Moore Threads Eyes Hong Kong: China's 'Little Nvidia' Posts 147% Revenue Surge and Plans a Second Listing

Moore Threads Eyes Hong Kong: China's 'Little Nvidia' Posts 147% Revenue Surge and Plans a Second Listing

On August 9, 2026, Moore Threads — the Beijing-based GPU startup that staged one of China's most spectacular stock market debuts in recent memory — announced that its board has approved plans to issue H shares on the Hong Kong Stock Exchange (HKEX). The announcement landed the same week the company disclosed first-half 2026 revenue of 1.74 billion yuan (approximately $257 million), a 147% year-on-year increase that has transformed the narrative around China's domestic AI chip sector from one of desperate improvisation to one of genuine commercial momentum.

The dual-listing strategy is not merely a capital-raising exercise. It is a statement: China's homegrown GPU ecosystem, long dismissed as a poor substitute for Nvidia's hardware, is now actively courting the international investors who once assumed the story would never get this far.

From Nvidia Exile to STAR Market Sensation

Moore Threads was founded in October 2020 by Zhang Jianzhong, the former Global Vice President and China General Manager of Nvidia, alongside a team of engineers drawn from Nvidia, Microsoft, and Intel. The founding premise was straightforward and audacious in equal measure: build a full-stack GPU company — hardware, software, and developer ecosystem — capable of serving China's AI infrastructure needs without relying on American silicon.

The company's December 5, 2025 debut on the Shanghai STAR Market↗ was a watershed moment. Shares opened at 650 yuan, a 468% premium over the issue price of 114.28 yuan, and closed the day up 425% at 600.50 yuan. The IPO raised approximately 7.576 billion yuan ($1.07 billion) — the largest STAR Market listing of 2025 — and briefly pushed the company's market capitalization above 300 billion yuan. The subscription multiple of 2,751 reflected the intensity of domestic investor demand for a credible Nvidia alternative.

Since that debut, shares have risen a further 420%, compounding the initial surge into a total return that has made Moore Threads one of the most closely watched names in Chinese technology investing.

"The IPO was not just a financing event — it was a signal that China's semiconductor self-sufficiency story had found a commercial anchor. Moore Threads gave investors something concrete to own." > — Analysis from The Next Web's coverage↗ of the Hong Kong listing announcement

The Financial Case for a Hong Kong Listing

The H1 2026 numbers that accompanied the Hong Kong listing announcement are the strongest argument for why international investors should pay attention.

Key Financial Metrics — H1 2026

• Revenue: 1.736 billion yuan — up 147.42% year-on-year, driven by mass production of the flagship MTT S5000 AI computing card↗ and accelerating adoption of the KUAE Intelligent Computing Cluster
• Gross profit: 989 million yuan — up 103.78% year-on-year, reflecting improving unit economics as production scales
• Net loss: 11.6 million yuan — down 95.73% from 270.9 million yuan in H1 2025, approaching breakeven
• Q1 2026 net profit: 29.4 million yuan — the company's first quarterly profit↗, reversing the loss recorded in Q1 2025
• Cumulative R&D investment since 2022: ~5.9 billion yuan, with 769 million yuan allocated in H1 2026 alone

The trajectory is unambiguous. Moore Threads is not a speculative bet on future chip production; it is a company that has already crossed into commercial-scale revenue generation, with a clear path to sustained profitability. A single 660 million yuan order disclosed in March 2026 illustrated the scale of enterprise demand the company is now capturing.

Why Hong Kong, and Why Now

The strategic rationale for a secondary HKEX listing is multi-layered. The Shanghai STAR Market, while providing domestic capital access, is effectively closed to most international institutional investors due to regulatory and operational barriers. A Hong Kong H-share listing opens the register to global funds, sovereign wealth vehicles, and the international technology investors who have been watching China's AI chip story from the sidelines.

