Houmo AI Completes Several-Billion-Yuan Series B Funding; M50 Chip Wins Over 100 Customers

Capital's attention is shifting from cloud computing clusters to the chip inside phones, computers, and robots. According to incomplete statistics, in September alone this year, nearly ten edge-side AI chip companies in China announced the completion of new funding rounds. The latest large financing came from Houmo AI. The reporter learned today from relevant sources that the compute-in-memory AI chip company Houmo AI recently completed a Series B funding round, with a single financing amount of up to several billion yuan, having completed multiple rounds of financing within the past year. This round of financing came from a collective bet by heavyweight upstream and downstream industrial capital, the joint entry of state-owned capital and international investors, and continued increases from multiple existing shareholders, behind which is Houmo AI's accelerated realization from technological breakthroughs to large-scale commercial deployment. It is reported that this round of funds will mainly be used to expand the large-scale delivery of M50, advance the research and development of the new-generation M6X and next-generation 3D CIM products, and continuously improve the software toolchain and industrial ecosystem, promoting the large-scale deployment of edge AI in more terminal scenarios. Capital turns to the edge side, compute-in-memory returns to the center of the table Compute-in-memory is not a new concept. Why is it being chosen simultaneously by capital and terminal manufacturers at this moment? And why is Houmo able to secure continuous capital injections on this track? The answer is first hidden in a set of shipment data. IDC data shows that in the first quarter of 2026, global edge-side/edge AI chip shipments grew 45% year-on-year, while cloud AI chip growth was only 12%. Capital and shipments move in the same direction; demand for edge-side AI chips does not rest only in investors' expectations. Industrial giants are also betting on compute-in-memory with real money. Last December, Nvidia reached a non-exclusive technology licensing agreement with AI inference chip company Groq and absorbed its core team, worth about $20 billion. In October this year, Nvidia also plans to invest in inference chip company d-Matrix, which adopts a digital compute-in-memory architecture; previously, d-Matrix had announced access to Nvidia's AI factory ecosystem through NVLink Fusion. In the past few years, the main axis of large-model competition was on the training side—stacking compute, stacking GPUs, stacking clusters—and compute-in-memory at the architecture level could hardly become the industrial mainline. Now the weight of inference is rising; why is it precisely now? There are three reasons. One layer is that models are becoming smaller and stronger. Taking the Qwen series as an example, to reach similar capabilities, the scale required by Dense models was compressed to about 1/8 within 17 months, and MoE models were compressed to about 1/6 within about 6 months; the models that edge-side chips can carry have risen from the 7B level in 2024 to the 120B level this year. The two curves move toward each other, giving large models a technical foundation for running on terminals. One layer comes from agents. Agents no longer merely wait for users to ask questions, but begin to continuously understand goals, call tools, and execute tasks. More and more tasks need to be completed locally, and terminals have also changed from devices that passively execute instructions into computing entrances that continuously produce and consume tokens. There is also the layer of storage itself. Terminals have limited space, power consumption, and cost, and chips must simultaneously improve computing efficiency, storage capacity, and memory access bandwidth within limited conditions; under the traditional architecture that separates computing and storage, model parameters need to be moved frequently, and the larger the data scale, the more prominent the latency and energy consumption. Storage is also becoming more expensive: according to TrendForce data, the contract price of general-purpose DRAM rose 90%-95% quarter-on-quarter in the first quarter of this year, and is expected to rise another 58%-63% in the second quarter. Deep cultivation of SRAM-CIM for six years, adapting to edge-side needs Houmo AI was founded in 2020 and has always taken SRAM-CIM as its root technology, placing part of the computation where the data resides. When the industry's exploration of compute-in-memory was mostly concentrated in low-compute and ultra-low-power scenarios, Houmo chose the direction oriented toward high compute and edge-side large-model inference. According to data disclosed by the company, after adopting SRAM-CIM, the area of the core computing unit was compressed to about 30% of the traditional architecture, and operating power consumption was reduced to about 10%. Behind this is a team with both academic and engineering strengths. Founder and CEO Wu Qiang holds a PhD from Princeton University in the United States, has worked at AMD and Meta, and served as CTO of Horizon Robotics; the core R&D team has published multiple compute-in-memory papers at top chip conferences such as ISSCC and has completed more than 10 SRAM-CIM chip tape-outs. At the just-held Houmo AI Ecosystem Partner Conference, the company disclosed that its edge-side large-model AI chip M50, less than a year into mass production, has obtained more than 150 product design wins and serves more than 100 customers. M50 has 160 TOPS@INT8 compute, 150 GB/s memory bandwidth, and typical power consumption of 10W; the first 3D CIM chip A20 has entered production, and the new-generation product of the M series, M6X, is under development. Board-level products cover PCIe standard cards and miniaturized modules such as M.2 and MXM; the combination of chips, modules, standard cards, and multi-chip systems addresses the localized operation needs of models from the 8B level to nearly 1,000B. Less than a year into mass production, M50 has entered the edge side with more than 150 design wins Recently, at the CSDN 1024 Programmer's Day and AI Builder Conference, the preview version of Lenovo's AI Host P7 was unveiled. This palm-sized host, weighing only 300 grams, uses Houmo AI's M50 for local large-model inference; M50 provides 160 TOPS of compute at 10W typical power consumption, the whole machine's maximum power consumption is controlled at 30W, and an ordinary power bank can supply power. The P7 preview version can run large models of up to 125B locally and supports 7×24-hour continuous task execution, with data never leaving the device. Lenovo summarizes the goal of this device as helping developers and workplace professionals achieve "token freedom." Houmo AI appeared at WAIC2026. P7 is not an isolated case. The Great Wall N90 Pro laptop can run a 35B-parameter model locally, putting local large models into mobile office scenarios; UGREEN's Home Agent HA100 Pro combines local AI capabilities with NAS storage and smart home devices. However, what explains the situation better than a single complete machine is the number of design wins. Less than a year into mass production, M50 has obtained more than 150 product design wins, covering four major directions: agent computers, intelligent mobile terminals, embodied robots, and edge inference. At the same time, the common problems encountered in more than 150 product design-win projects have also continuously pushed Houmo to fill in software and development support capabilities. On this basis, Houmo AI has open-sourced the "Houmo AI Ecosystem Software," distilling model adaptation, compilation optimization, and development experience into reusable development support to reduce partners' repeated adaptation investment; related open-source projects will be launched successively before January next year. Design wins prove it can sell; open source determines whether it can be replicated at low cost by more partners. IDC predicts that in 2026, global generative AI phone shipments will reach 432 million units, and China's AI phone shipments are expected to reach 147 million units, a year-on-year increase of 31.6%. From the backing of several billion yuan in capital and product deployment with more than 100 customers to the continuous iteration of the compute-in-memory architecture and the open-sourcing of the software ecosystem, Houmo AI is forming a development flywheel driven by technology, industry, and capital. This is enough to explain why capital is rushing to the edge side; what the investor structure answers is why it is this company.