Cambricon's market value exceeding one trillion is the most unadorned realism.

On June 30, the stock price of Cambricon broke through 1,600 yuan during intraday trading, and its total market value crossed the threshold of one trillion RMB. This is the first stock with a market value of one trillion on the STAR Market since its opening seven years ago.

If someone had told Chen Tianshi about this when he founded Cambricon in 2016, I think he probably wouldn't have believed it. And this lack of confidence probably wouldn't be due to technology, but because at that time, hardly anyone could see where the path of AI chips was leading.

The current scene of Cambricon reaching a market value of one trillion also evokes many of my memories. For example, Cambricon became a unicorn just one year after its establishment. Another example is that Cambricon was constantly badmouthed and discouraged before its listing. And also, after its listing, Cambricon was deeply trapped in the whirlpool of losses and had a hard time turning things around.

One of the most profound memories is that in 2020, before Cambricon's listing, someone close to the company told me, "It's either going public or going bankrupt. These are the only two paths in front of Cambricon." And now, just six years later, Cambricon has not only become the leading representative of domestic computing power but also the most beloved and frequently mentioned project among chip investors.

What does Cambricon's one - trillion - market - value mean? The answer an investor gave me is "the dividend of the era and a fortunate encounter for investors."

Ten Years of Entrepreneurship: From 900,000 to One Trillion

I won't go into details about Cambricon's entrepreneurial story.

As mentioned earlier, when Cambricon was established in 2016, it was still at a stage where it didn't know "where its products would be used." The starting point of the Chen brothers, Chen Yunji and Chen Tianshi, was very simple. They wanted to establish a company to combine their research directions - AI algorithms and chips. So, AI chips became the top choice.

At that time, AI chips were still a typical non - consensus track. Cambricon's registered capital was only 900,000 yuan. Among them, Chen Tianshi subscribed to 630,000 yuan, holding 70% of the shares, and the industrialization platform under the Institute of Computing Technology of the Chinese Academy of Sciences subscribed to 270,000 yuan, holding 30% of the shares.

In 2018, Cambricon released its first cloud - based AI chip, the MagicMind 100, but the market reaction was lukewarm. At that time, NVIDIA was at the peak of its power, and the CUDA ecosystem was deeply rooted. In the eyes of most people, an AI chip made by a Chinese startup was just a "PPT product." In 2019, the MagicMind 270 took the edge inference route. There were technological innovations, but it still failed to attract the industry's attention.

This stage also coincided with the change in the semiconductor cycle, which started to move from the previous upward cycle to the downward cycle. On the industrial side, the smartphone market was saturated, the price of memory chips crashed, and the Philadelphia Semiconductor Index (SOX) dropped by about 30% from its peak. The global chip stocks were in a slump.

Of course, the landmark event during this period was the escalation of the China - US trade friction into a technology blockade. ZTE and Huawei were successively punished and sanctioned. The problem of "lacking core chips and operating systems" suddenly became a reality that had to be faced. Correspondingly, the demand for domestic substitution began to emerge. However, Cambricon's product capabilities at that time were not sufficient to meet such demand.

Subsequently, the pandemic swept across the globe, and the semiconductor industry also faced an unprecedented imbalance between supply and demand. On one hand, the work - from - home trend gave rise to an explosive demand. The global electronics and automotive industries were competing for chips at the same time. On the other hand, the chip manufacturing industry came to a standstill. The delivery cycle of NVIDIA's GPUs exceeded twenty weeks, and the price of an RTX 3080 graphics card in the second - hand market was inflated to three times its original price.

Riding on this wave of industry benefits, Cambricon officially listed on the STAR Market at an issuing price of 64.39 yuan. The stock price soared on the first day of listing, enjoying the premium of being "the first AI chip stock." However, the benefits brought by this upward cycle to Cambricon were limited to this. On the product and commercialization fronts, Cambricon's situation was still not optimistic.

In 2021, it incurred a loss of 825 million yuan, 1.256 billion yuan in 2022, and 848 million yuan in 2023. Meanwhile, its annual revenue stagnated at around 700 million yuan. For a listed company, this figure could be euphemistically called "gathering strength for a future leap" or, more bluntly, "standing still."

In the second half of 2022, the global semiconductor industry entered the downward cycle again. The inventory accumulated during the pandemic due to over - ordering began to backfire. The demand for PCs and mobile phones softened, and the collapse of the cryptocurrency market led to a cliff - like drop in the demand for cryptocurrency mining.

The Chinese market was not spared and was in an even more difficult situation. In October 2022, the US Department of Commerce issued the strictest - ever export control on chips to China, restricting the export of advanced computing chips (including AI training chips), semiconductor manufacturing equipment, and supercomputer - related technologies to China, and restricting Americans from supporting the research and development of advanced chips in China.

This means that NVIDIA's GPUs are no longer an option. Domestic chips have to step up whether they are "ready or not." At the same time, the global semiconductor industry has entered the so - called "AI super - cycle."

It's all about timing and luck. At this time, Cambricon's MagicMind series of products just entered the mature stage. The MagicMind 370 is based on a 7nm process and is the first domestic AI chip that can truly be used for large - model training. The MagicMind 590 goes a step further, with its performance more than doubling that of the 370. Although there is still a significant gap compared with NVIDIA's A100/H100, the fact that it "can be used" is its greatest competitiveness.

After that, Cambricon's story started to take a turn for the better. It quickly obtained orders from intelligent computing centers and large enterprises. Its revenue soared from 700 million yuan in 2023 to 6.5 billion yuan in 2025. The company achieved a historic turnaround from losses to profits, with a net profit of 2.059 billion yuan.

