Enflame Technology Passes STAR Market IPO Review: All Four Chinese GPU Dragons Now Listed on A/H Shares
N.R. Finch
Enflame Technology (燧原科技) cleared its STAR Market IPO review, targeting ¥6 billion in proceeds. With this approval, all four of China's so-called 'GPU dragons' have now secured A- or H-share listings — shifting the domestic GPU race from lab to capital market.
Who are the "four GPU dragons," and why does a full listing matter?
Enflame, Moore Threads (摩尔线程), Metax (沐曦), and Biren (壁仞) — all founded in 2019–2020 as Nvidia's high-end chips faced tighter export controls on China.
Moore Threads and Metax already trade on the A-share STAR Market. Biren chose Hong Kong. Enflame, now cleared, will be the last of the four to list.
This means → the rules of competition have changed. The race shifts from "who can raise venture capital" to "who can prove themselves on a public market" — with disclosure requirements, earnings scrutiny, and investor accountability all kicking in at once.
Do they all make the same thing?
Not quite. Enflame and Biren focus on cloud-based AI training and inference chips, competing head-on with Nvidia's data-center lineup.
Moore Threads straddles graphics rendering and AI compute — one of few Chinese players pursuing a full-function GPU roadmap. Metax seeks a middle ground between general-purpose computing and AI acceleration.
In plain terms = the four are not running the same race. Each is staking out a different niche — whoever secures their segment first survives to the next round.
What other signal dropped the same day?
On the same day, CanSemi (粤芯半导体) — dubbed "Guangzhou's first fab" — also cleared its ChiNext IPO review, potentially becoming ChiNext's first wafer-manufacturing listing.
This reflects a broader pattern: China's semiconductor capitalization is not a single-point push but a simultaneous advance across design, fabrication, and packaging.
In plain terms = a GPU design house and a chip foundry clearing IPO review on the same day signals that the capital market is issuing a pass to the entire supply chain, not just one link.
Is ¥6 billion enough?
Enflame plans to raise ¥6 billion (~$830 million), earmarked for cloud AI chip R&D iterations and capacity buildout. For context, Nvidia spends on the order of tens of billions of dollars on R&D per year.
The deeper moat is not just money: Nvidia's CUDA ecosystem — a software platform that lets developers write programs for its GPUs — creates enormous switching costs. Changing chips means rewriting large amounts of code.
This means → ¥6 billion is an entry ticket, not a winning hand. Capital efficiency — whether the gap to leading international architectures can be narrowed — is the real test.
What to watch next?
R&D iteration speed: whether the architecture and process-node gap keeps closing.
Software ecosystem: whether customer migration costs away from CUDA can be lowered — the critical bottleneck to commercialization.
Customer validation: whether orders shift from policy-driven "must-buy" to performance-driven "want-to-buy."
Content is for reference only, not financial advice.