Chinese Nvidia rival Enflame soars 206% on stock market debut as AI demand stays hot

Chinese AI-chip maker Enflame surged 206% in its Shanghai trading debut after its retail IPO drew orders exceeding available shares by more than 6,000 times. The Tencent-backed company reported 2025 revenue of 990 million yuan ($147 million), up 37% from 722 million yuan, although it remains unprofitable and will use listing proceeds to develop fifth- and sixth-generation AI processors. The rally reflects investor confidence that domestic suppliers can displace Nvidia as U.S. export controls and China’s self-sufficiency push accelerate semiconductor investment, which Goldman Sachs forecasts will reach $82 billion by 2030.

Analysis

The listing frenzy is more important as a signal of China’s capital-allocation regime than as validation of any individual accelerator’s economics. Public-market funding at elevated valuations lowers the cost of capital for domestic challengers, allowing them to subsidize software porting, customer pilots and successive tape-outs despite negative operating cash flow. That raises the probability of a multi-year margin and share-pressure cycle for NVDA’s China-adjacent revenue pool, but does not imply near-term displacement in workloads requiring mature CUDA tooling, high interconnect performance, or reliable supply.

The second-order beneficiary is the domestic stack rather than any single chip designer: BABA can monetize demand through cloud capacity, model hosting and software optimization even if it does not win accelerator share directly. Huawei, local foundry capacity, HBM/DRAM suppliers, advanced packaging and server integrators are likely to capture more durable value than highly volatile pre-profit chip IPOs. The key bottleneck is usable cluster-level performance—yield, memory bandwidth, networking, compiler maturity and power efficiency—not advertised single-chip specifications.

Near term, extreme retail demand creates a scarcity-and-momentum setup that can detach listed Chinese AI hardware from fundamentals for weeks. Over 1-3 months, quarterly order disclosures, gross-margin trends and evidence of repeat deployment by major cloud customers will determine whether the theme broadens. Over 6-18 months, domestic substitution becomes material only if Chinese chips achieve acceptable total cost of ownership at scale; a relaxation of U.S. export rules, or evidence that domestic systems require materially more chips/power per unit of model output, would weaken the thesis.

Consensus appears too quick to translate China accelerator enthusiasm into a large immediate NVDA earnings headwind. China replacement is likely additive semiconductor capex initially because customers will dual-source and overbuild to mitigate supply and policy risk. The more actionable bearish NVDA catalyst is not local IPO performance, but a disclosed reduction in China-related data-center demand or a broader pricing response from domestic alternatives.

Key Decisions for Investors

• Maintain NVDA as a watch-short rather than initiating on this news alone. Trigger a 1-3 month tactical short only if management guides to a renewed China-related data-center revenue decline or gross-margin pressure; use a defined-risk put spread because non-China hyperscaler demand can overwhelm the China narrative.
• Prefer a 6-12 month long BABA position over direct participation in newly listed Chinese AI-chip names. The cloud/model/software layer has broader exposure to domestic compute substitution and less single-product execution risk; exit or reduce if Alibaba Cloud growth fails to accelerate over two reporting periods or AI investment materially compresses consolidated margins.
• For China-tech exposure, use a barbell: long BABA or KWEB against a modest short SMH or NVDA put-spread hedge. This expresses domestic-stack funding and deployment growth while limiting exposure to a global AI-hardware valuation correction; reassess if U.S.-China export policy softens.
• Do not chase Chinese AI-chip IPO momentum without independently verified backlog, customer concentration, gross margin and cash-burn data. Treat a post-lockup supply event, weak first earnings report, or a failure to disclose repeat cloud deployments as a potential short catalyst rather than a reason to add.