Analysis of Iluvatar CoreX's Hong Kong Listing and Domestic GPU Industry Development

Based on the latest market data and industry trends, I have prepared this in-depth analysis report on the development of the domestic GPU industry for you.

---

On January 8, 2026, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. (Stock Code: 09903.HK) was officially listed on the Main Board of the Hong Kong Stock Exchange, making it the second domestic GPU company to debut in the Hong Kong stock market[1][2].

• Opening Performance
  : The opening price was HK$190.2, a sharp
  31.54%
  increase from the issue price of HK$144.6, reflecting a warm market response[1]
• Midday Pullback
  : As of midday, the stock closed at HK$160.7, with the gain narrowing to
  11.13%
  , and the total market capitalization was approximately HK$41.3 billion[2]
• Grey Market Performance
  : The gain reached 36%-37% in the grey market trading on January 7[3]
• Subscription Status
  : The Hong Kong public offering received approximately
  414.24 times
  oversubscription, while the international offering received
  10.68 times
  oversubscription[1][2]

This IPO raised a total of HK$3.677 billion, with 18 cornerstone investors introduced, subscribing to a total of HK$1.583 billion. The cornerstone investor lineup includes industry leaders such as ZTE (Hong Kong) and 4Paradigm, as well as top domestic and international financial institutions like UBS, Fullgoal Fund, and China Asset Management[1].

Based on the prospectuses and inquiry responses of multiple domestic GPU players, the market share of domestic GPU enterprises in the AI chip market is still relatively low:

Player	Market Share	Data Scope	Remarks
Muxi	~1%	2024 China AI Accelerator Market	Based on sales amount[4]
Moore Threads	Less than 1%	2024 Domestic Market Segments	AI intelligent computing products, graphics acceleration products, etc.[4]
Biren Technologies	0.16%	2024 China Intelligent Computing Chip Market	[5]
Iluvatar CoreX	0.3%	Domestic General-Purpose GPU Market	[5]

Player Type	Market Share	Remarks
NVIDIA (Full-Featured GPU)	54.4%	Ranked first in the domestic AI chip market[4]
HiSilicon (ASIC)	21.4%	[4]
AMD (GPGPU)	15.3%	[4]
Total of Other Domestic Players	~1%	Including Muxi, Moore Threads, Biren Technologies, etc.[4]

This data indicates that although domestic GPU players are highly anticipated by the capital market for the rise of domestic AI chips, they are still in the early stage of large-scale commercialization, with a significant gap in market share[4][5].

The competitive barrier established by NVIDIA through its CUDA ecosystem is one of the core challenges faced by domestic GPU players.

1. Nearly 20 Years of Continuous Investment
   : CUDA was officially launched in 2006, and NVIDIA has invested more than 20% of its revenue in ecosystem construction for 10 consecutive years[6][7]

2. Vertical Integration of Software and Hardware
   :
   • The architecture is updated every two years, with CUDA updated simultaneously
   • As soon as a new card is launched, the entire suite including compilers, drivers, libraries, and the debugging tool Nsight becomes immediately available
   • Vertical optimization of software and hardware delivers “out-of-the-box” performance[6][7]

3. Developer Ecosystem Barrier
   :
   • NVIDIA has established CUDA research centers in universities worldwide
   • Deeply participates in open-source community development
   • Millions of developers worldwide have become accustomed to the CUDA programming model[6][7]

4. Market Share Monopoly
   : According to third-party data, NVIDIA holds
   94%
   of the global GPU market share in the AI and data center sectors in Q2 2025[6]

: For customers, migrating from NVIDIA to domestic GPUs incurs costs far higher than purchasing new hardware, equivalent to rewriting the entire software stack[6][7].

Facing NVIDIA’s ecosystem barrier, domestic GPU players are exploring multiple breakthrough paths:

Heterogeneous mixed training refers to breaking brand barriers in a computing cluster by simultaneously using NVIDIA, Huawei, Cambricon, and other domestic GPUs to create a highly compatible and engineering-robust computing environment[6][7].

: Shifting from “replacing NVIDIA” to “coexisting and collaborating with NVIDIA” is the most realistic breakthrough path in the short to medium term. This strategy reduces customers’ migration risks and provides domestic GPUs with a market entry point.

Currently, multiple domestic GPU players adopt the approach of supporting a CUDA-compatible layer in their own chip architectures. Although this “compatibility” approach only scratches the surface of ecosystem breakthroughs, it can reduce developers’ learning costs and accelerate ecosystem migration[6][7].

: This approach cannot fundamentally shake CUDA’s dominant position in the ecosystem and may face legal and patent risks.

Players represented by Iluvatar CoreX adhere to the full-stack independent R&D route for software and hardware, building independent software stacks and AI computing power solutions[1].

