Developing personalize our customer journeys to increase satisfaction & loyalty of our expansion recognized by industry leaders.
Artificial intelligence has rapidly become the backbone of global technological competition. Governments, corporations, and research institutions are investing billions to build the computing infrastructure that will power next-generation AI systems.
On February 28, the city of Hangzhou took a significant step toward strengthening its position in this race by hosting a major AI development summit at the Hangzhou Civic Center. The event brought together policymakers, technology companies, and investors with one clear objective: transforming Hangzhou into one of China’s leading artificial intelligence hubs.
During the summit, 12 AI-focused projects were signed, representing a combined investment of CNY 25.5 billion (approximately US$3.71 billion). The projects cover a wide range of initiatives, including AI computing infrastructure, chip research and development, intelligent data platforms, and enterprise AI applications.
Among these agreements, one project stood out for its direct impact on AI computing hardware.
The only project specifically dedicated to AI inference GPUs was signed by Sunrise, a company focused on high-performance GPU and inference chip development.
The initiative, titled “High-Performance GPU and Inference Chip R&D Project,” will support Hangzhou’s broader computing infrastructure strategy. The goal is to strengthen the city’s ability to support AI training, inference workloads, and large-scale industrial applications.
According to Sunrise Co-CEO Wang Zhan, the next stage of AI development will not be defined solely by model innovation.
Instead, computing power will be the decisive factor.
In his remarks at the summit, Wang emphasized that:
“Future industrial development will depend less on model breakthroughs and more on computing capacity and deployment infrastructure.”
This perspective reflects a growing consensus across the global AI industry: the real bottleneck is no longer algorithms — it is compute.
The agreement also marks a new phase of Sunrise’s expansion within Hangzhou. The company stated that the project will contribute directly to the city’s AI computing infrastructure and innovation ecosystem.
Sunrise has an interesting corporate background. The company was originally the large-chip division of SenseTime, one of China’s major AI software firms.
At the end of 2024, the unit was spun off as an independent entity focused entirely on AI hardware development.
Today, Sunrise is developing:
The company claims that its chip designs have already gone through multiple product generations and validation cycles.
Some public statements reference “ten-thousand-card-scale deployments,” although these claims largely originate from corporate and regional disclosures rather than independent third-party verification.
Nevertheless, the project reflects a broader trend: local governments are actively supporting domestic AI chipmakers to build national computing capacity.
But Sunrise is only one part of a much larger ecosystem.
For years, the global AI computing market has been dominated by companies such as Nvidia. NVIDIA’s GPUs power a majority of the world’s large-scale AI training infrastructure.
However, geopolitical restrictions and export controls have accelerated China’s efforts to develop a domestic alternative ecosystem.
Instead of relying on a single replacement supplier, China is now building a multi-vendor AI chip ecosystem.
This approach spreads risk, accelerates innovation, and increases supply resilience.
At the center of China’s domestic AI hardware strategy is the Ascend series from Huawei.
The Ascend platform includes processors such as:
These chips are widely considered China’s primary domestic alternative for large-model AI workloads.
The Ascend platform is designed to support:
Reports indicate that Huawei is preparing large-scale shipments of Ascend processors to meet growing demand for domestic AI hardware.
However, supply constraints remain.
Estimates from industry analysts suggest that Ascend shipments in 2025 may reach around 200,000 units, a meaningful number but still far below global GPU demand.
Manufacturing capacity, advanced semiconductor processes, and ecosystem maturity continue to limit expansion.
Still, the platform has already been deployed across multiple Chinese cloud and enterprise AI projects.
Beyond Huawei, a growing group of Chinese semiconductor companies is developing AI accelerators and inference processors.
One of the most prominent is Cambricon, known for its MLU series AI accelerators.
These processors have appeared in several government procurement programs and enterprise infrastructure deployments aimed at strengthening domestic technology capabilities.
Another major example comes from the telecommunications sector.
According to infrastructure reports, the China Unicom data center in Xining, Qinghai, has deployed multiple domestically developed AI chips, including processors from:
This deployment demonstrates that Chinese AI chips are moving beyond prototypes and into real-world, large-scale infrastructure projects.
Future expansion phases of the same data center are expected to include processors from:
Enflame, in particular, has attracted significant attention in China’s semiconductor sector. The company is reportedly advancing plans for a STAR Market IPO, indicating strong investor interest in AI computing infrastructure.
While GPUs and accelerators power AI training, traditional CPUs still play a critical role in servers, cloud infrastructure, and enterprise computing.
China has therefore also accelerated the development of domestic CPU architectures.
Two key players are:
These companies are increasingly deployed in:
China has gradually restricted the use of foreign processors in certain government projects, encouraging the adoption of domestic alternatives instead.
This policy shift aims to reduce reliance on companies such as:
Although domestic CPUs do not directly replace AI training accelerators, they play an important role in building a fully independent computing stack.
China’s AI computing push is happening at an unprecedented scale.
Some key indicators include:
Cities such as Hangzhou, Shenzhen, Beijing, and Shanghai are competing to become regional AI superclusters.
Hangzhou’s latest summit signals the city’s ambition to join the top tier of AI development hubs.
The signing of Sunrise’s GPU project illustrates how local governments and emerging chipmakers are collaborating to build domestic AI infrastructure.
However, the bigger story lies in the diversification of China’s AI hardware ecosystem.
Current deployment trends show that:
• Huawei Ascend remains one of the few domestic platforms capable of supporting large-scale AI training workloads.
• Chips from Alibaba T-Head, MetaX, Biren, and Cambricon are already appearing in major data center deployments.
• Moore Threads and Enflame are emerging as additional supply sources for cloud AI infrastructure.
• Domestic CPUs from Loongson and Phytium are expanding in government and enterprise server environments.
Taken together, these developments indicate that China is moving toward a multi-vendor AI chip ecosystem rather than relying on a single national champion.
Despite rapid progress, China’s AI chip industry still faces several structural challenges:
1. Manufacturing Constraints
Advanced semiconductor production remains limited by access to leading-edge fabrication technology.
2. Software Ecosystem Maturity
Developer ecosystems for domestic AI chips still lag behind Nvidia’s CUDA platform.
3. Global Supply Chain Integration
International partnerships and supply networks remain complex due to geopolitical tensions.
4. Computing Scale
Even with multiple vendors, the overall computing capacity still trails the infrastructure available in the United States.
Hangzhou’s $3.7 billion AI investment program represents more than just a regional development initiative.
It is part of a broader national effort to build:
As AI increasingly becomes the foundation for industries such as autonomous driving, robotics, biotech, and smart cities, the availability of computing power will define which economies lead the next technological era.
