On November 3, 2026, sparse-computing chip company Moxin (墨芯) announced it had closed a Series C+ funding round of nearly 1 billion yuan. Just five months after a previous round of nearly 1 billion yuan, the two rounds together total nearly 2 billion yuan. Investors in this round included Renai Capital, Haitang Fund, and Dongheng Oil from the industrial side, as well as Bichu Capital, Paila Capital, and Zhongcai Capital. The funds will go toward the full-stack upgrade of its sparse, specialized inference chips, pre-research on its next-generation chip, and building an industry-academia alliance for sparse computing. The real question this round has to answer is not the amount, but whether the products can keep shipping.
According to information from the company disclosed via the Chinese financial outlet Pedaily (投资界), revenue grew several-fold in the third quarter of 2026, though no specific figure was published. Inference clusters built on its own chips have been deployed in China's northwest, southwest, east, and north regions, and, in cooperation with a leading telecom operator, its sparse-computing inference offering has been folded into that operator's compute service system. The idea behind sparse computing is to skip calculations in model inference that contribute nothing and compute only what is useful, thereby lowering compute costs; Moxin describes its approach as algorithm, software, and hardware designed together. Whether that claim holds will be decided by cost and stability under real business workloads, not by the pace of fundraising.
Two products due before the end of the year
On the product side there are two clear dates: the next-generation SP200 board is due before the end of the year, and the upgraded SECA software stack is also about to be released, with the goal of lowering the barrier to deploying large models for inference and running clusters. For buyers, those two milestones matter more than the funding news: whether the board ships on time, and whether the software stack lowers migration costs, will decide directly whether sparse computing is a money-saving solution or one more system to maintain. The funding announcement disclosed neither a post-money valuation nor new customer names, gaps that will have to be filled by later company disclosures or public information from the customer side.
Competition among sparse-computing chip makers is not about a single funding round, but about who first gets cluster scale, a software ecosystem, and operator channels working at the same time. By bringing industrial investors into this round, Moxin has tied distribution channels and application scenarios into its shareholder base; if the SP200 slips or the software stack lands flat, the window opened by the funding will close quickly. That large model companies are looking for compute beyond Nvidia — as in reports that OpenAI plans to buy Google TPU chips — shows the opportunity window for Chinese chip companies is real, but only open to those that can deliver on time.