:

• R&D and commercialization of general-purpose GPU chips and accelerator cards
• Construction of proprietary software stacks and AI computing power solutions
• Practical deployment covering key fields such as finance, healthcare, transportation, and manufacturing
• As of June 30, 2025, cumulative shipments have exceeded 52,000 units, and the number of served customers has jumped from 22 in 2022 to 290[1]

Different players have chosen different technical routes and market strategies:

Player	Technical Route	Market Strategy
Moore Threads	AI Training-Inference Integrated Chips + Graphics Chips + AI SoC	Diversified scenario implementation, cooperation with AI startups[7]
Muxi	Data Center and High-Performance Computing	Partnering with large B-end customers such as H3C, focusing on large-scale B-end business first[7]
Iluvatar CoreX	Full-Stack General-Purpose GPU Solutions	Coverage of multiple fields including finance, healthcare, transportation, and manufacturing[1]
Biren Technologies	Data Center-Grade GPUs	Targeted at large model training and inference scenarios

Player	2022 Revenue	2023 Revenue	2024 Revenue	Compound Annual Growth Rate (CAGR)
Iluvatar CoreX	RMB 189 million	-	RMB 540 million	68.8%[1]
Moore Threads	RMB 46 million	RMB 124 million	RMB 438 million	-
Muxi	RMB 426,400	RMB 53.02 million	RMB 743 million	-
Biren Technologies	RMB 499,000	RMB 62.03 million	RMB 337 million	-

All domestic GPU players are currently in a loss-making state:

• Moore Threads
  : Accumulated losses of RMB 5.005 billion from 2022 to 2024[4]
• Muxi
  : Accumulated losses of over RMB 3.2 billion from 2022 to 2024[4]
• Biren Technologies
  : Accumulated losses of RMB 4.75 billion from 2022 to 2024[4]
• Iluvatar CoreX
  : Accumulated losses of over RMB 2.2 billion in three years[5]

High R&D expenses are the main reason for the losses:

• Moore Threads’ R&D expenses accounted for
  309.88%
  of its revenue in 2024[4]
• Biren Technologies’ R&D expenses accounted for
  245.5%
  of its annual revenue in 2024[4]

Domestic GPU players face the issue of high customer concentration:

• In 2024, the revenue from the top five customers accounted for over
  70%
  of total revenue for Muxi, Biren Technologies, Moore Threads, and Iluvatar CoreX[5]
• Customer structure changes frequently; only one of Muxi’s top five customers in 2024 remained in its top five list in Q1 2025[5]

According to industry forecasts, the market share of domestic general-purpose GPUs (calculated by shipment volume) has increased from 8.3% in 2022 to 17.4% in 2024, and is expected to exceed

by 2029[1]. Behind this accelerated replacement is the continuous realization of the logic of “demand-driven - technological breakthrough - commercial verification”.

1. Breakthrough in Manufacturing Capability
   : High-end chip production capacity and yield control are key bottlenecks, requiring collaborative breakthroughs in the domestic foundry industry chain[6]

2. Software Ecosystem Construction
   : Referencing Google’s TPU experience, software ecosystem construction requires continuous investment from an engineering team of over 2,000 people[6][7]

3. Differentiated Competitive Advantage
   : Establish technological leadership and cost advantages in specific vertical scenarios

4. Policy Support
   : The policy orientation of domestic substitution provides continuous impetus for industry development

1. Supply Chain Risk
   : Restrictions on advanced process foundry may affect product iteration
2. Market Competition Risk
   : NVIDIA continues to innovate, and competitors such as Huawei are also accelerating their layout
3. Commercialization Risk
   : Cash flow management is crucial amid sustained losses
4. Valuation Risk
   : The market value of some companies has deviated from fundamentals, leading to potential correction risks

---

[1] Securities Times Network - “China’s First General-Purpose GPU Player Iluvatar CoreX Officially Lists on HKEX, Opening Up 31.54%” (https://www.stcn.com/article/detail/3578078.html)

[2] Eastmoney - “Iluvatar CoreX Rings the Bell on HKEX: What Gives This Shanghai GPU Player Confidence?” (https://finance.eastmoney.com/a/202601083612718026.html)

[3] Caixin - “Domestic GPU Player Iluvatar CoreX Opens Over 30% Higher on First Day of HK Listing” (https://companies.caixin.com/2026-01-08/102401588.html)

[4] Securities Times - “Domestic GPU Players Race to List, None Have Exceeded 1% Market Share” (https://www.stcn.com/article/detail/3546594.html)

[5] Yicai Global - “Year-End Review: Domestic Computing Power Underpins the Surge of GPU New Stocks in the Capital Market” (https://www.yicai.com/news/102976152.html)

[6] Huxiu - “China’s NVIDIA-like Players List One After Another, But the Test for Domestic GPUs Has Just Begun” (https://m.huxiu.com/article/4819351.html)

[7] Sina Finance - “Breaking Through in 2025: Domestic GPUs List Collectively, What’s Next?” (https://finance.sina.com.cn/stock/relnews/cn/2025-12-09/doc-inhaeyuc1398122.shtml)