Keep pulling the thread on Bill Gurley.
Jensen Huang predicts that OpenAI is likely to become the next multi-trillion dollar hyperscale company.
Jensen Huang predicts that the demand for AI inference compute will increase by a factor of one billion due to the adoption of multi-step reasoning techniques.
NVIDIA will invest $100 billion over time in OpenAI as part of the 'Stargate' project partnership.
The OpenAI 'Stargate' project, which aims to build 10 gigawatts of AI infrastructure, could represent up to $400 billion in potential revenue for NVIDIA if its hardware is used.
NVIDIA is working directly with OpenAI to help it build its own self-built AI infrastructure for the first time, a project separate from its cloud provider partnerships.
Jensen Huang asserts that the entire multi-trillion dollar global computing infrastructure will be refreshed from general-purpose computing to accelerated and AI-based computing.
Jensen Huang projects that the annual capital expenditure required to support a $10 trillion AI token generation market could be approximately $5 trillion.
Alibaba CEO Eddie Wu stated that the company will increase its data center power capacity by 10x by the end of the decade.
Jensen Huang asserts that the AI-driven revenue market has already surpassed $1 trillion because the entire revenue base of hyperscalers like TikTok and Meta is now AI-driven.
Jensen Huang believes the risk of an AI compute glut is extremely low until the multi-year transition from general-purpose computing to accelerated AI systems is complete.
NVIDIA will soon announce a major initiative for accelerated data processing to move workloads from companies like Databricks, Snowflake, and Oracle from CPUs to its platform.
NVIDIA has moved to an annual release cycle for its AI platforms, with Grace Blackwell in 2025, Vera Rubin in 2026, Ultra in 2027, and Feynman in 2028.