NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

Moore Threads Lists as China's 'First Domestic GPU Stock,' Raising RMB 8 Billion at RMB 114.28

Moore Threads (摩尔线程) debuted on Shanghai's STAR Market, raising about RMB 8 billion at RMB 114.28 a share while still posting billion-scale losses.

2026-11-14 · 1171 words · NeuroAI
Moore Threads Lists as China's 'First Domestic GPU Stock,' Raising RMB 8 Billion at RMB 114.28

A chip company that loses money by the billion still convinced public investors to hand it eight billion yuan.

Moore Threads makes the kind of processor that powers AI workloads and on-screen graphics.

The market's real question is whether a homegrown GPU can ever catch the incumbents it is meant to replace.

Key takeaways

  • Moore Threads (摩尔线程) listed on Shanghai's STAR Market (科创板) on 5 December 2025 under code 688795.SH, branded by local media as China's "first domestic GPU stock" (国产GPU第一股).
  • It sold 70,000,000 shares at RMB 114.28 each — raising about RMB 8.0 billion (≈US$1.1B) gross, with net proceeds near RMB 7.58 billion (≈US$1.05B).
  • The deal was fast-tracked: from SSE acceptance on 30 June 2025 to IPO approval on 26 September was just 88 days, and it became the largest STAR Market IPO of 2025 by proceeds.
  • The financials are heavy: 2022–2024 revenue climbed from RMB 46 million to RMB 438 million, but cumulative net losses over those three years reached about RMB 5.0 billion; the first half of 2025 alone lost RMB 271 million.
  • The cash is earmarked for next-generation domestic AI training/inference chips, graphics chips and AI SoCs, yet the company remains unprofitable with a large accumulated deficit.

What Moore Threads actually makes

A GPU (graphics processing unit, 图形处理器) is the workhorse behind both video rendering and the parallel maths that modern AI runs on. Moore Threads, founded in 2020 and based in Beijing, builds what it calls "full-function" GPUs — chips that handle AI computing, graphics rendering, scientific computing and media in one architecture, rather than specialising in just one. By its own account the company has shipped four generations of GPU architecture.

That breadth is the pitch. Nvidia dominates both the AI-training and graphics markets globally, and Chinese buyers have been pushed to find domestic alternatives as advanced exports tighten. A "full-function" local GPU is therefore as much a supply-security story as a commercial one, which is exactly why a company of this profile attracts state-linked and strategic capital.

The numbers behind the eight billion

The offering was large by any measure. Seventy million new shares at RMB 114.28 priced the deal at gross proceeds of RMB 7,999,600,000 — round it to RMB 8.0 billion (≈US$1.1B) — with net proceeds after fees near RMB 7.58 billion (≈US$1.05B). The RMB 114.28 issue price was the biggest STAR Market debut of the year by money raised, behind only a much larger main-board utility listing.

The speed was the other headline. Moore Threads filed with the Shanghai Stock Exchange on 30 June 2025, cleared its hearing on 26 September, and listed on 5 December — about 88 days from acceptance to approval. For a loss-making hardware firm, that is an unusually compressed timeline and reflects how deliberately the listing framework has been tuned to channel capital into hard-tech.

Shares more than quadrupled on the first trading day, closing at RMB 600.50 against the RMB 114.28 issue price, before giving back a large share of that gain in subsequent sessions — a reminder that debut pops on hot tech listings say more about scarcity and sentiment than about underlying earnings.

Why a loss-making GPU gets a listing

The prospectus tells the cautious half of the story. Revenue grew fast on a small base: RMB 46 million in 2022, then RMB 438 million in 2024, and RMB 702 million in the first half of 2025 alone — already above any full prior year. But the company has never turned a profit. Net losses were RMB 1.84 billion, RMB 1.67 billion and RMB 1.49 billion across 2022–2024, a three-year total near RMB 5.0 billion, plus a RMB 271 million loss in the first half of 2025. By mid-2025 its accumulated uncovered loss stood at about RMB 1.6 billion. These are two measures: the ~RMB 5.0 billion is the cumulative consolidated net loss attributable to the parent (2022–H1 2025), while the ~RMB 1.6 billion — about RMB 1.48 billion per the prospectus (累计未弥补亏损) and CNR — is the carried-forward uncovered deficit as of 30 June 2025.

The risks the company itself lists are candid: it is unprofitable with an uncovered deficit, it carries large inventory, and it faces the usual pre-profit hard-tech exposures. The growth is real, but it is growth from a low base bought with heavy R&D spend that has not yet reached scale.

According to Li An, chief scientist at BrainNet (脑机网), a domestic GPU listing like this is less a verdict on today's product and more a capital vote that China intends to fund the substitute for restricted foreign chips through public markets. Li An's comments reflect his own views and do not represent NeuroAI's editorial position or an endorsement by BrainNet (脑机网).

What the cash is for — and the risks

The proceeds plan is tightly focused on silicon. The company said it intended to raise RMB 8.0 billion for three research tracks — a next-generation autonomous, controllable AI training-inference integrated chip, a next-generation autonomous graphics chip, and a next-generation autonomous AI SoC — plus working capital. In plain terms: keep building the domestic GPU roadmap and try to reach the volume where engineering cost per chip finally falls.

The tension is the same one every pre-profit chipmaker lives with. Public markets supplied the war chest precisely because the strategic prize is large, but the path from a working domestic GPU to meaningful share against an entrenched incumbent runs through years of R&D, ecosystem software and customer qualification — none of which a listing cheque shortens by itself.

Honest limitations

The issue price, share count, gross and net proceeds, and "largest STAR Market IPO of 2025 by proceeds" claim are taken from the company's issuance result announcement and corroborated by China National Radio, China Economic Net and SSE-linked market data; the 88-day acceptance-to-approval timeline comes from Xinhua and CNR reporting. Revenue, loss and accumulated-deficit figures are from the prospectus via CNR and CE.cn, and could be restated in later filings. The first-day closing price (RMB 600.50) is a single-session market datum, not a valuation judgement, and the subsequent pullback is not tracked here. "First domestic GPU stock" (国产GPU第一股) is a media label, not an official designation, and "full-function GPU" is the company's own description. I have not verified post-listing financials, lock-up expiries or the realised progress of the three chip programmes. Nothing here is a recommendation to buy.

What readers can do now

  • If you follow China's chip listings, separate the strategic case from the income statement: Moore Threads raised on a supply-security thesis, not on profit, and the losses are structural, not temporary.
  • If you evaluate pre-profit hard-tech IPOs, watch whether revenue can scale past the RMB 438 million-ish annual base fast enough to bend the loss curve before the accumulated deficit grows further.
  • If you track the domestic GPU race, the numbers that matter after listing are progress on the three next-generation chip tracks, gross-margin trajectory, and whether design wins convert into volume shipments rather than pilot orders.

Related coverage

Cite this article

NeuroAI (脑机网), "Moore Threads Lists as China's 'First Domestic GPU Stock,' Raising RMB 8 Billion at RMB 114.28," 14 November 2026, https://neuroai.site/a/na-capital-moore-threads-ipo

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Capital & Markets” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories