Nvidia projects $1 trillion AI chip opportunity as inference demand accelerates
Nvidia said the revenue opportunity for its artificial intelligence chips may reach at least $1 trillion through 2027, up from a $500 billion forecast made in February. CEO Jensen Huang announced the projection at the company's annual developer conference in San Jose on March 16, outlining a strategy to compete more aggressively in inference computing-the process of running AI queries in real time.
Huang unveiled a new central processor and an AI system built on technology from Groq, a chip startup from which Nvidia licensed technology for $17 billion in December. The moves target inference, where Nvidia's graphics processors face growing competition from central processing units and custom chips built by Google and others.
Nvidia's chips have dominated AI model training, which requires massive computational power upfront. Inference-answering questions and executing tasks with trained models-represents a different challenge. "The inference inflection has arrived," Huang said at the four-day conference, which drew more than 18,000 attendees.
Breaking inference into two stages
Huang said inference will split into two distinct steps. Nvidia's Vera Rubin chips will handle "prefill," converting user requests into tokens that AI systems understand. Groq's chips will handle "decode," where the AI system provides the answer.
This division reflects how companies are shifting spending priorities. After investing hundreds of billions in training chips, firms like OpenAI, Anthropic, and Meta are now scaling to serve hundreds of millions of users. That shift is driving demand for CPUs, traditionally dominated by Intel.
Huang said Nvidia is already selling CPUs separately. "This is already for sure going to be a multi-billion-dollar business for us," he said.
Broader infrastructure play
Huang also presented Nvidia's Feynman roadmap, expected in 2028 after the company's Rubin Ultra chips. The architecture includes AI processors and networking chips, signaling Nvidia's move beyond standalone processors toward complete systems.
The company is also targeting autonomous AI agents with NemoClaw, integrating with the OpenClaw platform to add privacy and safety controls for tasks that require minimal human guidance.
Nvidia stock briefly jumped on the $1 trillion forecast but closed up 1.2%. Emarketer analyst Jacob Bourne said the projection "underscores the durable demand for Nvidia's AI infrastructure despite investor concerns" and signals the company is sustaining its leadership as the broader AI industry moves beyond early experimentation into large-scale deployment.
For sales professionals, understanding Nvidia's infrastructure roadmap matters. As companies invest in AI deployment at scale, the hardware and systems they choose will shape procurement decisions, vendor relationships, and competitive positioning. Learn more about AI for Sales and how infrastructure decisions impact revenue strategy.
Your membership also unlocks:
