Kunlunxin Series D Financing and Analysis of AI Chip Technology Barriers

Kunlunxin’s financing history reflects continuous attention from the capital market to its development prospects. According to public information, Kunlunxin’s financing trajectory is as follows [1]:

• Seed Round Financing
  : Completed the seed round when it became independent in 2021, led by CPE Source Peak, followed by IDG, Legend Capital, YH Capital, etc., with a company valuation of approximately RMB 13 billion [1][2]
• Series A to Series D
  : From 2022 to 2025, Kunlunxin completed multiple financing rounds from Series A to Series D, attracting well-known investors including BYD, CITIC Construction Investment Capital, Legend Capital, Zhongguancun Science City, and Beijing Artificial Intelligence Industry Investment Fund [1][2]

• Baidu remains Kunlunxin’s largest shareholder, holding a 59.45% stake
• Beijing Kunlun Legend Technology Partnership holds approximately 4.61%
• Tianjin Jiuzhang Kunlun Enterprise Management Partnership holds approximately 3.57% [2]

Baidu has decided to spin off Kunlunxin for independent listing on the Hong Kong Stock Exchange, and is currently undergoing the listing application process. The outside world expects its valuation to reach the HK$100 billion level, mainly based on its commercial progress such as winning a 1-billion-level (RMB) bid from China Mobile [2][3].

---

Kunlunxin adopts a self-developed XPU architecture, which has gone through multiple generations of iterations. According to market research feedback:

Technical Indicator	Kunlunxin’s Performance	Industry Position
CUDA Compatibility	Reaches or even surpasses industry leading enterprises	Obvious competitive advantage [2]
2024 Shipment Volume	69,000 units	Approximately 2.65 times that of Cambricon [1]
Product Iteration Speed	Launches new products annually	Maintains technical update pace [1]

• Kunlunxin M100
  (launched in 2026): Optimized for large-scale inference scenarios, focusing on cost-effectiveness
• Kunlunxin M300
  (launched in 2027): Targets ultra-large-scale multimodal model training and inference needs
• Tianchi Super Node Product
  (2026): A single Tianchi 512 super node can complete trillion-parameter model training [1]

• In August 2025, in China Mobile’s 2025-2026 AI General Computing Equipment Procurement Project, AI server products based on Kunlunxin won 70% share in Bid Package 1 and Bid Package 2, and 100% share in Bid Package 3, indicating a substantial breakthrough in its commercialization path [1][2]

• BOC International’s research report points out that Kunlunxin has reached or even surpassed many leading enterprises in the industry in many aspects, especially having obvious advantages in CUDA compatibility [1][2]
• This reduces the migration cost for developers and is conducive to ecosystem construction

• According to IDC’s 2024 China Accelerated Computing Chip Shipment Report, Kunlunxin’s shipment volume reached 69,000 units, ranking among the top domestic AI chips [1]

---

The track where Kunlunxin is located is extremely competitive; the global market faces strong competition from NVIDIA and AMD [3]:

• International Competition
  : NVIDIA has built extremely high competitive barriers relying on its GPU ecosystem
• Domestic Competition
  : Huawei Ascend, Cambricon, Moore Thread, Muxi, and other enterprises each have their own focuses
• Cloud Vendor Self-Development
  : Large cloud vendors such as Alibaba’s Pingtouge are also developing AI chips independently [3]

• For a long time, Kunlunxin has mainly served the Baidu ecosystem, so its external awareness is not high
• Compared with competitors like Ascend that have formed a relatively complete industrial ecosystem, Kunlunxin’s ecosystem construction still needs to be strengthened [2]

• Huawei Ascend 910B is basically comparable to A100, and the performance of 910C is expected to be comparable to H100
• New generation products from Cambricon, Hygon Information, and other enterprises are also catching up, leading to fierce substitution competition [2]

:

• Significant commercial progress (winning a 1-billion-level bid)
• Leading shipment volume in the industry
• Technical iteration capability of launching new products annually
• Supporting role of the Baidu ecosystem

:

• Whether the valuation has fully reflected the risk of intensified competition
• Whether it can maintain technical leading advantages after independent listing
• Whether ecosystem construction can catch up with industry leading enterprises

---

:
There is a certain risk that Kunlunxin’s AI chip technology barrier is overestimated, but it is not completely overestimated. Its CUDA compatibility and commercialization capability are its core competitive advantages, but ecosystem construction and market competition are still challenges that need continuous attention.

:
Open Source Securities’ research report points out that considering Kunlunxin is expected to drive valuation improvement, it maintains a “Buy” rating on Baidu [1]. If Kunlunxin can successfully list in Hong Kong and continue to maintain technical iteration and commercial advantages, its HK$100 billion-level valuation may be recognized by the market.

:

• Intensified market competition may lead to erosion of market share
• Ecosystem construction progress may fall short of expectations
• Technical iteration speed needs to remain leading in the industry

---

[1] Time Finance - “Baidu Decides to Spin Off Kunlunxin for Independent Listing in Hong Kong; Stock Price Hits Two-Year High” (https://www.tfcaijing.com/article/page/746667453236743947494a366c522b4a726537424b773d3d)

[2] Sina Finance - “Baidu Kunlunxin Goes Viral: Comparing to Cambricon, Valuation May Reach HK$100 Billion?” (https://finance.sina.com.cn/jjxw/2025-09-06/doc-infpqkky1110458.shtml)

[3] Sina Finance - “Baidu Kunlunxin Confidentially Submits Hong Kong Listing Application; AI Chip Business Spinoff Accelerates” (https://finance.sina.com.cn/stock/t/2026-01-02/doc-inhewqny9428276.shtml)