Cambrian's First Profitable Year in Decade, Why Did Stock Price Fall Instead of Rise?

Due to factors such as large domestic chip manufacturers developing their own chips and limited production capacity of advanced domestic processes, the stock performance of domestic listed chip companies has been relatively weak recently. Industry experts believe that domestic models have a strong demand for computing power, and the performance of domestic chip manufacturers is expected to further improve.

Text | Special Contributor Kang Guoliang, Reporter Zhang Jianfeng

Editor | Yang Xiuhong

Recently, several domestic AI (artificial intelligence) chip companies have announced their 2025 performance forecasts, generally showing explosive growth.

On the evening of March 12, Cambrian-U (688256.SH) released its 2025 financial report first. The report shows that Cambrian achieved revenue of 6.497 billion yuan in 2025, a significant increase of 453.21% year-over-year; net profit attributable to shareholders was 2.059 billion yuan, turning from loss to profit, compared to a net loss of 452.3 million yuan in 2024. This is Cambrian’s first annual profit since its founding in 2016.

On March 13, the first trading day after the earnings release, Cambrian’s stock opened slightly higher but then retreated. By the close, the stock price fell 0.26% to 1,096.10 yuan per share.

According to the financial report, after turning profitable, Cambrian plans to distribute its first dividend, proposing a cash dividend of 15.00 yuan (tax included) per 10 shares to all shareholders. Additionally, it will transfer 4.9 shares for every 10 shares from capital reserve, which will significantly lower the stock price after the stock split. In August 2025, Cambrian’s stock price once exceeded that of Kweichow Moutai (600519.SH), becoming the highest-priced stock in A-shares, attracting market attention.

Two other chip companies listed in 2025, Moore Thread-U and Mu Xi Co., Ltd.-U, also recently released their 2025 performance forecasts, both showing significant narrowing of losses. The forecasts indicate that Moore Thread’s revenue in 2025 will increase 2.43 times to 1.506 billion yuan, with net loss narrowing by 36.70% to 1.024 billion yuan. Mu Xi’s revenue is expected to grow 1.21 times to 1.644 billion yuan, with net loss narrowing by 44.53% to 781 million yuan.

In fact, the accelerated turnaround of domestic AI chip companies is closely related to the surge in demand for computing power driven by AI development and the accelerated trend of domestic chip substitution. Their revenues have been growing rapidly in recent years. The financial reports show that from 2022 to 2025, Cambrian’s revenue increased from 729 million yuan to 6.497 billion yuan, a 7.9-fold increase; Moore Thread’s revenue grew from 46 million yuan to 1.506 billion yuan, a 31.7-fold increase; Mu Xi’s revenue surged from 426,400 yuan to 1.644 billion yuan, a staggering 3,854.5-fold increase.

Especially after going public, domestic AI chip companies have entered a new development stage. “The GPU (graphics processing unit) industry has high technical barriers, requires large R&D investment, and has long R&D cycles. Capital market support not only provides us with more abundant R&D funds but also significantly enhances brand influence and attracts top talent, enabling us to invest more strongly in independent innovation of domestic GPUs,” said Moore Thread, a leading domestic GPU chip company, to Caijing.

However, the stock prices of domestic AI chip companies have performed poorly this year. Cambrian, Moore Thread, and Mu Xi have all experienced declines since the beginning of the year.

“Recently, the stock prices of Cambrian and other domestic computing power sector stocks have been weak, partly due to the previous performance vacuum caused by rumors, and also related to factors such as NVIDIA’s H200 export approval to China, large companies developing their own chips, and limited capacity of advanced domestic processes, which have affected optimistic expectations for future performance. Especially regarding capacity, Cambrian in 2025 faces some delivery issues, but the steep growth in domestic demand for computing power this year means that the demand for domestic models is very strong, and delivery issues will gradually be resolved, allowing performance to further release,” said a fund manager from a private equity firm in Beijing to Caijing.

GF Securities stated that Cambrian’s revenue in 2025 grew significantly, with impressive expansion in the internet sector. It is expected that in 2026, the company’s AI chip sales to internet and other commercial clients will reach a new level. The company’s inventory increased sharply in the fourth quarter, reflecting optimism about product delivery in subsequent quarters and supporting sustained high growth and certainty of future revenue. Compared to NVIDIA’s H200 chip sales uncertainty in China, the improved performance of domestic AI chips and the recovery of the supply chain provide more stable computing power support for domestic tech companies. As the company’s industry position and technological competitiveness are increasingly recognized by the market, its long-term premium is expected to further emerge.

