Moxin AI closes ~¥1B Series C as China's sparse computing chip bet scales
What's the deal? Moxin AI, a Chinese AI chip startup specialising in sparse computing, has closed a Series C round of nearly ¥1B (roughly $140M). The round drew state-backed investor Shenzhen Capital Group alongside industry players including Yanshan TechnologyDealroom has a profile for this one. Try Dealroom →, Greater Bay Area Common Home, Liding CapitalDealroom has a profile for this one. Try Dealroom →, and Yunsheng CapitalDealroom has a profile for this one. Try Dealroom →. Existing backers Triumph CapitalDealroom has a profile for this one. Try Dealroom →, Chuangxiang Investment, and Shengjing JiachengDealroom has a profile for this one. Try Dealroom → also followed on.
Moxin said it will use the funds to develop and mass-produce SparsePrime, a next-generation high-performance AI inference card set to launch by year-end. The card is built on Moxin's proprietary Antoum 2.0 chip architecture and targets data centres running large language models.
Why now? China's AI inference demand has surged to 10–15 times the volume of training compute, according to multiple industry estimates. Daily token calls in China hit 140 trillion by March 2025 — a thousand-fold increase from early 2024.
That explosive growth has turned inference into the main battleground for AI chip makers. At the same time, US export controls on advanced Nvidia GPUs have intensified demand for domestic alternatives, pushing capital into homegrown chipmakers at a rapid clip.
What could go wrong? Moxin's sparse computing approach — which strips redundant calculations from neural networks at the hardware level — remains a niche architecture. Winning benchmark titles is one thing; convincing large-scale customers to shift away from established GPU ecosystems is another. The company expects to reach breakeven only in 2026 or 2027, leaving a long runway of cash burn ahead.
Scaling from thousand-card cluster deployments to nationwide coverage across cloud, edge, and compute networks will also test Moxin's supply chain and engineering resources.
The signal: The participation of Shenzhen Capital Group — one of China's largest state-backed investment funds — alongside a mix of corporate and market-oriented investors underscores how domestic AI chip development has become a strategic priority, not just a venture bet. With inference demand now dwarfing training compute by an order of magnitude, the economics increasingly favour specialised, cost-efficient architectures over general-purpose GPUs — a structural shift that plays directly to Moxin's sparse computing pitch.
Read more: 36Kr
