NVIDIA’s Jensen Huang Predicts $1 Trillion AI Revenue by 2027

NVIDIA's Jensen Huang predicts $1 trillion in AI chip orders by 2027. Key focus on lowering cost per token to drive inference inflection.

Artificial intelligence inference has reached an inflection point, Huang says. The shift from model training to advanced inferencing is transforming AI’s main focus. 'It’s way past training now,' Huang stated. 'Inference is your workload and tokens are your new commodity.'

Huang unveiled new GPU and CPU models, emphasizing Nvidia’s extreme co-design process. This approach has reduced cost per token, making it the cheapest in the world. 'Our cost per token is the lowest in the world. You can't beat it,' he asserted.

Huang’s vision reflects growing pressure on AI vendors to justify investments. Enterprises have invested millions in AI infrastructure, expecting 2026 to bring returns. Nvidia plans to boost token generation rate from 2 million per second to 700 million.

Nvidia also introduced NemoClaw, an enhanced version of OpenClaw, with security protocols ensuring privacy and cybersecurity. The growing influence of agents and the shift in AI economics highlight the importance of token generation and infrastructure.