CITIC Securities noted that autonomous control of computing power is a clear direction. The development of domestic computing power will focus on achieving independence and reducing reliance on external technologies, enhancing the competitiveness of the domestic AI chip industry. As AI computing power demand increases, domestic chips will gradually evolve from single cards to system-level integration, improving overall performance and efficiency. 2026 will be a key year for the volume growth of domestic computing power from 1 to N.

Performance Accelerates Turnaround

Based on the performance forecasts of Cambrian, Moore Thread, and Mu Xi for 2025, all three domestic AI chip companies are expected to see significant revenue growth and either turn profitable or accelerate their loss reduction.

As a domestic leader in computing power, Cambrian achieved its first annual profit since its founding in 2016. According to the financial report, Cambrian’s 2025 revenue was 6.497 billion yuan, up 453.21% YoY; gross profit was 3.583 billion yuan, up 437.99%. While revenue surged, the company also turned from loss to profit for the full year, with net profit attributable to shareholders of 2.059 billion yuan, and net profit after non-recurring gains and losses of 1.77 billion yuan.

Cambrian stated that benefiting from the continuous rise in AI industry demand for computing power, the company has expanded its market through excellent product competitiveness and actively promoted AI application scenarios, leading to a substantial YoY increase in revenue in 2025.

The financial data shows that Cambrian’s profitability improved significantly, with gross margin reaching 55.15% and net margin 31.68% in 2025. Although gross margin slightly declined from 56.71% in 2024, net margin improved sharply from -38.91% in 2024. In 2025, R&D expenses were 1.351 billion yuan, up 11.1% YoY, with R&D expense ratio dropping sharply from 103.57% in 2024 to 20.79% in 2025. Return on equity reached 26.96%. However, compared to NVIDIA’s net margin of 55.6% and ROE of 101.49% in fiscal year 2026 (from January 27, 2025, to January 25, 2026), there is still a gap.

The report also shows that in 2025, Cambrian produced and sold 128,000 and 117,400 units of intelligent chips and boards, respectively. Its inventory at the end of 2025 reached 4.9 billion yuan, about 1.2 billion yuan more than at the end of Q3 2025. According to the annual report, Cambrian’s inventory of chips and boards was 857,000 units, supporting further delivery to customers.

After turning profitable, Cambrian plans to distribute dividends for the first time, proposing a cash dividend of 15.00 yuan per 10 shares (tax included), totaling 632 million yuan, accounting for 30.71% of net profit attributable to shareholders in 2025. It will also transfer 4.9 shares for every 10 shares from capital reserve.

The financial report also shows that well-known market investor Zhang Jianping increased his holdings of Cambrian by 408,400 shares to 6.8149 million shares, accounting for 1.62% of circulating shares, ranking fifth among shareholders. Based on the March 13 closing price, his holdings are worth nearly 7.5 billion yuan. This marks Zhang Jianping’s second consecutive quarter of increasing his position in Cambrian, having added 320,200 shares in Q3 2025.

Notably, with the achievement of annual profitability in 2025, Cambrian will successfully delist from the “Science and Technology Innovation Board Growth Tier” on March 16, and its stock name will change from Cambrian-U to Cambrian. The Growth Tier was established in July 2025 specifically for unprofitable tech companies, with a new and old tier separation rule: companies that become profitable for the first time can exit the tier, while new listings must meet higher profitability standards. Cambrian’s exit from the Growth Tier makes it one of the first “delisted” companies from this tier.

Meanwhile, two domestic GPU leaders listed on the Hong Kong Stock Exchange, Moore Thread and Mu Xi, also achieved rapid revenue growth and significant reduction of losses in 2025.

According to previous forecasts, Moore Thread’s 2025 revenue will reach 1.506 billion yuan, up 2.43 times; net loss attributable to shareholders will narrow by 36.70% to 1.024 billion yuan compared to 2024’s loss of 1.618 billion yuan. The loss reduction is driven by the booming AI industry and high-performance GPU demand. Moore Thread stated that the company is still in a high R&D investment stage and, compared to international giants, still lags in overall R&D strength, core technology accumulation, and product ecosystem. It is not yet profitable and has accumulated losses.

Another listed domestic GPU company, Mu Xi, forecasted revenue of 1.644 billion yuan in 2025, up 1.21 times; its net loss narrowed by 44.53% to 781 million yuan from 1.409 billion yuan in 2024. This is also Mu Xi’s first annual loss reduction since 2022.

