- Investors backed AI memory infrastructure and satellite computing startups in the latest funding rounds
- The proceeds will fund AI agents, memory systems and next-generation space computing technologies
AI memory maker MemTensor raises nearly $15 million from pre-Series A
Shanghai-based AI memory startup MemTensor (记忆张量) has raised 100 million yuan ($14.76 million) in a pre-Series A round backed by Huawei Hubble, Honor’s CVC, SenseCapital, Shenzhen Capital Group and MSA Capital.
The company previously raised nearly 100 million yuan in an angel round in May 2025 from investors including Fortera Capital, Suanova Technology and CICC Capital.
Founded in November 2024 and incubated by the Shanghai Institute for Advanced Algorithms, MemTensor is developing technologies to give large language models long-term memory, experience accumulation and continual learning capabilities.
Its technology roadmap spans Memory³ hierarchical memory theory, the MemOS memory operating system and memory-native foundation models.
The company said MemOS ranked first in a benchmark covering 14 mainstream AI memory frameworks while reducing token consumption by nearly 70% on complex long-horizon agent tasks.
The new funding will support further development of MemOS, commercialization of AI agent and memory infrastructure products, deployment across key industries and continued research on memory-native foundation models.
RedBear AI lands Series A+ at $44 million post-money valuation
Shanghai-based AI memory startup RedBear AI (红熊AI) has raised several hundred million yuan in a Series A+ round, valuing the company at nearly 3 billion yuan post-investment.
The round was backed by Jiuwei Fund, Jiaxing Zhangyuan Venture Capital and existing investor Geruifeng, marking the company’s sixth financing round in 15 months.
Image source: RedBear AI
Founded in April 2024, RedBear AI has developed the MemoryBear AI memory system, which integrates ACT-R dual-memory architecture, the Ebbinghaus forgetting curve, semantic pruning and a self-reflection engine to enable dynamic AI memory management.
Earlier this year, the company launched its OpenBear foundation model for enterprise applications and CodeBear coding model, both integrated with MemoryBear to support native AI agent applications.
RedBear AI said it has more than 500 enterprise customers across finance, retail, manufacturing, education, healthcare and internet services, including Fortune 500 companies, with first-half 2026 revenue already surpassing its full-year 2025 total. The company did not specify, though.
The funding will support AI memory research, deeper integration of its models with MemoryBear and expansion into customer service, marketing, education and business intelligence applications.
Stardubhe bags angel funding for satellite AI computing
Stardubhe (星联天枢) has raised tens of millions of yuan in an angel round less than six months after its founding, although the investors were not disclosed.
The Shanghai-based startup develops integrated satellite communications and computing services that allow drones, autonomous vehicles and unmanned vessels to connect directly to satellites through lightweight terminals, process AI workloads in orbit and transmit results back to Earth.
Founder Cheng Yinghui (程英辉) spent more than a decade at China Electronics Technology Group’s No. 54 Institute, where he held senior roles in satellite communications and low-Earth orbit satellite internet projects.
The founding team also includes veterans from CETC, the Shanghai Academy of Spaceflight Technology and the Chinese Academy of Sciences.
The company plans to deliver its first prototypes in September and begin in-orbit validation of its computing payloads, with commercial deployment targeted for 2027.
