Log in Sign up
Gemini Google Takes Dreambeans to India, Its First Market Outside the US, as American Downloads Slump Business Chinese Sparse-Computing Chip Startup MoFETT AI Raises Nearly 1 Billion Yuan in a C-Plus Round News Qualcomm Validates HBC Memory Stacks That Push 133 TB/s Toward Its AI250 Accelerator ChatGPT Altman Says the Model That Cracked Navier-Stokes Is the Next GPT, and It Is Coming Soon News Embodied-AI Startup RoboScience Closes Another Funding Round as 360's Strategic Arm Backs Its Push Into Logistics and Retail Robots Security Anthropic Tells Investors It May 'Imminently' Cross Its Recursive Self-Improvement Threshold — and Might Only Notice After It Happens AI Agents Trajectory Raises 45 Million Dollars From Sequoia and Nvidia to Let Nontechnical Experts Fine-Tune AI Models Business Salesforce Pays 2.7 Billion Dollars for Cresta in Its Second AI Contact-Center Acquisition This Fall Gemini Google Takes Dreambeans to India, Its First Market Outside the US, as American Downloads Slump Business Chinese Sparse-Computing Chip Startup MoFETT AI Raises Nearly 1 Billion Yuan in a C-Plus Round News Qualcomm Validates HBC Memory Stacks That Push 133 TB/s Toward Its AI250 Accelerator ChatGPT Altman Says the Model That Cracked Navier-Stokes Is the Next GPT, and It Is Coming Soon News Embodied-AI Startup RoboScience Closes Another Funding Round as 360's Strategic Arm Backs Its Push Into Logistics and Retail Robots Security Anthropic Tells Investors It May 'Imminently' Cross Its Recursive Self-Improvement Threshold — and Might Only Notice After It Happens AI Agents Trajectory Raises 45 Million Dollars From Sequoia and Nvidia to Let Nontechnical Experts Fine-Tune AI Models Business Salesforce Pays 2.7 Billion Dollars for Cresta in Its Second AI Contact-Center Acquisition This Fall

Chinese Sparse-Computing Chip Startup MoFETT AI Raises Nearly 1 Billion Yuan in a C-Plus Round

Business · By: AIReport Admin · 1 hour ago · 👁 11
Chinese Sparse-Computing Chip Startup MoFETT AI Raises Nearly 1 Billion Yuan in a C-Plus Round

MoFETT AI, a Shenzhen startup building sparse-computing inference chips, has closed a C-Plus round of nearly 1 billion yuan, about five months after a similarly sized round that took its total funding to roughly 2 billion yuan. The company says third-quarter revenue grew severalfold, and its SP200 accelerator board is due by year-end. Sparse computing skips unnecessary multiplications so the same silicon serves more inference.

MoFETT AI, a Shenzhen chip startup that has spent eight years on sparse computing, announced it has closed a C-Plus funding round of nearly 1 billion yuan, roughly 140 Million Dollars. The round came just five months after a previous round of similar size, taking total funding to about 2 billion yuan.

Investors include Ren'ai Capital, Haitang Fund and Dongheng Petroleum alongside other industrial players, as well as financial institutions Biqiu Capital, Paila Capital and Zhongcai Capital. Proceeds go toward a full-stack upgrade of the company's sparse-computing inference chips, pre-research on the next generation of silicon, and a newly formed industry-academia alliance for sparse computing.

Sparse computing attacks the cost problem of AI inference at the source of computation: by exploiting sparsity in weights and activations, the chip identifies and skips multiplications that contribute nothing to the result, so the same silicon serves more tokens. With MoE architectures making inference workloads more conditional and sparse by design, the approach has also attracted policy backing — China's Ministry of Industry and Information Technology and the National Development and Reform Commission recently named dynamic sparse computing and tightly coupled software-hardware inference chips as priority directions.

The company says its third-quarter revenue grew several times over, though it did not disclose absolute figures — a company-reported claim. Its inference clusters, built on in-house silicon, are deployed across four regions of China and have been incorporated into a top telecom operator's computing services. MoFETT also took part in what it describes as China's first offshore computing token factory, a Shanghai state-backed project that pairs offshore wind power and seawater cooling with sparse inference to cut per-token cost.

On the product side, the SP200 accelerator board is due by the end of the year, and an upgraded SACA software stack is imminent. The company recently reported that adapting two mainstream remote-sensing foundation models to its cards raised inference throughput by up to 2.88 times and compressed the models by up to 7.64 times without retraining — again, company-reported results.

Sources: Jiemian News; Beijing News; ChinaVenture; company statement.

Comments (0)

Log in to join the discussion

Log In

No comments yet