Nvidia has made a strategic decision to halt production of its H200 AI chips, redirecting manufacturing capacity at Taiwan Semiconductor Manufacturing Company (TSMC) toward the next-generation Vera Rubin hardware, according to reports from the Financial Times.
The move marks a significant pivot in Nvidia’s chip roadmap, as the company accelerates its transition to newer architectures despite continued strong demand for current-generation products. The H200, which launched in late 2024, has been one of the most sought-after AI accelerators in the industry, particularly in China’s rapidly growing AI market.
Production Shift Details
According to multiple sources, Nvidia requested TSMC to reallocate manufacturing capacity away from H200 production toward Vera Rubin chips, which are expected to deliver substantial performance improvements over the current Blackwell architecture. This decision comes as Nvidia develops the B30A, a new chip based on the Blackwell architecture intended to succeed the H20 in the Chinese market.
The production suspension is not due to any technical defects in the H200 chips, but rather reflects geopolitical and regulatory considerations, including export restrictions and ongoing tariff negotiations. The United States announced limited export approvals for newer chips to China in early 2026, with shipments either minimal or blocked entirely.
Market Implications
Nvidia has effectively written off the China market in its forecasts, citing inspections and 25% tariffs on H200 chips. The stock dipped approximately 1.5% on the news as investors processed the implications of the company’s strategic shift.
The decision also signals Nvidia’s confidence in the Vera Rubin platform, which was first announced earlier this year. The company appears to be betting that customers will prefer waiting for the next-generation hardware rather than purchasing current-generation chips that face regulatory uncertainty.
Broader Industry Context
This development occurs amid a broader restructuring of the AI chip supply chain. Competitors including AMD and Intel continue to gain ground in certain segments, while custom silicon efforts from major cloud providers like Google, Amazon, and Microsoft create additional competition.
The shift also highlights the increasing importance of manufacturing capacity in the AI era. With TSMC’s advanced nodes in high demand, chipmakers are making tough prioritization decisions that could shape the industry’s trajectory for years to come.