Cambricon founder’s net worth soars to $22.5b

Chen Tianshi, founder of Beijing-based AI chip maker Cambricon Technologies, has seen his net worth rise to US$22.5 billion as of November 2025, according to the Bloomberg Billionaires Index.

Cambricon, which designs AI accelerators, reported its first net profit in Q4 2024 after a surge in demand for domestic chips in China.

The company’s revenue jumped over 500% in the past year, fueled by Beijing’s push for local tech companies to use Chinese-made chips following US export controls on advanced semiconductors.

Cambricon’s shares have climbed more than 765% in the last two years, with most of Chen’s wealth coming from his 28% ownership stake.

The company faces competition from Chinese tech firms like Huawei, and remains under US sanctions, which restrict its access to Western technology.

Analysts say Cambricon’s growth is closely tied to government support and policy mandates, rather than the competitiveness of its products.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Cambricon depends on a concentrated client base

In H1 2025, 99.6% came from cloud AI chips (data center accelerators) 1. Top five buyers made over 94% of sales, and the largest near 80% is likely Alibaba or its cloud unit 1. ByteDance pre-ordered 200,000 chips, yet deliveries slipped due to bottlenecks 2. Alibaba, Tencent, and Baidu are testing with no large buys confirmed 2.
Any cutbacks or delays by big clients open a revenue hole 1. The stock trades at a rolling Price-to-Earnings (P/E) of 5,117.75 versus an industry average of 88.97 2. That mismatch looks risky even with strong growth.

AI chip service firms can fill the software integration gap

Cambricon shipped just over 10,000 chips in 2024, while Huawei moved 300,000 to 400,000 Ascend units 2. Demand from ByteDance, Tencent, and others pushes inference workloads (running trained AI models) to domestic alternatives 3. That opens room for third-party AI infrastructure firms that handle porting, optimization, or managed inference on Cambricon gear.
The Siyuan 590 data center accelerator delivers about 80% of NVIDIA’s A100 performance on China’s 7nm process 1. It fits mid-size server clusters of 1,000 to 3,000 cards 4. Providers that migrate clients from NVIDIA’s CUDA to Cambricon’s inference toolchain (software stack) can tap a market set to reach ¥178 billion ($24.28 billion) in 2025 5.
Investors can find which Chinese clouds already offer Cambricon instances (virtual machines powered by Cambricon chips) to spot near term integration targets and partners.