Nvidia ships first H200 chips to China, but forecasts no data-centre computing revenue

Nvidia's H200 chip shipments to China are a technicality, not a revenue driver. The sales accounted for less than 1 percent of the company's US$89 billion data-center revenue in the second quarter, and Nvidia's US$108 billion third-quarter forecast still assumes no data-center computing revenue from China. This reflects the strict US licensing scheme, which effectively caps the commercial upside for advanced AI chips in the Chinese market. For Chinese AI developers, this means continued reliance on less powerful, custom-built, or domestically produced alternatives, limiting the scale and speed of advanced AI model training compared to global peers. The constraint on high-performance chips will likely accelerate China's domestic chip development efforts, but the performance gap remains significant. Nvidia CEO Jensen Huang's global demand outlook is the more important takeaway for Asia. He forecasts 70 percent revenue growth in the 2028 financial year, describing it as supply-constrained. Huang stated actual demand is "much greater than 70 percent," with hyperscalers accounting for only half the opportunity. This suggests strong demand across enterprises, neo clouds, and sovereign AI projects globally, including in other Asian markets. The challenge for Nvidia and its customers in Asia will be securing sufficient capacity to meet this unconstrained demand, as the company still needs "a lot more" supply.