Nvidia is in discussions with TSMC to significantly increase production of its H200 AI accelerators, as demand from Chinese technology companies far exceeds current supply.
According to people familiar with the matter, Nvidia has approached TSMC about securing additional manufacturing capacity for the H200, its most advanced data center GPU currently cleared for limited export to China .
Orders from China exceed two million units
Chinese customers have reportedly placed orders for more than two million H200 chips for delivery starting in 2026, dwarfing Nvidia’s available inventory of roughly 700,000 units. The surge reflects aggressive investment by Chinese firms in large-scale AI infrastructure, even as domestic chip alternatives remain under development .
The H200 offers a substantial performance increase over Nvidia’s earlier China-compliant accelerators, making it a key option for training large AI models under current export rules.
Production ramp targets mid-2026
Sources say Nvidia and TSMC are exploring ways to begin expanded H200 production as early as the second quarter of 2026. Final volumes have not been agreed, and discussions remain ongoing, with capacity constraints at TSMC acting as a limiting factor .
TSMC already manufactures Nvidia’s latest-generation AI chips, including parts of the Blackwell platform, which could complicate allocation decisions as demand grows across multiple product lines.
Pricing and regulatory uncertainty remain
Nvidia reportedly prices the H200 at around $27,000 per unit, positioning it as a premium accelerator for large data centers. While U.S. policy now allows controlled exports of the chip to approved Chinese customers, Chinese regulatory authorities have not yet completed formal approvals for large-scale imports .
Any changes in U.S.-China technology policy could still affect shipment timelines or volumes.
Market impact and outlook
Investors responded positively to reports of the production talks, with Nvidia shares rising as expectations grew around sustained AI hardware revenue. Analysts see China demand as a short- to medium-term growth driver, even as geopolitical risks continue to shape long-term strategy.
If Nvidia secures additional capacity at TSMC, the move could ease near-term supply pressure but may intensify competition for advanced semiconductor manufacturing slots across the AI industry.
What’s next will depend on how quickly Nvidia and TSMC can finalize capacity commitments, and whether regulatory conditions on both sides remain stable through 2026.
Discussion (0)
Be the first to comment.