In mid-April, an investment firm most retail investors have never heard of quietly set a record. It did not shout about a single breakthrough. It simply closed a fund—and the size of that fund said more about where China's AI money is heading than any product launch this year.
The firm is GL Ventures (光合创投), the early-stage arm built from what was once Lightspeed China. The person behind the raise is Michael Mi (宓群), a founding partner who has now spent fifteen years backing Chinese technology from the seed stage.
Fifteen years, one bet
Mi is not a newcomer to this cycle. The institution's roots go back to 2011, when Lightspeed China was set up as an independently run local vehicle. It raised its first renminbi fund in 2016, and in 2021 it rebranded as GL Ventures (光速光合). Through two booms and a long fundraising drought, Mi kept the same thesis: get in early, sit with the company for years, and let the portfolio compound.
That patience is now paying off visibly. In the past year alone, GL Ventures saw six of its holdings go public. They include Hanshi Technology (汉朔科技), a store-digitalization leader on the Shenzhen exchange, and Muxi (沐曦股份), a GPU chip designer that debuted on the STAR Market. Early in 2026, Zhipu (智谱)—billed as the first AI large-model (大模型) stock—listed in Hong Kong. The firm also points to a pipeline of nearly ten more listings expected in 2026, among them Unitree (宇树科技), the humanoid-robot maker that has cleared the path to an IPO, and METi (剂泰科技), an AI drug-delivery company.
The 5-billion-yuan raise
The new vehicle is a dual-currency fund. Its sixth early-stage dollar fund closed at $460 million (about 3.3 billion yuan), and its third renminbi fund carried a 2 billion yuan target with a first close of 1 billion yuan already in hand. Together the two funds exceed 5 billion yuan (roughly $700 million), and both were oversubscribed—the latest in a run of consecutive oversubscribed rounds. Industry trackers called it the largest China VC raise of the opening months of 2026.
After this close, GL Ventures said its investable capital topped 7 billion yuan. The dollar fund's limited partners are a global mix: sovereign funds, university endowments, charitable foundations, pension funds, fund-of-funds, top family offices and large technology firms. The firm noted that five new sovereign funds joined this time. The renminbi side drew science-and-technology guidance funds, industrial investors, market-oriented FoFs, university foundations and founders from the firm's own network.
What the money traces back to
The returns that made this raise possible trace back to a specific reading of the AI stack. GL Ventures did not bet only on one layer. It backed compute and models and the next terminal form at once: the optical-module leader Eoptolink (中际旭创), the GPU designer Muxi, the humanoid-robot maker Unitree, and the large-model (大模型) developer Zhipu, plus Shengshu (生数科技, the video-model company), the AI biology firm Stellar Bio, and toolmakers such as OpusClip.
Mi's line is that AI is a five-to-ten-year track, not a quarterly trade. The firm's pattern is to enter before the inflection is obvious, then stay through the public listing. On Unitree, he has said the team tracked the company for years and moved only when the technical turning point was clear.
According to Li An, chief scientist at BrainNet (脑机网), the signal from a raise this size is less about one firm and more about where long-term capital now wants exposure: at the infrastructure and application layers of Chinese AI, not just the headline models.
Why a dollar-and-renminbi fund
The dual structure is not cosmetic. A dollar fund can bring global LPs and cross-border optics; a renminbi fund can line up with local industrial policy and domestic listings. For a portfolio that spans chips, robots and models—some of which will list in Hong Kong or Shanghai, some of which may never touch a foreign exchange—having both currencies lets the firm support a company across its whole life without forcing a refinancing at the wrong moment.
Mi has said the firm deliberately capped the final size even after exceeding its targets, to protect operating efficiency. Whether that restraint holds through the next cycle is a question only the 2026 vintage's returns can answer.
Honest limitations
This profile relies on fund-announcement reporting and portfolio disclosures, not audited performance figures. "Investable capital above 7 billion yuan" and "nearly ten 2026 IPOs" are the firm's own statements, repeated by financial media; they are not independently verified outcomes. The largest-deal claim depends on how one counts the year and the peer set. And a strong recent track record says little about the new fund's results, which will not be knowable for years. Readers should treat the raise as a meaningful sentiment signal, not proof of future returns.
What readers can do now
- If you are an LP or allocator, note that global sovereign and pension money is again entering China AI via credible early-stage vehicles—watch follow-on raises for confirmation.
- If you build or run an AI company, the pattern here is "infrastructure plus application, held long"; match your fundraising story to that layered thesis.
- If you follow the China AI market, treat the size of this fund as a proxy for where patient capital is concentrating in 2026: compute, robots and models together.