London: Nvidia, the U.S.-based leader in artificial intelligence hardware, is reportedly considering increasing production of its powerful H200 AI accelerators in response to unexpectedly strong interest from technology firms in China, according to sources familiar with the matter. The deliberations come as Beijing’s companies seek access to advanced computing hardware capable of powering large-scale AI workloads.
The H200, part of Nvidia’s previous generation Hopper-based lineup, offers significantly higher performance than the versions tailored for restricted markets. Its computational capabilities are reported to be roughly two to three times greater than China’s most advanced domestically produced AI accelerators, making it highly attractive to cloud service providers and enterprise users.
This surge in interest follows a notable shift in U.S. policy. Earlier this month, the Biden administration agreed to allow Nvidia to export H200 chips to China under a licensing regime, with a 25% levy on sales to authorised customers a decision that marked a departure from previous strict export controls on advanced AI semiconductors.
Major Chinese companies, including e-commerce giant Alibaba and tech powerhouse ByteDance, have reportedly contacted Nvidia in hopes of placing sizeable orders for the H200. These enquiries have highlighted just how robust demand has become, with some Chinese firms indicating their interest could outstrip the current volume of chips available.
Nvidia’s current production capacity for the H200 is constrained, in part because the company is prioritising manufacturing resources for its latest Blackwell series and the upcoming Rubin architecture, which represent the next evolution of its AI hardware. The H200 is manufactured using Taiwan Semiconductor Manufacturing Company’s (TSMC) 4-nanometre process, a technology also in high demand from other global customers.
An Nvidia spokesperson emphasized that the company is managing its supply chain carefully to ensure that licensed sales to China “will have no impact” on fulfilling commitments to customers in the United States and other regions. However, expanding production capacity poses logistical and technical challenges at a time when global demand for advanced chips remains extraordinarily strong.
Despite the U.S. policy change, the Chinese government has yet to authorise any official purchases of the H200. Sources suggest that Chinese officials have held internal discussions to weigh the implications of allowing imports of the powerful chips particularly as Beijing also seeks to foster the development of its own domestic AI semiconductor industry. One proposal reportedly under consideration would require foreign chip acquisitions to be bundled with purchases of locally made accelerators.
The debate underscores the broader tension between China’s ambitions to advance its AI capabilities and its strategic imperative to support homegrown technology. If approval is granted, and if Nvidia does increase H200 production, it could reshape competitive dynamics in the global AI chip market, offering Chinese firms access to hardware that significantly outperforms locally produced alternatives.
At the same time, Nvidia must navigate broader geopolitical sensitivities, supply chain limitations, and competition for advanced manufacturing capacity particularly from other technology leaders also eyeing scarce semiconductor fabrication resources. How the company balances these pressures will be closely watched by industry analysts and markets alike.