Since September 24, the A-share market has witnessed an astonishing surge, with the STAR Market, which had been on a prolonged downturn, among the top-performing sectorsNotably, Cambricon Technologies, a significant player within this sector, has reached an all-time high in its stock price, pushing its market value beyond 140 billion yuan (about 20 billion USD).

Cambricon, which made its debut in the stock market four years ago, has experienced significant stock price declines mirroring those of the STAR Market as a whole, resulting in skepticism surrounding its performanceNow, having rebounded to historic highs, the company needs to validate its worth through tangible achievements.

In this recent stock rally, Cambricon's upward trajectory commenced at the end of 2022, well before the broader market began to rise.

During that period, Cambricon's stock had plummeted over 80% from its debut price, seemingly following the path of other large companies that faced similar downtrends.

The catalyst for Cambricon's revival can be traced back to the advent of ChatGPT and the resulting surge in the AI sector, along with Nvidia's robust expansion

Advertisements

This frenzy has spurred intense competition among Chinese AI giants and signaled the emergence of a domestic substitution trend, thereby creating a burgeoning demand for domestic computing power and AI chips, which has significantly lifted Cambricon's stock price.

In 2023 alone, Cambricon's stock price skyrocketed by 535%, the highest among nearly 700 AI-related stocks listed on the A-shares; it also recorded a remarkable 157% increase in 2024, consistently ranking among the best performers in the AI field.

However, the rapid rise in Cambricon's stock price appears somewhat disconnected from realityUnlike companies that dominate the upstream supply chain—securing vast profits long ago—Cambricon finds itself in the middle of the industry with no significant earnings backing its newfound wealth.

Despite eight years of operation, Cambricon has reported losses across all seven of its annual financial statements, accumulating over 5 billion yuan in total losses

Advertisements

In the first half of this year alone, the company faced a loss of 530 million yuan.

If the early-stage profitability challenges of tech companies are somewhat understandable, the decline in R&D and revenue points to a chilling trendIn 2023, Cambricon's revenue dropped to 709 million yuan, a 2.7% decline compared to the previous yearFurthermore, its research and development expenses fell to 1.118 billion yuan, illustrating a significant 26.6% dropThis marked the first annual decrease in both crucial financial indicators that define nascent tech companies.

In the first half of this year, the company's revenues were recorded at just 64.76 million yuan, reflecting a staggering 43% drop year-on-year, with R&D expenditures decreasing by 7.3% to 447 million yuan.

As the stock price diverges from actual performance metrics, some savvy investors seem poised to exit, with several pre-IPO shareholders having already chosen to reduce their stakes at higher prices or even completely cash out

Advertisements

For instance, iFlytek, which used to hold 1.17% of shares, exited from the top 10 circulating shareholders by the third quarter of 2022. Similarly, Alibaba's investment arm, which previously owned a 1.74% stake, also vanished from the top shareholder list by 2023.

As of October 2023, all five founding shareholders of Cambricon, including GITIC Venture Capital and various financial institutions, had realized substantial reductions, cashing out more than 4.1 billion yuan collectivelyDespite this, the enthusiasm among investors in the secondary market appears to persist, with the number of shareholders increasing, reflecting a level of confidence in Cambricon even at its historic highs.

Founded only eight years ago, Cambricon has aligned itself closely with the AI craze, emerging amidst a time of explosive technological development.

In 2016, the AI landscape took a significant turn when Google's AlphaGo defeated renowned Go player Lee Sedol, signifying an acceleration in the development of global AI technologies

• Yen Sees Increased Volatility
• Fed Signals New Direction for Interest Rates
• The Collapse of the Korean Stock Market
• CNY Faces Further Depreciation After Breaching 7.30
• Gold Prices Surge to Two-Week High

It was also in that year when the Institute of Computing Technology under the Chinese Academy of Sciences introduced the world's first "deep learning" neural network processor chip, which was aptly named "Cambricon."

In a milestone for the sector, Chen Tian Shi, one of the project's leaders, drew upon this innovation to transition it into a commercially viable business, assembling a team of industry veterans from the research institute to establish one of China's leading firms in the AI chip field.

Just a year after its inception, Cambricon secured a partnership with Huawei, whose Kirin 980 chip's AI capabilities were enhanced by Cambricon's 1H processorFrom 2017 to 2019, sales from licensing its IP to Huawei exceeded 117 million yuan, making it nearly the entirety of the company’s income.

By 2018, Cambricon had begun to develop its own chips, releasing the Suyuan 100 cloud-based intelligent chip, followed by the Suyuan 270 and Suyuan 220 chips, establishing a comprehensive product array that spanned edge, cloud, and terminal applications.