According to SCMP's reporting on the announcement↗, Moore Threads cited four specific objectives for the listing:

• Deepening its international strategic footprint — signaling compliance with global financial transparency standards to enterprise clients outside China
• Attracting and retaining global engineering talent — a Hong Kong listing provides equity compensation structures that are more legible to international hires
• Accessing international capital markets — enabling foreign investors to participate in the company's growth story
• Improving corporate governance — the disclosure requirements of HKEX listing rules impose a discipline that strengthens institutional credibility

The timing is also deliberate. With Nvidia's China market share having effectively collapsed under U.S. export controls — Tom's Hardware reported↗ that Nvidia's Jensen Huang confirmed his company's China AI market share has fallen to zero — the window for a domestic GPU champion to establish itself as the default infrastructure layer for Chinese AI is open and narrowing. Moore Threads is moving to lock in that position before the competitive landscape consolidates further.

The Hardware: MTT S5000 and the Huashan Roadmap

The commercial engine behind Moore Threads' revenue surge is the MTT S5000, the company's flagship AI accelerator built on its fourth-generation "Pinghu" architecture. The card's specifications position it as a credible, if not yet equal, alternative to Nvidia's Hopper-class hardware for Chinese AI workloads:

MTT S5000 Key Specifications

• Peak AI compute: 1,000 TFLOPS (dense, single card)
• Memory: 80GB with 1.6 TB/s bandwidth
• Inter-card connection bandwidth: 784 GB/s
• Precision support: FP8 through FP64, including native hardware-level FP8 — the first Chinese GPU to offer this
• Framework compatibility: PyTorch, Megatron-LM, vLLM, SGLang via the MUSA software stack
• Benchmark performance on DeepSeek V3: 1,000 tokens/second decode, 4,000 tokens/second prefill — competitive with Hopper-class hardware on inference workloads

The MTT S5000 is being deployed in major Chinese AI computing hubs including Beijing, Wuxi, and Hangzhou, and forms the core of the KUAE Intelligent Computing Cluster that has driven the company's enterprise order book.

Looking further ahead, Moore Threads has unveiled the Huashan AI accelerator line, built on the new "Huagang" architecture announced in December 2025. The Huashan design uses a dual-chiplet configuration with 9 HBM modules, supports the MTLink 4.0 interconnect enabling clusters of over 100,000 units with 1,314 GB/s bandwidth, and targets compute performance comparable to Nvidia's Blackwell generation. Independent verification of these claims remains pending, but the architectural ambition is clear: Moore Threads is not designing for today's workloads, it is designing for the training runs of 2027 and beyond.

The Software Moat: MUSA and the Developer Ecosystem

Hardware alone does not build a platform. The lesson of Nvidia's dominance is that CUDA — the software ecosystem that makes Nvidia GPUs programmable for AI — is as much a competitive moat as the silicon itself. Moore Threads has invested heavily in its answer: the MUSA (Meta-computing Unified System Architecture) software stack.

As of mid-2026, MUSA supports over 800,000 registered developers, a milestone that reflects both the scale of China's domestic AI developer community and the company's deliberate investment in ecosystem building. The stack includes:

• MCC (Moore Compiler) and the MUSA Runtime for native GPU kernel development
• MUSA-X, a specialized library collection for AI workloads
• Musify, a toolkit that translates existing CUDA code into MUSA-compatible equivalents, lowering the migration barrier for teams with established Nvidia codebases
• MusaCoder, a full-stack training framework↗ developed on Moore Threads GPU clusters, featuring a progressive data synthesis pipeline and reinforcement learning guided by MooreEval, a distributed execution sandbox

The Musify toolkit is particularly significant. The single largest barrier to switching from Nvidia hardware is not the hardware itself — it is the accumulated CUDA code that runs on it. By providing a translation layer, Moore Threads is reducing the switching cost for the Chinese AI labs, cloud providers, and enterprises that have built their infrastructure on Nvidia's ecosystem but can no longer reliably access it.

"The MUSA ecosystem is the real product. The GPU is the delivery mechanism. If Moore Threads can make MUSA sticky enough that developers write to it natively rather than porting from CUDA, the competitive dynamics shift permanently." > — Fanatical Futurist analysis↗ of Moore Threads' 2026 GPU roadmap

Competitive Landscape: A Crowded Race for Nvidia's Vacated Ground

Moore Threads does not operate in isolation. China's domestic GPU sector has attracted a cohort of well-funded competitors, each targeting different segments of the market that Nvidia's export restrictions have opened up.

Huawei's Ascend series — particularly the Ascend 950DT, which debuted on Huawei's cloud platform in August 2026 — targets the highest-performance training workloads and has the advantage of Huawei's existing enterprise relationships and domestic chip manufacturing scale. Cambricon, which reported a 108% revenue surge in H1 2026, focuses on inference accelerators and has established a strong position in edge and cloud inference deployments. MetaX, Biren Technology, and Enflame round out the competitive field, each with distinct architectural approaches and customer bases.

Moore Threads' differentiation lies in three areas: the breadth of its software ecosystem (MUSA's 800,000+ developers dwarfs most competitors), the scale of its institutional backing (Tencent, ByteDance, Sequoia China, GGV Capital, and government-backed funds), and its status as the only domestic GPU company to have completed a major public listing — giving it a balance sheet and a public market currency that private competitors cannot match.

The U.S. Entity List designation, which Moore Threads received in October 2023, remains a constraint. The company cannot access TSMC's advanced nodes and must rely on domestic foundries including SMIC, which face higher costs and lower production yields than their Taiwanese counterparts. This manufacturing gap is real and will not close quickly. But it has not prevented the company from achieving commercial scale — a fact that would have seemed implausible two years ago.

What the Hong Kong Listing Means for the Broader Ecosystem

The Moore Threads HKEX announcement is significant beyond the company itself. It represents the first major attempt by a Chinese AI chip company to access international public capital markets at scale, and its reception will be closely watched by the broader domestic semiconductor sector.

A successful Hong Kong listing would accomplish several things simultaneously. It would provide Moore Threads with a currency — H shares — that can be used for international acquisitions, talent compensation, and strategic partnerships in ways that STAR Market A shares cannot. It would give international investors a liquid, regulated vehicle for exposure to China's domestic AI infrastructure buildout. And it would signal to the global investment community that the Chinese AI chip story has matured beyond the "national champion" narrative into something that can be evaluated on conventional financial metrics.

The risks are real. The HKEX listing process is rigorous, and the geopolitical environment — with U.S.-China technology tensions showing no sign of abating — creates uncertainty about how international investors will price a company that sits squarely in the crosshairs of export control policy. Moore Threads' Entity List status means that any international partnerships or technology acquisitions will face scrutiny that domestic competitors without that designation do not.

But the financial trajectory is hard to argue with. A company that was losing 270 million yuan per half-year in 2025 and is now approaching breakeven, with revenue growing at 147%, has earned the right to make its case to a global audience.

Practical Takeaways for Developers and Enterprises

For the AI practitioners and enterprise technology teams who follow this desk, the Moore Threads story has immediate practical implications:

• The MTT S5000 is a deployable alternative for Chinese AI workloads today. Its DeepSeek V3 benchmark performance — 1,000 tokens/second decode — is competitive with Hopper-class hardware for inference, and the MUSA stack supports the major inference frameworks (vLLM, SGLang) that production deployments require.
• The Musify CUDA translation toolkit lowers migration costs. Teams with existing CUDA codebases can evaluate Moore Threads hardware without a full rewrite, reducing the friction of a hardware transition.
• The KUAE Intelligent Computing Cluster is the enterprise product. Individual GPU cards are not how Moore Threads sells to large customers; the cluster product, which has already attracted orders in the hundreds of millions of yuan, is the relevant unit for enterprise procurement decisions.
• The Hong Kong listing, if completed, will improve price discovery. International investors will be able to benchmark Moore Threads against global semiconductor peers, which may provide a more accurate valuation signal than the domestic STAR Market, where sentiment around AI hardware has been exceptionally elevated.

The Huashan next-generation architecture, with its Blackwell-class performance targets and 100,000-unit cluster scalability, is the product that will determine whether Moore Threads can compete at the frontier of AI training — not just inference. That story will unfold over the next 12 to 18 months. For now, the H1 2026 numbers and the Hong Kong listing announcement confirm that the foundation is real, the revenue is real, and the ambition is being backed by capital that is increasingly willing to bet on it.