China’s AI Chip Surge Amidst Nvidia H200 Uncertainty

The nation’s tech giants are signaling a significant ramp-up in the production and deployment of homegrown chips this year. This strategic push for self-sufficiency comes at a crucial juncture, with reports suggesting the potential return of U.S. chip titan Nvidia to the Chinese market.

Internet behemoth Tencent indicated on Wednesday that the output of domestically developed Chinese chips is poised for substantial growth in the coming year. Concurrently, e-commerce leader Alibaba detailed its aggressive expansion in the utilization of self-designed semiconductors.

These pronouncements underscore a broader trend: in the face of U.S. export restrictions that have curtailed access to cutting-edge Nvidia technology, China has been compelled to accelerate its development and adoption of indigenous chip capabilities to power its ambitious artificial intelligence initiatives.

Tencent’s Chief Strategy Officer, James Mitchell, articulated a vision of “substantial increase” in capital expenditure, particularly in the latter half of the year, driven by the increasing availability of China-designed chips. He noted that the supply of these homegrown graphics processing units (GPUs) would “progressively” escalate throughout 2024.

Mitchell further elaborated that the manufacturing capacity for China-designed chips is expanding, with production facilities within China and in “neighboring countries” contributing to the supply chain.

The burgeoning domestic chip ecosystem is populated by a growing number of local players who have actively pursued public offerings and product launches. Companies like Moore Threads, MetaX, and Huawei have been at the forefront of efforts to bridge the technological gap left by Nvidia’s restricted access to the Chinese market over the past year. This surge in domestic production has directly contributed to record revenues for Chinese chip manufacturers, fueling a domestic AI boom amidst global technological trade restrictions.

Alibaba’s strategic advantage lies in its proprietary AI chips, which are integral to powering its extensive cloud computing infrastructure. An executive from Alibaba stated during the company’s recent earnings call that “T-Head’s proprietary GPU chips have achieved scaled mass production.” This self-reliance offers a significant competitive edge in an environment characterized by semiconductor scarcity.

The executive further highlighted that “in an environment of compute scarcity, this structural advantage is favorable to our revenue growth and gross margin improvement.” Beyond internal utilization, Alibaba is exploring opportunities to extend its influence in the semiconductor space by offering servers equipped with its custom chips to other enterprises building data centers, or by co-developing such facilities with partners. This indicates a strategic intent to play a more prominent role in China’s evolving semiconductor landscape.

The discussions from Tencent and Alibaba emerged just prior to a Reuters report on Thursday, which indicated that the U.S. had granted approval for several Chinese firms, including Alibaba and Tencent, to procure Nvidia’s H200 chips – among the most advanced GPUs available. However, it’s crucial to note that no H200 chips have been manufactured yet, according to the report.

The definitive status of U.S. government approval for Nvidia’s sales remains somewhat ambiguous. When questioned by CNBC’s Joe Kernen, U.S. Treasury Secretary Scott Bessent stated, “This is news to me.” He acknowledged ongoing deliberations, adding, “I know there’s been a lot of back and forth … and we’ll have to see on that. That’s a Commerce Department function.”

Throughout the past year, there have been intermittent reports regarding Washington’s approvals for Nvidia to supply specific, less powerful chips, such as the H20, to China. However, preceding reports suggested that China was actively encouraging its domestic companies to opt for locally developed alternatives, even when U.S. export curbs were eased.

Neil Shah, a partner at Counterpoint Research, commented that as Chinese enterprises increasingly focus on “agentic AI” – the development of AI systems capable of performing complex tasks autonomously – the demand for more advanced chips will inevitably rise. He believes that Nvidia’s H200 product would therefore be a welcome addition to the market, particularly for the burgeoning inference stage of AI development. Shah noted that while China’s AI roadmap is “pivoting towards ‘domestic-only’ with AI training infrastructure,” the “race towards Agentic AI” is shifting “from training to massive inference scaling.”

Shah elaborated, “Chinese hyperscalers simply cannot afford to wait.” He posited that “the timing is right for NVIDIA H200 to get adopted in what could be a hybrid AI inferencing infrastructure based on Chinese and U.S. chips to scale the infrastructure sooner rather than later.” This suggests a pragmatic approach where China might integrate leading foreign components into its domestic AI infrastructure to accelerate progress.