The combination of a technically astute founder, a robust research team, an extensive product line, and the backing of Huawei, dubbed the AI juggernaut, ensured that Cambricon received substantial attention from venture capitalists

Within its first two years, the company completed six rounds of financing, raising a total of 4.6 billion yuan and achieving a pre-IPO valuation exceeding 20 billion yuan, placing it among China's most noted AI unicorns.

In July 2020, as the STAR Market surged through a bull phase, Cambricon entered the market as the first AI chip company, with its opening day stock price soaring threefold, propelling its market valuation into the trillion yuan range.

However, this zenith from its initial public offering (IPO) also heralded an undercurrent of challengesIn 2019, Huawei embarked on self-sufficiency in AI chips, which not only deprived Cambricon of a vital client but introduced a formidable competitor.

Between 2020 and 2022, the revenue growth rate of Cambricon plunged significantly without Huawei’s patronage, while Huawei's Ascend line rapidly garnered market share in the AI chip sector.

Statistics reveal that in 2022, out of approximately 1.09 million AI accelerator cards shipped in China, NVIDIA dominated with 85% market share, while Huawei claimed 10%, BAIDU 2%, and both Cambricon and Suiyuan technology barely reached 1%. The market evidently recognized Huawei as a more triumphant local alternative after the restrictions placed on NVIDIA’s chips.

Shifts in the industry structure and management circumstances have exerted formidable pressure on Cambricon’s stock value

Over the two-plus years post-IPO, the company has seen its stock price nosedive, and its once-pristine reputation is fading.

Despite being a relative newcomer compared to established heavyweight competitors like NVIDIA and Huawei, Cambricon's singular focus on AI chips has equipped the young company with a certain degree of competitive strengthThis foundation may play a critical role in re-propelling its stock prices back to historic highs.

As early as 2021, Cambricon launched its first intelligent training chip, the Suyuan 290, followed by the Suyuan 370—an integrated training and inference productIt could deliver an inference capability of 256 TOPS, or 40% of the NVIDIA A100’s potential, with training capabilities reaching 128 TFLOPS—approximately 20% of the A100's powerSignificant quantities were sold to server producers such as Alibaba Cloud and Inspur, as well as algorithm companies like iFlytek.

In 2023, the company introduced the next-generation Suyuan 590 smart chip, aimed directly at the A100 market

Internal testing by Baidu has indicated that the Suyuan 590 nearly accommodates all mainstream models, although performance varies significantly across different models; some approaches closely match the A100, while others register about 50% of its training performance, suggesting an overall capability approximating 80% of the A100's level.

From a technical standpoint, AI chips mainly comprise three categories: GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application-Specific Integrated Circuit). NVIDIA's GPU architecture is a well-established, versatile type of AI chip, while the Chinese manufacturers, including Cambricon, primarily utilize ASIC architecture tailored for specific AI requirementsIn 2022, GPUs accounted for a staggering 89% of the global AI chip market, while ASICs represented a mere 1%.

Over the long run, ASIC chips, currently with minimal market share, hold tremendous potential for growth.

Research by CBI indicates that the value of China's AI chip market was estimated at 40 billion yuan in 2021, with forecasts predicting it to soar beyond 230 billion yuan by 2024, signifying a fivefold increase over three years

Such rapid growth undoubtedly presents significant opportunities for mainstream Chinese ASIC chips to accelerate growth and achieve market penetration.

The technological attributes of ASIC chips hint at positive prospects ahead.

Unlike the more universal GPU technology path, the ASIC architecture offers less versatility but can achieve higher processing capabilities in certain specified applications by embedding algorithms directly into the hardwareBecause of their low power consumption and high efficiency, ASIC chips are particularly well-suited for smaller, low-power smart endpoints, such as smartphones, smart security cameras, smart homes, and drones, which may facilitate early-scale rollouts of ASIC technologyAs ASIC chips continue to evolve and as downstream demand changes, they are expected to replace GPU chips in emerging fields, including data centers, autonomous driving, and edge computing.

As a frontrunner in China's AI chip landscape, Cambricon stands as one of the primary beneficiaries of domestic substitution trends

However, to capture a more substantial market segment, it still faces overwhelming pressures.

The primary challenge stems from financial limitations.

At its IPO in 2020, Cambricon raised only 2.582 billion yuan, a minuscule fraction compared to domestic competitor SMIC’s fundraising during the same periodIn its third year post-IPO, when it proposed a capital increase plan at 26.5 yuan per share, it faced regulatory inquiries that ultimately reduced its fundraising target significantlyThe colossal investments required for chip development raise questions about Cambricon's ability to sustain its technological edge while lacking the capability for self-financing.

In this unprecedented epoch of technological tide, investors may exhibit faith and passion, but such sentiments cannot translate into infinite patience for any one company.

As the fervor surrounding thematic investments wanes, the question remains whether Cambricon can extricate itself from the supply chain's grip and swiftly march toward profitability to establish a cycle of self-financing and continuous technological innovation, thereby restoring investors' confidence.