TL;DR
- Scaled-Back Deal: Nvidia is close to finalizing a $30 billion equity stake in OpenAI, replacing the original $100 billion partnership announced in September 2025.
- New Structure: The revised deal is a pure equity investment with no chip-purchase obligations, giving OpenAI full discretion over its hardware spending.
- Circular Loop: A substantial portion of Nvidia’s investment is expected to cycle back as GPU purchases, making Nvidia both a hardware supplier and equity stakeholder.
- Broader Round: Nvidia’s stake is part of a larger OpenAI funding round that could value the company at $730–$850 billion, per CNBC and the Financial Times.
Nvidia is close to finalizing a $30 billion equity stake in OpenAI, replacing the $100 billion partnership announced in September 2025, the Financial Times reported Thursday. That news arrives weeks after Nvidia CEO Jensen Huang dismissed reports of deal trouble as “complete nonsense” in Taipei, insisting the investment would proceed without interruption. CNBC independently confirmed the revised terms Thursday.
“Sam is closing the round, and we will absolutely be involved in the round,” Huang told reporters in late January. Financial Times reporting described that involvement as a pure equity stake, potentially closing as early as the weekend of February 21–22.
From $100 Billion to $30 Billion
When Nvidia announced the September 2025 partnership with OpenAI, the scale was the story: up to $100 billion directed toward AI infrastructure, a commitment WinBuzzer covered at the time. That announcement drove Nvidia’s stock past $5 trillion in market capitalization, reinforcing the chipmaker’s standing as the indispensable financial engine of the AI expansion. At the time, it was framed as a long-term structural commitment, not a transactional investment.
Subsequently, the investment plan was reported to have stalled in late January 2026. Huang’s response was sharp. At a press encounter in Taipei on January 31, he dismissed the stalling report as “nonsense” and offered a full-throated endorsement of the AI lab:
“I believe in OpenAI. The work that they do is incredible — they’re one of the most consequential companies of our time.”
Jensen Huang, CEO of Nvidia
Yet a widening gap between Huang’s January remarks and February’s reported terms indicates the deal was in active renegotiation even as he spoke. At its announced scale, the multi-year partnership will not proceed; the revised arrangement preserves Nvidia’s role as a direct equity stakeholder in OpenAI.
The New Deal Structure
Nvidia’s original letter of intent tied its investment to chip purchases: Nvidia would provide capital, OpenAI would use it to buy and deploy Nvidia hardware, making the deal as much a vendor subsidy as an investment. That structure, the Financial Times reported, has been abandoned entirely. According to the Financial Times, Nvidia will instead take a dedicated equity stake in OpenAI, not tied to any deployment milestones, removing any requirement for OpenAI to direct spending toward Nvidia hardware.
However, Huang had publicly declared at the time, “We will invest a great deal of money,” without specifying the final scale.
Reuters could not immediately verify the report; Nvidia declined to comment.
Consequently, this creates flexibility for OpenAI to direct Nvidia’s investment proceeds toward competing hardware vendors like AMD and Broadcom without violating any contractual commitment, removing the vendor-subsidy aspect from what is now a straightforward equity relationship. That structural change establishes a cleaner dynamic: Nvidia becomes a pure financial stakeholder, though OpenAI retains full discretion over its hardware spending.
Inside OpenAI’s Giant Funding Round
Moreover, Nvidia’s stake is one component of a broader capital effort. OpenAI is simultaneously seeking to close a larger equity round drawing participation from Amazon, Microsoft, and SoftBank’s OpenAI stake. CNBC reported the round could reach a pre-money valuation of $730 billion, per CNBC, placing OpenAI among the top-valued private companies globally, trailing only SpaceX.
Furthermore, the same round was valued closer to $850 billion per the Financial Times, a divergence that reflects both the round’s ongoing nature and the difficulty of establishing a single number for a company of OpenAI’s scale.