“In 2025, the company adhered to the ‘1+6+X’ development strategy, increased market expansion, and continuously enhanced its position and influence in the high-performance GPU industry, promoting deep integration of AI technology across various industries,” said Mu Xi. The company’s GPU shipments increased significantly, driving revenue growth. The substantial YoY revenue increase and reduced share-based payment expenses positively impacted 2025 profits, reducing losses and showing a trend of improving performance.

Mu Xi also released a forecast for Q1 2026, expecting further narrowing of losses: revenue is expected to be 400 million to 600 million yuan, a YoY increase of 24.84% to 87.26%; net loss attributable to shareholders is expected to be 91 million to 182 million yuan, a reduction of 21.93% to 60.97%.

In its Q1 forecast, Mu Xi stated that benefiting from the rapid development of the AI industry and relying on excellent product performance and a complete software ecosystem, its products and services have been widely recognized by downstream clients, leading to significant YoY growth in business scale.

It is noteworthy that although Moore Thread and Mu Xi are not yet profitable in 2025 and still in loss, both companies have made predictions about when they will turn profitable.

Mu Xi said that based on its operational situation, market potential, market share changes, customer repurchase and new customer expansion, the earliest expected breakeven point is 2026.

Moore Thread predicts that it may achieve consolidated profitability in 2027, which will include government subsidies. After deducting this subsidy, the company will be in a marginal profit state.

Currently, the two domestic GPU companies listed in Hong Kong have not disclosed their 2025 performance, but their interim reports show increasing net losses compared to the previous year.

Biren Technology achieved revenue of 58.9 million yuan in the first half of 2025, up nearly 50%, but its net loss attributable to shareholders was 1.601 billion yuan, an increase of 80.17%. Tianzhi ZhiXin achieved revenue of 324.3 million yuan, up 64.24%, but its net loss was 609.3 million yuan, an increase of 50.82%.

Why is revenue growing?

As a leading domestic AI chip company, Cambrian was the first to break even in the industry and successfully achieved annual profit in 2025. This was mainly due to its domestically produced ASIC (Application-Specific Integrated Circuit) chips used in AI data centers gaining recognition from major domestic cloud providers, entering the supply chain of large companies and securing large orders. Cost dilution effects helped it turn losses into profits.

According to the financial report, Cambrian’s revenue exploded starting from Q3 2024. In Q3 2024, quarterly revenue reached 120.5 million yuan, a 284.59% increase YoY; in Q4, revenue further increased by 75.51% to 989.2 million yuan, with the company first achieving quarterly net profit of 272.2 million yuan.

The 2024 annual report shows that the largest customer’s sales reached 930 million yuan, nearly doubling from 467 million yuan last year, a 99.1% YoY increase. The combined sales to the second to fifth largest customers totaled 182 million yuan, slightly down from 188 million yuan last year, indicating that the first customer’s procurement significantly boosted Cambrian’s revenue.

Notably, entering 2025, Cambrian’s customer concentration decreased. In 2023-2024, the top five customers accounted for 92.36% and 94.63% of revenue, respectively; in 2025, the top five customers’ sales totaled 5.76 billion yuan, accounting for 88.66% of revenue. Among these, the third customer is a long-term partner, while the others are new customers. The sales of the top five customers were 1.703 billion, 1.401 billion, 764 million, and 655 million yuan, respectively.

In late February, industry sources reported that Cambrian’s 590-chip entered Tencent’s supply chain. Previously, in September 2025, there were reports that Alibaba had increased its order of Cambrian’s Si Yuan 370 chips to 150,000 units. However, these reports have not been confirmed.

Cambrian’s 2025 financial report mentions that its products are widely used in large model algorithm companies, server manufacturers, AI application companies, and support the rapid development of AI across industries such as cloud computing, energy, education, finance, telecommunications, healthcare, and the internet. In the internet sector, the company focuses on large models and multi-modal applications, collaborating deeply with industry clients on operator development, performance optimization, framework improvements, and communication enhancements.

An industry insider told Caijing that domestic AI chip companies mainly fall into two categories: large established firms and independent startups. Each has its advantages and disadvantages.

He explained that, in terms of channels, when product performance is comparable, large firms’ chips are more favored by major buyers, as these companies prefer to support their own chip brands. However, these firms tend to have weaker customer acquisition capabilities outside their own ecosystem. Non-large-firm AI chip companies face higher hurdles entering major procurement systems but may have stronger expansion capabilities outside large firms. Fortunately, non-large-firm chips may also enter major procurement channels if they demonstrate good performance and cost-effectiveness, which is increasingly possible as the market values performance-to-price ratio.

Why are stocks cooling off?

