Nvidia unveils AI chips, $1T forecast

Nvidia CEO Jensen Huang unveiled new central processors and AI systems at the company's annual GTC developer conference in San Jose, California. Nvidia projected a revenue opportunity of at least $1 trillion through 2027 for its artificial intelligence chips, a significant increase from its previous forecast. The company is aggressively competing in the fast-growing market for running AI systems in real time, known as inference computing. Nvidia's strategy involves its new Vera Rubin chips handling the 'prefill' stage of inference, while Groq's chips will manage the 'decode' stage. This move addresses the increasing demand from companies like OpenAI, Anthropic, and Meta Platforms, which are shifting from AI model training to serving millions of users. Nvidia also introduced a new Vera CPU, expecting it to become a multi-billion-dollar business, and showcased its Feynman roadmap for future chip architectures. The company's shares briefly jumped on the forecast, closing up 1.2% to 1.6%.