Hangzhou Pivots to Multi-Vendor AI Chips in $3.7 Billion Infrastructure Push

2026-03-06 global

Hangzhou, Friday, 6 March 2026.
Hangzhou committed $3.7 billion on February 28, 2026, to a multi-vendor AI infrastructure strategy, effectively insulating its technology sector from foreign supply chain disruptions by prioritizing diverse domestic chip sources.

Investing in the Foundation

The agreement, finalized on February 28, 2026, at the Hangzhou Civic Center, comprises 12 distinct projects with a total valuation of CNY 25.5 billion (US$ 3.71 billion) [1][2]. This capital injection is designed to position Hangzhou as a premier hub for artificial intelligence, but the composition of the deal reveals a more defensive posture regarding hardware procurement. Anchoring these investments is Sunrise, a high-performance GPU developer spun off from SenseTime in late 2024, which signed a dedicated R&D project for inference chips during the summit [1][2]. Sunrise Co-CEO Wang Zhan emphasized the strategic necessity of this infrastructure, noting that “computing power, not model capability alone, will determine future industrial development” [1][2]. This statement reflects a broader industry realization that without robust, accessible compute capacity, the theoretical advancements in AI models cannot be translated into industrial output.

Strategic Diversification of Silicon

The significance of Hangzhou’s approach lies in its departure from relying on a single dominant supplier. While Huawei’s Ascend series, particularly the 910C, remains a primary platform for large-model training in China [1][2], the ecosystem is aggressively widening to include alternative domestic architectures. This follows a precedent set on September 17, 2025, when a China Unicom data center in Xining deployed a mix of chips from Alibaba’s T-Head, MetaX, and Biren [1]. Looking ahead, manufacturers like Moore Threads and Enflame Technology are slated to supply subsequent phases of such cloud projects [2]. This multi-vendor orchestration is not merely a technical preference but a survival mechanism; by cultivating a heterogeneous chip environment, Chinese municipalities are mitigating the risks associated with external supply chain caps, which US officials estimated could limit domestic AI chip shipments to approximately 200,000 units in 2025 [1].

The ‘AI+ Manufacturing’ Imperative

This infrastructure push aligns with the central government’s intensified focus on integrating AI with the real economy, a theme dominating the “Two Sessions” currently underway in Beijing. On March 5, 2026, Minister of Industry and Information Technology Li Leheng declared that “AI+ Manufacturing” is a “must-answer question, not a multiple-choice question” for the nation’s industrial sectors [6]. The objective is to move beyond abstract model development to tangible industrial applications. For instance, Hunan Changsha Shengxiang Biotechnology Co., Ltd. has utilized AI tools to compress enzyme development cycles from months to mere weeks [3]. Similarly, educational institutions like Hangzhou Baochuta Experimental School have already integrated general AI courses where students program robot dogs, signaling a long-term commitment to workforce readiness [3].

Economic Context and Future Outlook

The urgency to solidify this digital foundation is underpinned by the country’s macroeconomic performance. In 2025, China’s GDP grew by 5.0% to 140.19 trillion yuan, with high-tech product exports rising by 13.2% [4]. To sustain this growth, officials are prioritizing the construction of “super-scale intelligent computing clusters.” Chen Changsheng, a member of the Government Work Report drafting team, stated on March 5, 2026, that the country must “lay a solid foundation for AI development,” explicitly linking the future of AI to energy and hardware infrastructure [6]. By securing a diverse array of domestic GPU suppliers, Hangzhou is effectively executing this national mandate, ensuring that its digital economy retains the computing power necessary to compete globally despite the tightening constriction of international semiconductor trade.

Sources

Semiconductors AI Infrastructure