Chinese technology powerhouses Tencent and Alibaba are significantly escalating their investment and deployment of domestically developed artificial intelligence chips. This strategic pivot underscores a broader national push towards self-sufficiency in advanced semiconductor technology, a movement intensified by prior export restrictions that limited access to high-end foreign components. The focus on homegrown solutions has spurred a notable boom in China’s domestic chip manufacturing sector.

Tencent’s Chief Strategy Officer, James Mitchell, indicated a substantial increase in the company’s capital expenditure, particularly in the latter half of the year, as the availability of China-designed chips is expected to grow month by month. Mitchell also noted a progressive ramp-up in the supply of domestically produced graphics processing units (GPUs) throughout the year, with manufacturing expanding both within China and in neighboring countries. Meanwhile, e-commerce giant Alibaba highlighted the scaled mass production of its proprietary T-Head GPU chips, which are integral to its cloud computing data centers. The company views these self-developed semiconductors as a crucial structural advantage in an environment marked by computing scarcity, suggesting potential revenue growth and margin improvement. Alibaba is also exploring opportunities to sell servers equipped with its chips or co-build data centers, signaling its expanding role in the national semiconductor landscape. Several local players, including Moore Threads, MetaX, and Huawei, have intensified their efforts to fill the void left by international suppliers, contributing to record revenues for Chinese chip firms.

This intensified domestic development unfolds against a backdrop of reports suggesting a potential shift in U.S. policy. Recent information indicates that the U.S. government may have granted approval for several Chinese companies, including Alibaba and Tencent, to procure Nvidia’s H200 chips—among the most powerful GPUs available. However, official confirmation remains elusive, with U.S. Treasury Secretary Scott Bessent stating that such approval was “news to him” and falls under the Commerce Department’s purview. It’s also noted that no H200 units have been produced yet.

Experts suggest that as Chinese firms advance towards “agentic AI”—AI systems capable of executing more complex tasks—the demand for highly advanced chips will intensify. Neil Shah, a partner at Counterpoint Research, anticipates that Nvidia’s H200 product would be well-received, particularly as the “race towards Agentic AI” shifts from initial training to massive inference scaling—the practical application of trained AI models. Shah emphasizes that Chinese hyperscalers cannot afford to delay, suggesting that the timing is opportune for the adoption of Nvidia H200s, potentially leading to a hybrid AI inferencing infrastructure incorporating both Chinese and U.S. chips to accelerate development.