Hangzhou at an inflection point after the US$3.7b AI GPU deal
hangzhou is pushing to position itself as China’s leading artificial intelligence hub after convening an AI development summit and signing 12 AI projects totaling CNY25. 5 billion (US$3. 71 billion). The summit, held February 28 at the Hangzhou Civic Center, placed special emphasis on computing infrastructure, including one project focused specifically on AI inference GPUs.
What happens when Hangzhou ties AI growth to computing infrastructure?
The city’s summit-centered push sets a clear near-term direction: building out computing capacity as the foundation for industrial-scale AI. Among the 12 signed projects, the only agreement dedicated to AI inference GPUs was Sunrise’s “High-Performance GPU and Inference Chip R& D Project, ” signed in Hangzhou as part of a computing infrastructure initiative.
Sunrise, a Zhejiang-based company, framed the deal as more than a single R& D program. In its statement, Sunrise said the agreement marks a new stage of its Hangzhou expansion and will support the city’s AI computing infrastructure and broader innovation framework.
Company leadership also linked the investment logic to a specific competitive thesis. Sunrise Co-CEO Wang Zhan said computing power, not model capability alone, will determine future industrial development, adding that Hangzhou’s AI strategy offers a supportive environment for the company’s expansion.
Sunrise’s corporate background is also part of why this move is being watched. The company was previously the large-chip unit of SenseTime and was spun off at the end of 2024. It develops high-performance GPUs and multimodal inference chips for commercial deployment, and it cites multiple generations of validated products. At the same time, some scale claims in the public domain are described as originating from corporate and local disclosures rather than independent third-party verification, underscoring the importance of separating signed investment intent from independently confirmed delivery capacity.
What if China’s multi-vendor chip strategy accelerates from local projects to national deployment?
The Hangzhou agreement lands inside a broader shift described as a “multi-player structure” in China’s AI computing supply and substitution strategy. Within that landscape, several domestic chip and accelerator vendors are being cited in connection with procurement, deployment, and capacity-building efforts.
Huawei’s Ascend series, including the 910C, has been cited as a primary domestic platform for large-model training and inference. Large-scale shipments to meet demand for alternatives to Nvidia have been described as a target, while manufacturing capacity and policy constraints are also described as limiting supply. Separately, US officials have indicated domestic production may remain limited, with one estimate suggesting 2025 shipments could be capped at about 200, 000 units, highlighting ongoing supply chain and technology constraints.
Beyond Huawei, Cambricon is frequently cited as a representative domestic AI accelerator vendor, with its MLU series appearing on government and enterprise procurement lists. In addition, a China Unicom data center in Xining, Qinghai was described as deploying multiple domestic AI chips, including products from Alibaba’s T-Head, MetaX, and Biren—an example of real-world deployment inside large-scale infrastructure. The same thread of planned expansion extends to expected later-phase suppliers: Moore Threads and Enflame Technology have been described as expected to supply chips for later phases of the data center and related cloud projects. Enflame has also been linked in Chinese media coverage to progress toward a STAR Market IPO, signaling continued capital market activity alongside product development.
Within that bigger map, Hangzhou’s approach stands out for how it packages local-government coordination, a summit-driven signing moment, and a computing-infrastructure framing. The immediate signal is not that one vendor will dominate, but that local initiatives can serve as on-ramps for multiple domestic providers as deployment expands across different chip categories and use cases.
What if the most challenging constraint is verification and throughput, not intent?
The near-term picture combines strong investment intent with open questions about execution speed and independently verifiable throughput. Sunrise describes itself as developing high-performance GPUs and multimodal inference chips for commercial deployment, and it cites multiple generations of validated products. However, the same context notes that certain scale statements circulating publicly originate from corporate and local disclosures rather than independent third-party verification.
That matters because Hangzhou’s broader ambition—positioning itself as a leading AI hub—implicitly depends on the sustained availability of reliable compute at meaningful scale. Even as local governments provide strategic support for domestic AI computing providers, the wider ecosystem still faces constraints described in connection with manufacturing capacity and policy limits. The practical outcome for buyers of compute, developers, and infrastructure operators may depend on how quickly signed projects translate into stable, deployable systems and how transparently delivery milestones can be confirmed.
For Hangzhou, the headline number attached to the 12 signed AI projects is a strong signal of direction. The next test is operational: whether the computing infrastructure initiative, including the inference GPU R& D effort, can support real-world deployments at the pace implied by the city’s AI-hub ambitions.
For readers watching what comes next, the key takeaway is that Hangzhou is aligning its AI narrative around compute-first industrial scaling, while China’s broader AI chip push is described as multi-vendor and constraint-aware. The balance between ambition and execution will determine how durable this moment proves to be for hangzhou
