Beijing-based AI chip startup SUNMMIO has completed its Pre-A2 and Series A financing rounds, bringing its total funding to nearly CN¥2 billion over the past year. The latest round was led by Beijing state-backed investment platforms including Yizhuang State Investment. The company plans to use the capital to expand its research and development team, advance its core technology, and accelerate mass production of its proprietary 3D TokenPU processors.
Funding Details and Investor Support
The Pre-A2 and Series A rounds were led by Yizhuang State Investment alongside other municipal and district-level state platforms in Beijing. Earlier financing over the past year attracted state investors such as CDB Capital, Beijing Industrial Investment, Beijing Shunxi, and Zhongguancun Science City. Market-oriented and industrial investors including Source Code Capital, Xianghe Capital, Jiayu Capital, Stone Capital, Legend Capital, and Zhiwei Capital also participated, with some existing backers increasing their commitments.
Beyond financial support, the state-backed investors will provide policy resources and industrial connections within the Beijing E-Town area. They aim to connect SUNMMIO with manufacturing, packaging, and large-model application partners to strengthen supply chain coordination and capacity planning. This collaboration is expected to accelerate the transition from chip design to large-scale commercial deployment.
The company has completed several rounds over the past year, backed by national and municipal state platforms, private institutional investors, and industrial funds. This layered investor base reflects confidence in SUNMMIO's technology roadmap, production capability, and commercial prospects. The funding also provides a stronger foundation for advancing product development and deepening collaboration in the domestic 3D IC supply chain.
3D TokenPU Technology and A4E Progress
SUNMMIO develops 3D-stacked chips designed specifically for large-model inference. Its first-generation A4E chip completed tape-out in June 2026, stacking eight memory wafers above a logic wafer through wafer-on-wafer hybrid bonding. The design provides a stated memory bandwidth of 16 terabytes per second, directly targeting the data-transfer bottlenecks that currently limit AI inference performance.
The company describes its 3D TokenPU architecture as a native design for the large-model era, focused on extreme energy efficiency in token processing. The A4E uses a mature domestic manufacturing process and supply chain, and the company says the architecture breaks through memory wall constraints. SUNMMIO reports that its core team has nearly seven years of 3D IC experience, including volume production of more than 10,000 wafers and cumulative revenue exceeding CN¥1.2 billion.
Strategic Outlook for AI Inference
SUNMMIO positions itself as a chip company built for the large-model era, bypassing the traditional general-purpose GPU path to focus on token-level processing efficiency. The new funding will support R&D team expansion and faster volume production of its 3D TokenPU products. The company also emphasizes tile-native hardware and software co-design as a way to address AI compute, memory, and communication bottlenecks simultaneously.
Wang Fuquan, founder and CEO of SUNMMIO, has positioned the company to avoid the conventional GPU route and concentrate on token-level energy efficiency for inference. Yizhuang State Investment Chairman Chen Zhicheng said the investment aligns with Beijing's ambition to build an artificial intelligence and integrated circuit cluster. Both leaders emphasized deeper collaboration with manufacturing, packaging, and application partners in the Beijing E-Town ecosystem.
The latest financing underscores growing investor confidence in China's domestic AI chip innovation and advanced packaging supply chains. With state-backed platforms and industrial funds supporting its commercialization, SUNMMIO is preparing to scale its A4E chips for a rapidly expanding large-model inference market. Successful delivery from tape-out to reliable volume production will be a key milestone as China builds out its AI infrastructure.
Source: EE Times China