NVIDIA dominates the global AI chip market. Meanwhile, China is building a parallel chip ecosystem, one largely forged by geopolitics. Since early last year, NVIDIA shares have risen about 270%, while Cambricon has jumped roughly tenfold. In 1H 2025, Cambricon said revenue was up 44 times.
Cambricon’s share price performance caused a brief flurry of attention on Wall Street. Many investors realized for the first time that there was a Chinese AI chip company worth several hundred billion RMB, and they knew little about it.
In the latest episode of the VOC podcast, our two hosts, Diane and Aaron, who track U.S. and China tech closely, highlight a turning point most people missed. It was not the stock surge of 2025. It was the U.S. sanctions in 2022.
From Near Collapse to Market Hero
Cambricon was founded by Chen Tianshi, formerly a researcher at the Chinese Academy of Sciences. In 2016, well before ChatGPT, he left academia to start the company.
In 2017 to 2018, Huawei HiSilicon contributed nearly all of Cambricon’s revenue. Cambricon’s model then was IP licensing: it designed chip architectures and licensed them to HiSilicon for use in smartphone processors. When HiSilicon decided to build its own AI processors and stopped buying, Cambricon’s revenue evaporated. The founder made a hard choice: do not quit, pivot. Cambricon moved from IP licensing to building full AI chips, a capital intensive, high risk bet.
Cambricon went public in 2020. Domestic investors piled in because it was the only listed pure play AI chip company in China. The stock hit 250 yuan, then reality arrived: massive R&D and annual losses. By 2023, the stock fell below 30 yuan. A U.S. hedge fund even published a short report. It looked like a failure story.
But as Aaron noted on the show, the real inflection point was not the 2025 rally. The real inflection point was 2022. In December 2022, the U.S. put Cambricon on the Entity List. Immediate threat: no more access to TSMC. At the same time, the U.S. barred NVIDIA from exporting high end AI chips to China.
Suddenly, China’s biggest cloud players, Alibaba, Tencent, ByteDance, could not get NVIDIA chips. They were building AI infrastructure with the supplier cut off. Cambricon became one of the few remaining options.
Here is the key idea. Aaron was blunt about it on the show: geopolitics is economics. Political power now drives tech demand. U.S. restrictions created the problem. The problem created a market. The market saved Cambricon.
There was still a big constraint: without TSMC, Cambricon had to rely on SMIC.
Back in 2022, SMIC was not yet producing at 7nm. It took roughly two years to get there. Then, in 2024, the pieces clicked: SMIC’s 7nm became available; the ChatGPT wave went global; and China’s tech giants realized they needed compute, fast. Capex for AI data centers at Alibaba, Tencent, and ByteDance grew more than 45%, outpacing the Magnificent Seven in the U.S., which grew around 30%.
Cambricon posted its first quarterly profit in Q4 2024. In the first half of 2025, revenue jumped 44x.
Success by Necessity
Aaron offered an uncomfortable observation: many successful Chinese tech companies did not win because of foresight, but because they were forced to adapt.
Cambricon’s two major pivots, losing Huawei in 2019 and being sanctioned in 2022, were not born of grand strategy, but survival. The company had no choice but to evolve.
Diane agreed. Resilience matters. Many firms faced the same pressures. Not all survived. Cambricon did.
Winners and Rivals
Cambricon now has competitive chips. Its Siyuan 590 is roughly comparable to NVIDIA’s A100 in performance. But NVIDIA does not just sell chips; it sells an ecosystem, CUDA, tools, libraries, community, built over decades with powerful network effects.
Cambricon has the hardware, but its ecosystem is far from CUDA level maturity. Both Diane and Aaron think there is a ceiling. Cambricon can lead in China, but it will not globalize easily. It will remain a Chinese company. That is not necessarily bad, China’s market is huge, but it means Cambricon is unlikely to be the next NVIDIA. It is more likely to be China’s NVIDIA.
The fiercest competition is domestic. Huawei is pushing Ascend. Alibaba has chip plans. Tencent and Baidu are also in the mix. Today, these companies say their chips are for internal use. Aaron is skeptical. Once the tech matures, they will sell externally. They will buy from each other and from Cambricon. China is likely to end up with three to five major players. No single winner takes all. Cambricon may win this round, but probably not the entire game.
Valuation and the Lack of Alternatives
Cambricon trades at over 150x earnings. NVIDIA is around 30x. It looks bubbly.
But as Aaron pointed out, Chinese investors may not care because they do not have many options. HiSilicon, Alibaba’s chips, Tencent’s chips, none are listed. Other startups are not public yet.
If you are a Chinese investor who wants AI chip exposure, your listed choice is basically Cambricon. Under that constraint, buyers show up regardless of valuation. The question is not Is it worth this price The question is Do you want exposure to China’s AI chips If yes, this is the one you can buy.
Founder Chen Tianshi and the leadership team have not sold large blocks of shares since the IPO. On paper, they are wealthy, but have not cashed out. In China, management selling heavily is often read as a negative signal. If they start selling, it could mean the story is topping out.
America’s Strategic Mistake
Aaron echoed a point Jensen Huang had hinted at: if the U.S. wanted to slow China’s tech progress, the most effective approach would be to sell China cheap, high quality chips so China remains dependent.
The bans did the opposite. They forced China to build. They sped up a process that might otherwise have taken decades.
That is the irony: export controls aimed at limiting China’s AI capacity may have accelerated the country’s push to develop an independent chip stack.
If this topic interests you, I highly recommend watching Diane and Aaron’s English language video conversation on this episode, and joining the discussion in our Substack comments.
Disclaimer: The information in this episode is provided for informational purposes only and does not constitute legal or financial advice. All commentary reflects the personal views of the guests, based on publicly available information and first-hand operational experience.
