uses TPUs for its Gemini system. These alternative chip architectures represent direct threats to Nvidia’s graphics processing units (GPUs), which currently dominate the AI training market. If OpenAI were forced to reduce its reliance on Nvidia chips due to financial constraints, it could accelerate adoption of competing technologies.

Nvidia’s Hedging Strategy with Anthropic

Nvidia’s investment strategy reveals a pragmatic approach to managing risk in the rapidly evolving AI sector. In November, the company committed to investing up to $10 billion in Anthropic, OpenAI’s primary competitor. This move provides Nvidia with exposure to multiple leading AI developers, reducing dependence on any single customer while ensuring demand for its chips regardless of which company ultimately achieves market dominance.

The Anthropic investment also sends a subtle message to OpenAI: Nvidia possesses alternatives and will not hesitate to support competitors if necessary. This dynamic fundamentally alters the power balance in negotiations between the two companies, potentially explaining why discussions have shifted from the original $100 billion infrastructure deal to a more modest equity investment.

Implications for OpenAI’s Path to Public Markets

The unraveling of the Nvidia megadeal complicates OpenAI’s preparation for its anticipated 2026 initial public offering. Public market investors typically demand clear visibility into a company’s infrastructure costs and capacity constraints, both of which remain uncertain for OpenAI given the unsettled state of its computing partnerships. The company must now demonstrate that it can secure adequate computational resources at sustainable prices without the guaranteed capacity the Nvidia deal was supposed to provide.

Furthermore, the situation highlights broader questions about OpenAI’s business model sustainability. The company’s massive computational requirements create enormous fixed costs that must be supported by growing revenue. While OpenAI has achieved impressive revenue growth, the gap between its income and infrastructure commitments remains vast. Potential public market investors will scrutinize whether the company can narrow this gap sufficiently to achieve profitability at scale.

The Broader Industry Implications

The collapse of this deal extends beyond the immediate parties involved, offering insights into the challenges facing the entire AI industry. The episode demonstrates that even the sector’s most powerful companies struggle to make long-term infrastructure commitments with confidence, given rapid technological change and uncertain demand trajectories. This uncertainty complicates capital planning for AI developers and chip manufacturers alike, potentially slowing the pace of infrastructure investment.

The situation also illustrates the risks inherent in the current AI business model, which requires enormous upfront capital expenditures years before revenue materializes. This dynamic creates vulnerability to shifts in competitive positioning, technology evolution, and market sentiment. Companies that announce major deals prematurely may find themselves unable to fulfill commitments if circumstances change, damaging credibility with partners and investors.

For Nvidia, the experience may prompt more cautious approaches to future partnership announcements, with the company likely to insist on finalized, binding agreements before public disclosure. For OpenAI, the setback underscores the need to balance ambitious growth plans with financial discipline and operational realism. As both companies navigate these challenges, their evolving relationship will provide important signals about the maturation of the AI industry and the sustainability of current business models in this transformative but uncertain sector.