Last Updated -

June 11, 2026

Enflame Technology

Company Profile and Market Insights

Explore the business model, global strategy, and market performance including insights into its position in China.

Enflame Technology
Key facts
Founded 2018 • Shanghai, China • Private AI accelerator company in STAR Market IPO review (RMB 6.0bn raise)
RMB 540.2m
2025 1-9M revenue
RMB 887.8m
2025 1-9M net loss
36.23%
2025 1-9M main-business gross margin
30,236
2025 1-9M accelerator cards/modules shipped
RMB 13,616.82
2025 1-9M avg selling price per card
76.51%
R&D staff share of employees (2025-09-30)

About

Enflame Technology, formally Shanghai Enflame Technology Co., Ltd., is a Chinese AI accelerator company founded in 2018 and headquartered in Shanghai. The company designs cloud AI chips, AI accelerator cards and modules, intelligent computing systems and clusters, and the TopsRider AI computing and programming software platform. It uses a fabless semiconductor model, meaning it focuses on chip design and product development while outsourcing wafer fabrication, packaging, and testing to manufacturing partners.

Enflame has developed from a start-up into one of China’s more visible domestic AI chip vendors, with four generations of self-developed architecture and five chips disclosed in its IPO materials. Its products target cloud training and inference workloads, where AI models need large amounts of parallel computing power. The company’s strategic purpose is to build a domestic AI computing stack across chips, hardware, software, and cluster solutions, supported by planned IPO funding for fifth- and sixth-generation AI chip development and software-hardware co-innovation.

Enflame remains private, but its Shanghai Stock Exchange STAR Market IPO application was accepted in January 2026 and moved to inquiry status in February 2026, with a planned RMB 6.0 billion raise. The latest detailed figures cover the nine months ended September 30, 2025, when revenue was RMB 540.189 million and main-business gross margin reached 36.23%, up from 30.59% in 2024. AI accelerator cards and modules accounted for RMB 411.718 million of main-business revenue in that period, with 30,236 cards shipped, while the company still posted a net loss of RMB 887.756 million and had 860 employees, including 658 in research and development.

Enflame Technology

Business Model and Market Position

Enflame Technology is a private Shanghai-based AI accelerator company built around a fabless semiconductor model. It designs cloud AI chips, accelerator cards and modules, intelligent computing systems, AI clusters, and the TopsRider AI computing and programming software platform, while outsourcing wafer fabrication, packaging, and testing to manufacturing partners.

The company is still pre-IPO. Its STAR Market listing application was accepted in January 2026 and moved to inquiry status in February 2026, with a planned RMB 6.0 billion raise. No Q1 2026 financial results are available because Enflame has not yet listed and does not publish public quarterly earnings. The latest detailed figures cover the nine months ended September 30, 2025.

Revenue comes mainly from selling AI accelerator hardware and related computing systems to Chinese customers. In the first nine months of 2025, Enflame generated RMB 540.2 million of revenue. Main-business revenue was RMB 536.6 million, led by RMB 411.7 million from AI accelerator cards and modules, RMB 123.6 million from intelligent computing systems and clusters, and RMB 1.3 million from IP licensing and other income. AI accelerator card and module shipments reached 30,236 units in the period, above 21,789 units for full-year 2024, with an average selling price of RMB 13,616.82 per card.

The business has three core operating areas

  1. AI chips and hardware: Enflame develops self-designed cloud AI processors, accelerator cards, and modules for training and inference workloads.
  2. Intelligent computing systems and clusters: The company supplies customized systems and cluster solutions for internet platforms, large-model customers, and intelligent-computing-center projects.
  3. Software and programming platform: TopsRider supports Enflame’s hardware ecosystem, although the company had not sold it as a standalone product during the reporting period.

Enflame sells mostly through direct sales. More than 97% of main-business revenue during the reporting period came from direct customer relationships, with only a small distributor contribution in 2024 and the first nine months of 2025. The customer base is concentrated in mainland China, which contributed 99.98% of main-business revenue in the first nine months of 2025 and 100% from 2022 through 2024. International sales are immaterial.

Its competitive position rests on domestic AI compute substitution, product iteration, and integration across chip, card, system, cluster, and software layers. Since 2018, Enflame has developed four generations of self-developed architecture and five chips. Its fourth-generation cloud AI chip is described in its IPO materials as one of the few domestic products with native FP8 low-precision support and is designed to support super-node and 10,000-card-plus cluster solutions.

Enflame does not follow Nvidia’s GPGPU and CUDA architecture. It uses its own instruction-set architecture, GCU-CARE AI compute units, and GCU-LARE high-speed interconnect technology, while benchmarking functions similar to Tensor Core and NVLink. This gives the company a differentiated domestic stack, but it also raises ecosystem risk because customers need software frameworks, tools, and workloads to run efficiently outside Nvidia’s CUDA environment.

The company remains small relative to the overall Chinese AI accelerator market. Media summaries of the prospectus state that its 2024 AI accelerator card and module volume represented about 1.4% of China’s AI accelerator-card market, while still placing it among leading domestic AI chip vendors. Its market position is therefore best understood as an emerging Chinese AI infrastructure supplier rather than an established global-scale semiconductor platform.

