Inside a Beijing office park, a team that started in 2020 finally rang a listing bell. The label attached to it was loaded: China's first homegrown GPU company to reach public markets. The story underneath is less about a triumph and more about what it costs to build a graphics chip from scratch in a constrained world.
Key takeaways
- Moore Threads (摩尔线程) debuted on the Shanghai STAR Market (科创板) on 5 December 2025 under ticker 688795.SH, the first domestic firm marketed as a "full-function GPU (全功能GPU)" maker to list (Xinhua, CNR).
- The IPO was unusually fast: the exchange accepted the application on 30 June 2025, the company passed review on 26 September, and it listed about ten weeks later; the acceptance-to-approval stretch was 88 days (Xinhua, CNR).
- It sold 70 million new shares at 114.28 yuan each — the highest issue price among A-share IPOs in 2025 — raising about 8 billion yuan (≈ US$1.13 billion / HK$8.7 billion), with net proceeds of about 7.6 billion yuan (≈ US$1.07 billion / HK$8.3 billion) (Xinhua, CNR).
- Revenue climbed from 46 million yuan in 2022 to 438 million yuan in 2024 — a compound annual growth rate above 200% — then to 702 million yuan in the first half of 2025, which the prospectus says already surpassed the combined full-year revenue of 2022–2024 (Xinhua, CNR).
- The company is not profitable: net losses were 1.84 billion, 1.67 billion and 1.49 billion yuan in 2022–2024 (a combined ≈ US$705 million / HK$5.4 billion), with another 271 million yuan lost in the first half of 2025; its prospectus put cumulative unrecovered losses at about 1.21 billion yuan at the end of 2024 (China Economic Net, citing the prospectus).
What "full-function GPU" actually means here
Most Chinese chip startups chasing AI build narrow accelerators — chips that only do the matrix math behind model training and inference. Moore Threads took a harder, broader path: a GPU that does both graphics rendering and AI compute on one piece of silicon. Its product lines span AI compute cards, professional graphics, desktop graphics, and smart system-on-chips (SoC).
That distinction matters because the two jobs use different software and different customers. A graphics GPU must run game and workstation APIs; an AI chip must run training and inference frameworks. Doing both means maintaining two stacks at once — more engineering, but a product that can sell into PCs, workstations and data centres, not just AI labs.
The pitch lands against a clear backdrop. With advanced chips subject to export restrictions, a domestic supplier of general-purpose GPUs is a strategic asset, not just a commercial one. Moore Threads is one of the few Chinese designers attempting the full-function route rather than a specialised accelerator.
The 88-day sprint
The speed of the listing is the part rivals notice. According to Xinhua and CNR, the timeline ran: application accepted on 30 June 2025, moved into the inquiry stage on 17 July, passed review on 26 September, won registration approval on 30 October, and listed on 5 December. The acceptance-to-approval window was 88 days.
For context, the STAR Market was built to fast-track "hard tech" — semiconductors, aerospace, advanced equipment. A GPU designer fits the mandate exactly, and the regulator's pace reflected that. The offer itself was 70 million shares, all new, representing 14.89% of post-issue capital. At 114.28 yuan a share, the issue valued the company at about 53.7 billion yuan (≈ US$7.6 billion / HK$58.5 billion) on debut day.
A small footnote worth keeping: 29,302 shares, worth about 3.35 million yuan, were abandoned by subscribers and covered by the underwriter. Minor in scale, but a quiet signal that not every investor wanted the year's priciest A-share ticket.
The growth, and the bill
The revenue curve is the optimistic half of the story. From 46 million yuan in 2022 to 702 million yuan in a single half-year in 2025, the trajectory is steep. The company attributes the jump to AI compute demand and broadening product coverage.
The losses are the other half. Across 2022–2024 the company burned 1.84 billion, 1.67 billion and 1.49 billion yuan (a combined ≈ US$705 million / HK$5.4 billion), then another 271 million yuan in the first half of 2025. The prospectus flags eight categories of risk, including unrecovered losses, triggers that could invite delisting warnings, and large inventory that carries write-down risk. The reason is plain: heavy R&D with scale effects not yet in view.
None of this is unusual for a fabless chip designer in a capital-hungry segment. What is unusual is the candour of the disclosure and the speed at which public markets opened the door.
Why a GPU maker matters now
China's compute self-reliance push has produced several AI-accelerator names, but fewer full-function GPU contenders. Moore Threads' listing puts a domestic general-purpose GPU in front of public investors and forces a question the industry has avoided: can a Chinese designer match not just the maths but the software ecosystem that makes a GPU useful?
That ecosystem question, more than any single benchmark, decides whether "first domestic GPU stock" becomes a durable category or a one-off label. The hardware is the easy part to announce; the drivers, libraries and developer trust are the years-long climb.
Honest limitations
This article is based on the company's prospectus and Chinese state-media reporting (Xinhua, CNR) around the December 2025 listing. We did not model post-listing share performance, and a listing is not commercial validation. No market-share, benchmark, or customer-concentration figures were disclosed in the sources we used. The "first domestic GPU stock" phrasing is a market label the company and outlets apply, not an audited ranking. Loss figures are historical; the path to profitability is the company's own projection and remains unproven. We have not verified any trading-price movement after debut and treat it as irrelevant to the business assessment.
What readers can do now
- If you follow China's semiconductor self-reliance push, track whether Moore Threads converts its prospectus pipeline — full-function GPUs for AI and graphics — into design wins beyond trials and into named customers.
- If you invest in A-share "hard tech," read the prospectus risk section (unrecovered losses, delisting triggers) before treating the "first GPU stock" label as a quality signal rather than a positioning one.
- If you build with Chinese compute, watch whether domestic full-function GPUs close the software gap (a CUDA-class ecosystem), because that matters more than the hardware spec sheet for real adoption.
