Shares of chip giant Nvidia (NVDA) fell 2.9% on February 2 amid confusing signals from OpenAI and CEO Sam Altman. Rumors swirled that Nvidia had halted plans to invest $100 billion in the AI startup. Meanwhile, a Reuters report, citing sources, stated that Altman is unsatisfied with some of Nvidia’s latest AI chips used for inferencing and has sought alternative deals with Cerebras and Groq since last year. However, both Altman and Nvidia CEO Jensen Huang have dismissed the claims.
Claim 50% Off TipRanks Premium
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Stay ahead of the market with the latest news and analysis and maximize your portfolio's potential
Altman clarified in an X post, “We love working with Nvidia and they make the best AI chips in the world. We hope to be a gigantic customer for a very long time.” Nvidia echoed this, stating, “Customers continue to choose Nvidia for inference because we deliver the best performance and total cost of ownership at scale.” For reference, Nvidia holds about 80% of the GPU market.
OpenAI’s Inferencing Challenge
Nvidia’s GPUs (graphics processing units) excel at heavy data processing needed to train big AI models like ChatGPT, fueling global AI growth. But AI workloads are shifting to inference and reasoning, which is the next major phase powering tools like OpenAI’s models. McKinsey projects that the inference market could be 2x-5x larger than training, reaching over $200 billion annually by 2028.
OpenAI is seeking GPUs with lots of built-in SRAM memory on each chip for faster chatbot responses to millions of users. Nvidia and Advanced Micro Devices (AMD) GPUs rely on external high-bandwidth memory (HBM), which slows processing and delays chatbot responses for users. Nvidia’s H100 pairs HBM with on-chip SRAM, but wafer-scale rivals like Cerebras pack far more SRAM per chip, reducing latency.
OpenAI spotted slowdowns in its Codex code-writing tool, with staff blaming Nvidia GPUs for the weakness. Altman has emphasized speed as critical for coding users and highlighted a deal with Cerebras to address it. Rivals like Anthropic’s Claude and Google’s (GOOGL) Gemini use in-house TPUs (tensor processing units), optimized for AI inference and faster than Nvidia GPUs.
Is it Too Late to Invest in Nvidia Now?
Analysts remain highly optimistic about Nvidia’s long-term prospects. On TipRanks, Nvidia stock has a Strong Buy consensus rating based on 38 Buys, one Hold, and one Sell rating. The average Nvidia price target of $262.79 implies 41.6% upside potential from current levels. Over the past year, NVDA shares have surged more than 59%.