The Wind Vane of Domestic Substitution

Although the fate of each chip company is deeply tied to the industry cycle it is in, Cambricon's one - trillion - market - value myth mainly comes from the increasing recognition of the domestic substitution narrative in the capital market.

"Domestic substitution" is an industrial trend and also a kind of belief. However, in the capital market, it gives people a sense of cycle, especially in the semiconductor field. Investors' recognition of "domestic substitution" has been intermittent and capricious.

The debate between technology - industry - trade and trade - industry - technology in the 1980s is too far in the past to discuss in detail. As mentioned earlier, in 2018, the US technology blockade was the starting point of the demand for domestic substitution. Immediately afterwards, the primary market showed a "panic - style" full - scale bet. At that time, as long as a company was related to the "choke - point" problem, whether it was a chip, equipment, or material company, even a "three - no company" with no revenue, no customers, and even no products could obtain high - value financing.

The secondary market was in the same situation. Boosted by the concept of domestic substitution, the semiconductor sector had a doubling - up market. The market was willing to tolerate the high valuation of unprofitable hard - tech companies, believing that "safety takes precedence over profitability."

In the first half of 2021, this wave of enthusiasm reached its peak. The valuation of chip companies in the primary market became bubble - like. A large number of investors from the consumer, software, and new - energy fields turned to the chip industry. At the same time, local guidance funds also flocked to the chip industry on a large scale. The entire industry was in a blind state of having many people, a lot of money, and investing without thinking.

However, in the second half of 2021, the situation in the chip field took a sharp turn, and a setback soon came. More and more investors found that after a large amount of capital flowed in, it did not bring technological breakthroughs and commercial implementation to the chip industry. Therefore, companies that couldn't raise funds, had their valuations halved, or even "closed down" began to flood the market. During this period, nearly 10,000 chip - related enterprises in China were deregistered.

For a while, domestic substitution only became a story - type substitution. For some time after that, funds were concentrated on leading companies, and the market further returned to rationality. Coupled with the shrinking demand caused by the downward cycle, investors began to disenchant domestic substitution. Cambricon also encountered significant financing difficulties at that time.

The summary of an investor is quite representative: "The slogan of domestic substitution is loud, but obviously there are many unreliable companies in the middle, which have taught investors some lessons."

Also during this stage, some Cambricon investors chose to exit. Some investors believe that although in the chip field, supply is determined by policies and international relations, it can't be said that the logic of domestic substitution is completely correct. If one day the market becomes mainstream again and NVIDIA resumes supply, many current products will be ignored.

In short, the narrative of domestic substitution has been up and down, but the overall trend is still spiraling upward. Now, we can see that the entire industry is rapidly establishing a domestic computing power ecosystem.

There's no need to mention Huawei's "Tao's Law" and the Ascend ecosystem. Companies like Cambricon and the "Four Little Dragons of GPU" are all trying to build the second and third technology stacks outside of NVIDIA.

Let me share two recent examples. First, Lisuankeji launched the Lisuankeji LX 7G100. Compared with before, this graphics card has made significant improvements in benchmarking and game frame - rate performance, which is a breakthrough. Another example is from Huawei. There are reports that the upcoming Ascend products have been greatly optimized with the help of Deepseek and Zhipu, which is expected to change the current situation of "using NVIDIA for training and Huawei for inference."

It's not just about chips. The infrastructure above and the applications at the top are all trying to break through the constraints brought by the technology blockade. For example, AI infrastructure manufacturers play a key role in adapting domestic chips to AI servers, trying to use software to shield the interference caused by hardware differences. Another example is that I recently talked to an AI application company. The CTO told me that through underlying technological innovation, they can now achieve the effect of using one graphics card as two when training large models with domestic computing power, which has completely broken the MOE efficiency record.

The sharp rise of Cambricon and the semiconductor industry on June 30 was directly catalyzed by two pieces of news. One is that there are reports that Meituan's LongCat - 2.0 series of models rely entirely on domestic computing power clusters such as Cambricon and Ascend throughout the training and inference process, with the proportion of NVIDIA GPUs being zero. This is a leap for domestic GPUs. In addition, DeepSeek announced the introduction of a peak - valley pricing mechanism, doubling the API price during peak hours, which further proves the scarcity of underlying computing power.

Therefore, Cambricon's one - trillion - market - value is not an isolated event. Viewed in the context of domestic substitution, it happens to be in the most prominent position.

First of all, from losses to profits, Cambricon's story itself has proven that although China's self - built computing power system is not perfect, it has already started to operate, and domestic substitution can also be profitable. Secondly, in the context of global chip competition, a Chinese AI chip company like Cambricon with a market value of one trillion may represent a strategic signal.

Different from many companies that can choose to leave or persevere, regardless of the external environment, Cambricon, with the title of "the first AI chip stock," has to stand there and has no choice. Therefore, its stock price has been like a wind vane in the past six years. Cambricon's one trillion is not the end but the starting point for more people to choose to believe.

A few days ago, Justin Sun shared his insights from his first space journey. In essence, when the spaceship crossed the Kármán line and he first looked down at the planet from space, he suddenly felt that national boundaries disappeared. He thought that what humanity should protect the most is not the territory of power or the ledger of interests, but every tiny yet real life and the planet itself.

That video actually made me feel Justin Sun's sincerity. I also admit that the world of universal harmony he mentioned is indeed something to look forward to. However, the reality is that we have to face various disputes over technology, money, and power. Tariffs still exist, bans still exist, and the entity list still exists. Whether it's domestic substitution or AI competition, at least for now, there is no choice for domestic substitution, and neither is there for Cambricon. This is not romantic and even a bit clumsy, but it is the most straightforward realism of this era.

This article is from the WeChat official account "China Venture Capital", author: Zhang Xue, published by 36Kr with authorization.