Multiple reliable sources have reported that NVIDIA has communicated to its Chinese clientele its intention to supply the H200, its second-most powerful AI chip, prior to the Lunar New Year, which falls in mid-February 2026. The anticipated total shipment quantity ranges from 5,000 to 10,000 chip modules, translating to roughly 40,000 to 80,000 individual chips. In response, NVIDIA clarified that it is consistently overseeing its supply chain operations and assured that the provision of H200 chips to its authorized Chinese customers will not compromise its global supply capacity.

The H200 chip boasts 141GB of HBM3e memory along with a staggering 4.8TB/s bandwidth, marking a substantial leap in performance compared to its predecessors. This enhancement adequately caters to the computational requirements of the vast majority of domestic commercial applications. Nevertheless, it is important to note that the H200 chip has not yet secured the definitive import approval from Chinese regulatory bodies. Consequently, the exact timeline for delivery remains subject to uncertainty.