Vanished: Nvidia’s $100 Billion OpenAI Deal Disappears
▼ Summary
– In September 2025, Nvidia and OpenAI announced a non-binding letter of intent for Nvidia to potentially invest up to $100 billion in OpenAI’s AI infrastructure, but the deal was never finalized.
– Nvidia’s CEO has since stated the $100 billion figure was “never a commitment” and that any investment would be made incrementally, though he confirmed Nvidia plans a “huge” and likely record investment in OpenAI.
– Reuters reported that OpenAI has been quietly seeking alternatives to Nvidia chips due to dissatisfaction with the speed of some Nvidia hardware for AI inference tasks, which generate model responses.
– The performance issue was reportedly observed with OpenAI’s Codex tool, where staff attributed some limitations to Nvidia’s GPU-based hardware.
– Following the Reuters report and a drop in Nvidia’s stock, OpenAI’s CEO publicly reaffirmed the company’s strong partnership with Nvidia, calling their chips the best and expressing a desire to remain a major customer.
The ambitious $100 billion partnership between Nvidia and OpenAI, once heralded as a landmark deal for artificial intelligence infrastructure, has quietly evaporated. Announced with great fanfare in September 2025, the letter of intent outlined a staggering investment for Nvidia systems requiring power equivalent to ten nuclear reactors. Five months later, no formal agreement has been reached, and Nvidia’s CEO has publicly stated the colossal figure was “never a commitment.” This dramatic shift coincides with reports that OpenAI has been actively exploring alternatives to Nvidia’s hardware, citing performance concerns for critical AI tasks.
According to a recent Reuters investigation, OpenAI’s dissatisfaction stems from the inference speed of some Nvidia chips. Inference is the process where a trained AI model, like OpenAI’s Codex tool for generating computer code, produces answers to user prompts. Internal staff reportedly linked certain performance bottlenecks in Codex directly to the limitations of Nvidia’s GPU-based systems. This search for other chip suppliers is said to have been underway for over a year, suggesting deeper strategic deliberations beyond a single deal.
The news prompted a noticeable dip in Nvidia’s stock price, leading both companies to issue public reassurances about their relationship. OpenAI CEO Sam Altman took to social media to emphasize his company’s admiration for Nvidia, calling their chips the world’s best and expressing a desire to remain a “gigantic customer for a very long time.” He questioned the origin of what he termed the surrounding “insanity.” Similarly, Nvidia’s Jensen Huang reaffirmed his company’s intention to be a major investor in OpenAI, potentially its largest ever, but carefully distanced himself from the specific $100 billion headline.
During the initial announcement, Huang described the project as matching Nvidia’s total annual GPU shipments, labeling it a “giant” undertaking. The core of the current ambiguity lies in the non-binding nature of the original letter of intent. Huang now clarifies that OpenAI invited an investment “up to” that extraordinary sum and that Nvidia’s approach would be incremental, investing “one step at a time.” When pressed on whether the final figure would approach $100 billion, his response was definitive: “No, no, nothing like that.”
This development highlights the volatile and high-stakes nature of the AI hardware race. While Nvidia remains the dominant force in AI chips, its key partners are actively pursuing a multi-vendor strategy to avoid over-reliance and seek optimal performance. The vanished deal underscores that even the most promising collaborations in this fast-moving field are subject to rapid renegotiation as technological needs and market dynamics evolve. The future relationship between the AI software pioneer and the chipmaking titan will likely involve significant investment, but on a scale far removed from the eye-popping numbers that first captured the world’s attention.
(Source: Ars Technica)