Direct competitors include Nvidia at the global level and Huawei Ascend in China’s domestic AI compute market. Domestic AI-chip peers and references in Enflame’s IPO materials include Cambricon, Moore Threads, MetaX, Biren Technology, and Iluvatar CoreX. Compared with Nvidia, Enflame has a much narrower customer base, lower scale, and a less mature software ecosystem. Compared with domestic peers, its advantages include meaningful accelerator-card shipment growth, deep R&D staffing, Tencent-linked commercial deployment, and disclosed technical reserves for high-card-count clusters.

Financially, Enflame is still in an investment-heavy stage. Main-business gross margin improved to 36.23% in the first nine months of 2025, up from 30.59% in 2024 and 22.60% in 2023, helped by product iteration and a higher contribution from accelerator cards and modules. The company remained loss-making, with a net loss of RMB 887.8 million in the first nine months of 2025 and accumulated consolidated unrecovered losses of RMB 4.17 billion at September 30, 2025. R&D employees totaled 658 out of 860 staff, or 76.51% of headcount, showing that the business model is still heavily weighted toward technology development and future product generations.

Enflame Technology

Performance in China

China is Enflame Technology’s core market and effectively its entire commercial base. In the nine months ended September 2025, 99.98% of main-business revenue came from mainland China, with the balance from Hong Kong. Revenue was RMB 540.2 million, led by RMB 411.7 million from AI accelerator cards and modules and RMB 123.6 million from intelligent computing systems and clusters. The company shipped 30,236 accelerator cards and modules during the period, already above its 2024 full-year volume of 21,789. Its strategy is to supply domestic AI compute alternatives for Chinese internet platforms, large-model developers and intelligent-computing-center projects. Tencent is a key investor and customer, with Tencent Technology (Shenzhen) reported to account for 57.28% of 2025 nine-month revenue. Enflame competes with Huawei Ascend, Cambricon, Moore Threads, MetaX, Biren Technology and Iluvatar CoreX. Its STAR Market IPO entered inquiry in February 2026, with RMB 6.0 billion planned for fifth- and sixth-generation AI chips and software-hardware co-innovation.

Growth and Future Prospects

Enflame Technology’s growth profile is tied to China’s demand for domestic AI compute infrastructure. The company remains private, so there are no quarterly public earnings releases. The latest detailed figures cover the nine months ended September 30, 2025, when revenue reached RMB 540.189 million, compared with RMB 722.387 million for full-year 2024 and RMB 301.187 million in 2023. The main turning point is that accelerator card and module revenue in 2025 1-9M, at RMB 411.718 million, already exceeded the full-year 2024 level, while main-business gross margin improved to 36.23% from 30.59% in 2024 and 22.60% in 2023. Losses remain large, with a net loss of RMB 887.756 million for 2025 1-9M and accumulated consolidated unrecovered losses of RMB 4.165 billion.

Key growth drivers

  1. Domestic AI infrastructure demand: U.S.-China technology restrictions and China’s domestic substitution policy support demand for China-made AI accelerators, software stacks, and cluster systems.
  2. Product iteration: Enflame has developed multiple chip generations since 2018. Its fourth-generation cloud AI chip supports native FP8 low-precision computing and high-card-count cluster configurations, which are relevant for large-model training and inference workloads.
  3. Higher-value mix: Accelerator cards and modules drove most 2025 1-9M main-business revenue, while intelligent computing systems and clusters added RMB 123.571 million. Product mix improvement has supported better gross margins.
  4. IPO-funded roadmap: The planned RMB 6.0 billion STAR Market raise is earmarked for fifth- and sixth-generation AI chip R&D and industrialization, plus advanced software-hardware co-innovation projects.
  5. Software and cluster ecosystem: TopsRider, Enflame’s AI computing and programming platform, is part of its attempt to build a broader hardware-software stack, although it was not sold as a standalone product during the reporting period.

Geographic expansion is limited at this stage. In 2025 1-9M, 99.98% of main-business revenue came from mainland China, with only a small Hong Kong contribution. The near-term expansion opportunity is therefore domestic, across internet platforms, large-model companies, and intelligent computing center projects, rather than international sales.

Challenges ahead

  1. Profitability: Enflame is still investing far ahead of earnings, and sustained R&D spending is necessary to keep pace with larger competitors.
  2. Customer concentration: Tencent-linked entities are strategically important as investors, customers, and suppliers, which creates dependency and bargaining-power risk.
  3. Ecosystem risk: Enflame uses a self-developed architecture rather than Nvidia’s CUDA model, so customer migration, software compatibility, and developer adoption are central execution risks.
  4. Supply chain exposure: Advanced-process manufacturing, packaging, memory, and export-control-sensitive equipment remain material constraints for Chinese AI chip vendors.

The future outlook is attractive but high risk. Enflame has clear demand tailwinds, improving gross margin, and a funded product roadmap if the IPO completes. Its ability to convert that position into durable value depends on scaling accelerator deployments, reducing losses, widening its customer base beyond a few large accounts, and proving that its software and cluster stack performs reliably in production AI workloads.

This Company Profile was written by Dominik Diemer

Dominik Diemer blends an investor mindset with execution discipline.

He is a SAFe Program Consultant (SPC) and Lean Portfolio Management (LPM) practitioner at DMG MORI Digital, working as a SAFe Release Train Engineer and internal consultant in the Lean-Agile Center of Excellence (LACE).

His focus is prioritization, flow, and dependency management that turns strategy into outcomes. With experience across Bertelsmann and the Founders Foundation, he bridges corporate and startup thinking.

He also invests privately in private equity deals, sharpening his view on business models, value drivers, and go-to-market.

StockCounterParts reflects that lens.