It is noteworthy that the stock performance of the five listed domestic AI chip companies has been generally poor this year.

Moore Thread and Mu Xi, listed on the Hong Kong Stock Exchange at the end of 2025, experienced a hot start but then fell into a slump. As of March 12, their stock prices had fallen 40.79% and 41.9% from their peaks, respectively.

Earlier, Cambrian’s stock price surged after listing from 2023 to 2025 but then slowed down in 2026. Although Cambrian achieved profitability in 2025, its stock price has been declining since reaching a high of 1,550 yuan per share on January 12, 2026. As of March 12, it had retreated 29.1%, with market capitalization dropping from 654.1 billion yuan to 463.4 billion yuan, a decrease of 190.7 billion yuan.

The two domestic GPU companies listed in Hong Kong, Biren Technology and Tianzhi ZhiXin, show divergent performance. Biren’s stock peaked at 42.88 HKD per share after listing but then declined, falling to 33.22 HKD as of March 12, a 22.53% drop from the high. Tianzhi ZhiXin’s stock has performed relatively well, continuously reaching new highs after listing.

Why do domestic AI chip companies’ earnings improve while their stock prices remain weak? Market analysts believe that, firstly, high valuations at listing affected their performance. Moore Thread and Mu Xi listed at the end of 2025, benefiting from the scarcity of targets, the global AI wave, and domestic substitution trends, as well as high initial returns and relatively small circulating shares, attracting speculative funds. Their initial peaks reached 941.08 yuan and 895 yuan per share, respectively, a 7.23 and 7.55 times increase over the issue prices of 114.28 yuan and 104.66 yuan. As of the third quarter of 2025, their price-to-sales ratios were extremely high—468 times and 189 times—much higher than Cambrian’s approximately 108 times.

As speculative funds retreat and restrictions on share lock-ups loosen, valuations are normalizing. As of March 11, the price-to-sales ratios of Moore Thread and Mu Xi were 285 times and 122 times, respectively.

Secondly, as more domestic AI chip companies go public in A-shares and Hong Kong stocks, the scarcity effect diminishes, and the market begins to scrutinize each company’s fundamentals more carefully, focusing on future performance trends and industry competition.

Thirdly, competition in the domestic AI chip market is intensifying. Cloud service providers are also rushing to develop their own AI chips. For example, Alibaba’s Pingtouge’s Wuzhi PP chips have shipped tens of thousands of units, surpassing Cambrian and leading among domestic GPU vendors, though this has not been officially confirmed by Cambrian. Baidu’s Kunlun Xin submitted a confidential listing application to HKEX in January, supplying chips for internal AI projects and other industry clients.

More notably, ByteDance, previously thought to be a major customer of Cambrian, is developing an AI chip codenamed SeedChip, has contacted Samsung for foundry services, and plans to receive samples by the end of March. It aims to produce at least 100,000 AI inference chips in 2026, gradually increasing to 350,000 units.

Additionally, the high-end AI chips from NVIDIA, the H200, were allowed for export to China by the U.S. Trump administration on December 8, 2025. This marked the first “limited relaxation” after the full ban from 2022 to 2025. However, on February 24, U.S. Commerce Department official David Peters confirmed at a congressional hearing that, although the export license was approved, NVIDIA has not yet sold any H200 chips to Chinese customers, with sales recorded as zero. Nonetheless, this remains an important factor affecting investor confidence and future optimistic performance expectations for domestic chip companies.

Apart from market competition, the limited capacity of domestic advanced process manufacturing is a major bottleneck. Delivery is slow, which hampers performance growth. “The biggest problem in China’s AI chip industry is actually the supply chain—capacity constraints,” said an insider from a domestic chip firm to Caijing. Foreign chip manufacturers can sustain supply of advanced process chips at 3nm or even more advanced nodes, but China’s most advanced mass-produced process is equivalent to 7nm, with limited output.

Guotou Securities believes that as China’s models surpass the U.S. in call volume, the explosive growth in inference demand is driving the growth of computing power chip demand. Under the constraints of limited high-end supply overseas, domestic computing power is making breakthroughs in density, ecosystem adaptation, and commercial application, gradually replacing foreign supply at scale.

Guotou Securities also states that under dual external and internal demand, domestic computing power is shifting from “available” substitutes to “mainstream” solutions. With increased capital expenditure by domestic cloud providers, widespread use of lightweight models, and a maturing domestic ecosystem, infrastructure for domestic computing power is expected to be deployed at scale in government, finance, internet, and smart manufacturing sectors, creating value.