Moffett AI Raises Nearly RMB 1 Billion Again, Five Months After Its Last Round, to Build Sparse-Computing Inference Chips
Shenzhen chipmaker Moffett AI said on November 3 that it has closed a Series C+ round of nearly RMB 1 billion, about five months after raising a similar amount, taking the two rounds to nearly RMB 2 billion. Moffett designs inference chips that skip unnecessary calculations in AI models, an approach known as sparse computing that China's industry ministry has said it will prioritise. The company says third-quarter revenue grew several-fold and its next SP200 card arrives before year-end.
Moffett AI has raised its second round of nearly RMB 1 billion in five months, on the bet that Chinese AI models will increasingly favour chips that skip work rather than do more of it. Moffett AI, a Shenzhen developer of AI inference chips, said on November 3 that it has completed a Series C+ round of nearly RMB 1 billion, TechNode reported. The round comes about five months after a previous raise of nearly RMB 1 billion, taking the two rounds to nearly RMB 2 billion, according to Jiemian.
What Moffett builds#
Moffett's chips are built around sparse computing, which cuts the calculations needed to run an AI model by skipping parts of the data and operations that do not affect the result. Jiemian's report says demand for this is rising as more large models use mixture-of-experts designs, which activate only part of the model for each request, and that China's Ministry of Industry and Information Technology and National Development and Reform Commission have said they will step up work on dynamic sparse computing.
Investors include Ren'ai Capital, Haitang Fund and Dongheng Petroleum alongside state-backed and financial investors, Jiemian said. The money will go to upgrading Moffett's chips and software, research on its next-generation chip, and a sparse-computing research alliance with universities that it launched in August.
Company claims#
Moffett says it has inference clusters running in four regions of China and has partnered with telecom operators to offer its chips through their computing services. It says third-quarter revenue grew several-fold, without giving figures, and that its next accelerator card, the SP200, will launch before the end of the year, alongside a new software stack called SECA. In tests with a Peking University institute on two remote-sensing models, it reports inference throughput up to 2.88 times higher and models compressed up to 7.64 times with almost no loss of accuracy.
The round follows other large raises by Chinese AI infrastructure companies, including Infinigence AI's RMB 2.3 billion Series C in October.
Limits#
- The round size, revenue growth and benchmark results are Moffett's own claims; it did not disclose a valuation, revenue figures or a lead investor.
- Investor names are translated from Jiemian's Chinese-language report.
- We have not seen independent tests of Moffett's chips.