Additionally, that range represents a valuation roughly double what Anthropic achieved in its February 2026 fundraising round, underscoring how rapidly OpenAI’s capital requirements have escalated. Not every anticipated co-investor has committed.
Moreover, SoftBank said on a recent earnings call that “nothing has been decided” on its participation in the broader round, despite the Japanese conglomerate making billions on its existing OpenAI holdings last year. OpenAI’s ability to command near-trillion-dollar valuations reflects confidence in its market position even as rivals close the competitive gap.
Prior Coverage and Context
SoftBank’s uncertainty echoes an established pattern. OpenAI had been in discussions about a large equity raise targeting an $830 billion valuation to fund its Stargate infrastructure project. Altman extended that campaign in January 2026, courting Middle East sovereign wealth funds for additional commitments.
However, SoftBank’s participation in the current round remains unconfirmed despite months of coverage treating it as settled, part of a consistent pattern in which announced commitments have diverged substantially from finalized terms.
The Circular Investment Loop
Nevertheless, at the center of the new deal lies a structural paradox. Nvidia is the dominant supplier of AI processors used to train and run the large language models powering ChatGPT, Google Gemini, and the rest of the generative AI market. That dominance has fed an industry directing enormous capital into GPU-packed data centers ahead of anticipated demand surges.
Meanwhile, OpenAI is both one of Nvidia’s largest customers and, under this deal, one of its equity portfolio companies.
Moreover, a substantial share of the new Nvidia capital would cycle back into Nvidia hardware purchases, the newspaper reported. That arrangement, if confirmed, creates a circular financial loop in which Nvidia writes a large check and a substantial portion returns as GPU orders. Alvin Nguyen, an analyst at Forrester Research, explained why this dynamic is structural rather than incidental: “They need chips. They need as many as possible,” Nguyen noted.
As a result, Nvidia benefits from the AI buildout simultaneously as hardware supplier and equity stakeholder, an alignment of incentives that few technology companies have managed at this scale.
Chip Diversification and Competitive Pressure
Despite this, OpenAI has been diversifying its chip supply. OpenAI announced a multi-billion dollar AMD-OpenAI partnership in October 2025, targeting viable alternatives to Nvidia’s H100 and Blackwell chips, and struck a separate custom chip deal with Broadcom.
Furthermore, that leverage likely explains why OpenAI was able to negotiate the new Nvidia deal without a chip-purchase obligation attached. Broadcom CEO Hock Tan had already signaled the limits of such arrangements in December 2025, telling investors the company did “not expect much in 2026” from its own OpenAI custom chip partnership.
Yet chip diversification alone cannot close OpenAI’s competitive gap. According to Menlo Ventures data tracking enterprise AI adoption, ChatGPT’s market share fell from 86.7% to 64.5% over the past year.
Consequently, Nguyen, commenting on the gap between the original headline and the reality that emerged, argued the discrepancy was predictable:
“They will not discourage people from overhyping. Why say something and immediately sucker punch your own share price?”
Alvin Nguyen, Analyst at Forrester Research (via The Guardian)
Therefore, market share decline indicates OpenAI faces competitive pressure that investment capital alone cannot resolve, making chip diversification as much a hedge against vendor lock-in as a tool for negotiating better hardware terms.
Nevertheless, with Nvidia declining to comment and the Financial Times report sourced entirely from unnamed parties, the reported figure carries the standard caveats of anonymous deal reporting. Earlier reporting had already found that Nvidia described the deal as “non-binding” and unfinalized, making the gap between widely reported commitments and the actual investment unsurprising.
Thus, AI megadeals follow a recognizable pattern: announced loudly at a figure that satisfies investors and press alike, then restructured quietly as financing mechanics are worked out. Neither party has much incentive to correct inflated headline numbers while they serve a purpose.
If the deal finalizes this weekend as reported, the finalized terms will fall well short of the September 2025 headline figure. That outcome marks the normalization of a mode in which the announcement is the product, and the actual terms are the fine print.