NVIDIA is set to resume shipping its H200 artificial intelligence (AI) acceleration chips to the Chinese market from mid-February 2026. For the initial phase, the company will rely on its existing stockpile. The first batch is projected to include roughly 5,000 to 10,000 modules, which translates to 40,000 to 80,000 individual chips. NVIDIA has already initiated efforts to ramp up its production capacity and intends to start accepting new orders from the second quarter of 2026 onwards.

However, given that all transactions are subject to approval from Chinese regulatory bodies, the actual delivery timeline might see adjustments. As of now, the H200 chips have not secured the final import approval, leaving the delivery schedule in a state of flux. Built on the 'Hopper' architecture, the H200, though not NVIDIA's cutting-edge offering, boasts substantial enhancements in memory bandwidth and energy efficiency, aligning well with the present requirements of Chinese customers.

In the meantime, China's domestic AI chip sector is rapidly advancing. The limited resumption of H200 shipments represents one aspect of the ongoing dynamic adjustments within the global AI supply chain. NVIDIA is actively engaging with regulators to guarantee that the deliveries comply with all relevant regulations.