MemSilicon raised more than 400 million yuan last October. That figure, roughly $55 million, is a serious check for a hardware company that is not yet a household name [Xinhua, October 2026]. The round, led by Boyuan Capital, bought a stake in a specific technical bet: that resistive RAM, or RRAM, is the right substrate for the next generation of AI inference chips [36Kr, October 2026].
The Bet on Compute-Storage Collaboration
The company's thesis is architectural. Today's large language models are bottlenecked by the movement of data between separate memory and processing units, a problem that worsens as models grow. MemSilicon's focus is on resistive RAM, a non-volatile memory technology that retains data without power and consumes less energy than conventional flash [Reuters, September 2010]. By integrating computation directly with this memory layer,a concept known as compute-storage collaboration,the startup aims to build chips that can run "fully featured" AI models with better speed and cost-performance than current solutions [Futunn, October 2026]. The Series A proceeds are earmarked for two tracks: commercializing RRAM and developing a first-generation "LPU," or large-model processing unit, designed for inference [PERPLEXITY SONAR PRO BRIEF, October 2026].
An Investor Syndicate With Strategic Depth
The capital did not come from generalist funds alone. The investor list reads like a roster of potential partners and customers, spanning automotive, gaming, and China's tech-industrial complex. The participation of strategic names like Bosch Venture Capital, game developer miHoYo, and semiconductor firm GigaDevice signals more than financial interest [36Kr, October 2026]. It suggests these players see a path where MemSilicon's memory and inference technology could slot into future products, from smart sensors to gaming hardware. The full syndicate underscores the project's national strategic importance as well.
| Investor | Type | Notable For |
|---|---|---|
| Boyuan Capital | Lead Investor | Chinese venture firm |
| Bosch Venture Capital | Corporate VC | Global automotive supplier |
| miHoYo | Strategic | Creator of Genshin Impact |
| GigaDevice | Strategic | Chinese NOR flash memory maker |
| Xiaomi Strategic Investment | Corporate VC | Consumer electronics giant |
| Zhongguancun Science City | Government | Beijing tech hub development |
The Graphcore Veteran at the Helm
Concurrent with the funding, MemSilicon installed a new CEO with a relevant track record. Lu Tao, the former president and Greater China general manager for UK AI chip challenger Graphcore, joined as CEO and co-founder in October 2026 [PERPLEXITY SONAR PRO BRIEF, October 2026]. His experience commercializing novel AI hardware in a fiercely competitive market against incumbents like Nvidia is a tangible asset. It provides operational credibility for the daunting task of moving from lab research to shipped silicon. The company's technical roots are academic, originating from a team at the Chinese Academy of Sciences with over two decades of experience in materials and chip design [Sina Finance, October 2026]. Lu Tao's role is to bridge that research with the realities of product definition and customer adoption.
Where the Wheels Could Come Off
The ambition is clear, but the path is littered with the classic hurdles of deep-tech hardware. MemSilicon is pursuing a dual-product strategy in one of the most capital- and expertise-intensive industries. Commercializing a new memory technology like RRAM is a monumental challenge on its own, requiring perfection in materials science, manufacturing yield, and long-term reliability. Simultaneously developing a competitive inference processor, the LPU, pits the startup against some of the best-funded companies on the planet. The risks are not theoretical.
- The manufacturing gauntlet. Moving from lab prototypes to high-volume, reliable semiconductor production is a famously expensive and difficult leap. Any stumble in yield or performance could delay timelines and burn capital.
- The software moat. AI chips are only as good as their software stack. Building the compiler and tools to make the LPU easy for developers to use is a massive software undertaking that must happen in parallel with hardware design.
- The incumbent response. While targeting a new architectural approach, MemSilicon will still be measured against the raw performance and established ecosystems of GPU giants and other AI accelerator startups. Winning design wins requires proving a decisive advantage.
The company's substantial Series A provides a long runway to tackle these problems. The strategic investor base offers potential pathways to early, supportive customers. But the clock is ticking, and the technical milestones ahead are steep.
The $55 million infusion, anchored by Boyuan Capital and joined by Bosch, miHoYo, and GigaDevice, gives MemSilicon the fuel for its ambitious two-track race [Xinhua, October 2026] [36Kr, October 2026]. The question for the next eighteen months is whether its academic foundation and new commercial leadership can translate RRAM's promise into a chip that large-model builders actually want to buy.
Sources
- [Xinhua, October 2026] MemSilicon完成超4亿元A轮融资 宣布战略升级 卢涛出任CEO | http://www.news.cn/finance/20261020/4c658288a276478eb73ed062510d47fe/c.html
- [36Kr, October 2026] Mingxin Qirui completes Series A financing of over RMB 400 million | https://eu.36kr.com/en/newsflashes/4032399614481033
- [Futunn, October 2026] MemSilicon Announces Completion of Over RMB 400 Million in Series A Financing | https://news.futunn.com/en/post/1001511496/memsilicon-announces-completion-of-over-rmb-400-million-in-series
- [PERPLEXITY SONAR PRO BRIEF, October 2026] Company, product, and funding summary | Sourced from web-grounded brief.
- [Sina Finance, October 2026] MemSilicon完成超4亿元A轮融资，宣布战略升级 | https://portal.sina.com.hk/technology/sina/2026/10/19/1959466/memsilicon%E5%AE%8C%E6%88%90%E8%B6%85%E4%BA%BF%E5%85%83a%E8%BC%AA%E8%9E%8D%E8%B3%87%EF%BC%8C%E5%85%AC%E5%8F%B8%E6%88%B0%E7%95%A5%E5%85%A8%E9%9D%A2%E5%8D%87%E7%B4%9A/
- [Reuters, September 2010] Resistive RAM background | Sourced for technical definition.
- [Tracxn, retrieved 2026] Funding amount conversion | Used for USD estimate.