๐Bloomberg TechnologyโขStalecollected in 13m
Tencent-Backed Enflame Heads to IPO as China AI Chip Wave Grows
๐กTracking the IPO of a major Chinese AI chipmaker to assess the competitive landscape for domestic hardware alternatives.
โก 30-Second TL;DR
What Changed
Enflame is the final member of China's 'four little dragons' AI chipmakers to pursue an IPO.
Why It Matters
The IPO provides Enflame with significant capital to scale production and R&D, potentially accelerating the availability of domestic alternatives to Nvidia GPUs in the Chinese market.
What To Do Next
Monitor Enflame's technical documentation and benchmark releases to evaluate their chips as a viable alternative for local AI infrastructure deployment.
Who should care:Founders & Product Leaders
๐ง Deep Insight
Web-grounded analysis with 17 cited sources.
๐ Enhanced Key Takeaways
- โขEnflame Technology's initial public offering is on China's STAR Market, with the company aiming to raise approximately 6 billion yuan (about $830 million USD).
- โขTencent Holdings Ltd. is Enflame's largest institutional shareholder, holding approximately 20% of the shares, and also serves as a major customer, deploying Enflame chips within its own data centers.
- โขEnflame has developed through four architectural generations, launching five cloud AI chips, including the third-generation S60, which shipped approximately 70,000 units by mid-2025, and the fourth-generation L600, unveiled in July 2025, featuring 144GB on-chip memory and 3.6TB/s memory bandwidth.
- โขThe other members of China's 'four little dragons' in the AI chip sector are identified as Moore Threads, MetaX, and Biren Technology, all of whom have also pursued public listings.
- โขEnflame has reportedly submitted downgraded designs of its chips to TSMC in late 2023 to comply with U.S. export restrictions on advanced computing hardware to China.
๐ ๏ธ Technical Deep Dive
- Enflame DTU 1.0 (CloudBlazer T10/T11):
- Manufactured on a 12nm FinFET process.
- Features a PCIe Gen4 x16 interface and 200GB/s interconnects.
- Includes HBM2 memory onboard.
- Comprises four core clusters with 32 AI compute cores and 40 data transfer engines.
- Utilizes a Very Long Instruction Word (VLIW) architecture, supporting data types from FP32 down to INT8, including mixed precision calculations.
- Incorporates Tensor ALUs to accelerate matrix/vector operations and exploits sparsity to optimize execution.
- Offers a compute capability of 20 TFLOPS@FP32 and 80 TFLOPS@FP16/BF16/INT16/INT8.
- Employs self-developed GCU-CARE accelerated computing unit and GCU-LARE inter-chip interconnect technology.
- Supported by the TopsRider full-stack software platform.
- Enflame S60 (Third-generation):
- Manufactured by TSMC using the N6NTO-HPC (6nm) process.
- Enflame L600 (Fourth-generation):
- Unveiled in July 2025.
- Equipped with 144GB of on-chip memory.
- Provides 3.6TB/s of memory bandwidth.
- Supports FP8 data format.
- Designed for both large-model training and inference applications.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
China's domestic AI chip sector will see increased public listings and state-backed investment.
The successful IPOs of Enflame and other 'little dragons' on the STAR Market, combined with significant state-affiliated funding, underscore a national strategic imperative for self-reliance in AI hardware, encouraging more companies to seek public capital and government support.
Chinese AI chipmakers will continue to face technological gaps and manufacturing constraints compared to global leaders like Nvidia.
Despite rapid advancements, industry experts suggest Chinese alternatives lag several years behind established ecosystems like NVIDIA's CUDA in optimization and library completeness, and U.S. export controls compel companies like Enflame to downgrade chip designs for production.
Tencent's deep integration with Enflame suggests a growing trend of captive supplier models for key Chinese tech giants.
Tencent's dual role as Enflame's largest shareholder and primary customer, accounting for over 70% of its 2025 revenue, highlights a strategic partnership aimed at securing a reliable domestic AI hardware supply amidst geopolitical uncertainties.
โณ Timeline
2018-03
Enflame Technology founded in Shanghai by former AMD executives Zhao Lidong and Zhang Yalin.
2018-08
Tencent leads a 340 million RMB pre-Series A investment in Enflame.
2020-12
Enflame launches its first-generation AI training chip.
2023-09
Enflame closes a Series D funding round, raising approximately 2 billion yuan ($274 million), bringing total funding to over $740 million.
2025-07
Enflame unveils its fourth-generation AI chip, the L600, at the World AI Conference in Shanghai.
2026-01
Enflame's IPO application is accepted by Shanghai's STAR Market, aiming to raise approximately 6 billion yuan.
๐ Sources (17)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: Bloomberg Technology โ