NextFin News - In a series of high-stakes maneuvers that have rattled the semiconductor market, OpenAI CEO Sam Altman clarified on February 2, 2026, that the artificial intelligence pioneer intends to maintain its status as a "gigantic customer" for Nvidia. The statement, made during a period of intense speculation regarding the relationship between the two tech titans, comes as U.S. President Trump’s administration continues to emphasize American leadership in the global AI arms race. While Altman reaffirmed the partnership, internal shifts at OpenAI suggest a more nuanced reality: the company is actively diversifying its hardware fleet to address specific performance bottlenecks in AI inference.
According to Reuters, OpenAI has been unsatisfied with the response speeds of certain Nvidia chips when handling specific tasks like software development and real-time AI-to-software communication. This dissatisfaction has led Altman to explore deals with Advanced Micro Devices (AMD) and specialized startups including Cerebras Systems and Groq. The news sent ripples through the Nasdaq on Monday, with AMD shares jumping roughly 4% to $246.27 in after-hours trading, while Nvidia shares slipped 2.9% to close at $185.61. The market reaction underscores a growing investor sensitivity to any potential erosion of Nvidia’s near-monopoly on AI compute.
The core of the technical friction lies in the architectural difference between "training" and "inference." While Nvidia’s H-series and Blackwell GPUs are the gold standard for training massive models, the "inference" phase—where a trained model generates a response to a user prompt—requires different optimizations. OpenAI sources indicate that for products like Codex, which powers automated coding, the latency involved in fetching data from external memory in traditional GPUs has become a competitive liability. Consequently, OpenAI is seeking hardware with higher on-chip SRAM (Static Random-Access Memory) to reduce data retrieval times, a move that Altman noted is essential because customers "put a big premium on speed for coding work."
Despite these exploratory forays into alternative silicon, the financial ties between the two companies remain deep. Nvidia CEO Jensen Huang recently dismissed reports of friction as "nonsense," confirming that Nvidia still plans a "huge" investment in OpenAI, though clarifying it would not reach the previously rumored $100 billion mark. Huang emphasized that Nvidia’s chips remain the most cost-effective solution at scale, a sentiment echoed by an OpenAI spokesperson who confirmed that Nvidia still powers the "vast majority" of the company’s inference fleet. This suggests that while OpenAI may shift approximately 10% of its specialized workloads to other providers, it remains tethered to Nvidia for its massive general-purpose computing needs.
From an industry perspective, this development signals the end of the "one-size-fits-all" era of AI hardware. As the industry matures, the focus is shifting from raw power to efficiency and latency. Google’s success with its in-house Tensor Processing Units (TPUs) has proven that specialized silicon can offer a significant edge in inference speed for models like Gemini. OpenAI’s recent deal with Cerebras is a direct attempt to replicate this advantage without being entirely beholden to a single vendor’s roadmap. For Nvidia, the challenge is no longer just building the fastest chip, but defending its ecosystem against a "best-of-breed" hardware strategy adopted by its largest clients.
Looking ahead, the relationship between Altman and Huang will likely be characterized by "co-opetition." OpenAI needs Nvidia’s massive supply chain and software stack (CUDA) to maintain its lead in model training, while Nvidia needs OpenAI’s prestige and massive order volumes to justify its $4.5 trillion valuation. However, as inference becomes a larger share of total AI spend—projected by some analysts to surpass training spend by 2027—the pressure on Nvidia to innovate in low-latency architectures will intensify. For investors, the volatility in AMD and Nvidia stock serves as a preview of a more fragmented AI infrastructure landscape where specialized silicon providers will increasingly capture high-margin niches previously dominated by general-purpose GPUs.
Explore more exclusive insights at nextfin.ai.
