Last Thursday, US chip giant Broadcom released an impressive earnings report that exceeded market expectations, leading to a significant surge in its stock priceThis positive news reverberated through the global markets, including the A-shares in China, where numerous semiconductor stocks, particularly those related to computing power, witnessed a notable rise in their prices.

In its 2024 fiscal year, Broadcom's revenue from artificial intelligence soared by an astounding 220%, reaching a staggering $12.2 billionOn Friday, the company's stock price jumped by 24.43%, pushing its market capitalization beyond the $1 trillion markA significant portion of this revenue growth came from components used in Ethernet networksFurthermore, Broadcom has partnered with three major cloud providers to develop customized AI chips—these specialized chips, known as Application-Specific Integrated Circuits (ASICs), stand out from conventional Graphics Processing Units (GPUs) when it comes to AI applications.

The reported success of Broadcom has cast a bright light on the A-share market, with many companies in the computing power sector benefiting from the upswing

For instance, during today’s trading session, shares of several notable firms made impressive gainsCambrian, a prominent player in AI chip design, saw its stock rise by 1.57%. Meanwhile, cloud computing and AI solutions provider Yuntian Lifei experienced a surging 9.08% increase in share priceOther gainers included Zhaoyi Innovation, Xin Yuan Co., and Canxin Technology, whose stocks rose by 3.22%, 3.96%, and 1.55% respectivelyAdditionally, Broadcom Integrated's shares soared by 8.84%, while the education firm Kede had a remarkable increase of 20.03%.

Digging deeper into the primary business areas, Cambrian specializes in chip design specifically for the AI sector, offering cloud-based intelligent chips and a range of software and IP licensing servicesYuntian Lifei employs a customized Instruction Set Processing (ASIP) architecture designed for edge AI scenarios and boasts a dual-platform strategy, focusing on both algorithms and chip technology with their neural network processors

• Energy Transition and Climate Change Solutions
• Boosting Culture-Tourism Integration
• ETF Market Boom Drives Public Fund Innovation
• Commodity Price Volatility and the Economy
• Metaverse Finance: Exploring New Frontiers

Zhaoyi Innovation specializes in memory and Microcontroller Units (MCUs), essential components for AI productsFurther, Xin Yuan Coprovides a one-stop chip customization service, including semiconductor IP licensing as its NPU has been integrated into various AI chip productsSimilarly, Broadcom Integrated offers integrated circuit design focused on wireless communication radio-frequency chips and solutions.

In recent discussions regarding advances in artificial intelligence, Xin Yuan Cohighlighted its capabilities in investor engagementsThe company states that, built on its proprietary IP, it offers a platform for customized hardware and software chips tailored for AI applicationsThis encompasses lightweight spatial computing devices such as smartwatches and AR/VR glasses, as well as high-efficiency edge computing devices like AI PCs, AI smartphones, smart cars, and robotsNotably, their integrated NPU IP-powered AI chips have made over 100 million units in shipments

NPU chips are classified as a type of ASIC.

Yuntian Lifei also shared insights in their recent investor relations activity, noting a clear trend shifting AI applications from cloud-based large models to edge devicesTheir Deep Edge series of inference cards have successfully integrated with approximately ten leading models, such as Cloud Tian Shu and Tong Yi Qian Wen.

Meanwhile, Broadcom Integrated announced this October on its investor interaction platform that it has launched multiple IoT chip products that integrate AI technologies, with plans to enhance development investment in AI algorithms further.

However, an analysis reveals that some of the companies involved in this sector initially were not directly related to AI chips but have ventured into AI by investing in or acquiring stakes in ASIC chip firms, which have not yet yielded substantial profitsCompanies like Aibulu, primarily known for environmental management, and Kede Education, involved in educational services, have both acquired stakes in Chip-Hero Technology, an ASIC chip company.

Chip-Hero Technology claims to be the only company in China mastering the core TPU (Tensor Processing Unit) architecture for AI chips

They have developed high-performance TPU chips specifically designed for large AI models and created large-scale AI computational clusters capable of supporting models with over 1 trillion parametersIt is noteworthy that TPU chips also fall under the category of ASICs.

As for investment stakes, Kede Education planned to invest 130 million yuan to increase its holding in Chip-Hero Technology by acquiring more equityFollowing this transaction, Kede would hold an 8.3791% equity stake in the companyIn September of this year, Aibulu reported that its controlling subsidiary invested 250 million yuan in capital to increase its stake in Chip-Hero Technology to 7.6923% after the transaction was completed.

Despite holding shares in Chip-Hero Technology, neither Aibulu nor Kede Education has seen any significant changes in their core business structuresAibulu continues to focus on environmental management engineering and operations, reporting a loss of 33.53 million yuan in the first three quarters of this year, with revenues decreasing by 24.14%. Kede Education, on the other hand, diversified into vocational education and various printing ink products that exhibit high gloss, high wear resistance, and quick-drying properties

Notably, Chip-Hero Technology is not included in Kede Education’s consolidated financial statements, and the investment in Chip-Hero Technology is classified as a long-term equity investment, which resulted in Kede recording a loss of 640,170 yuan based on Chip-Hero's profits in the first half of the year, with revenue from Chip-Hero set to be accounted for later this year.

It is crucial to note that Aibulu's disclosures also indicate that while Chip-Hero Technology is among the few companies in China possessing core technology for TPU's training and inference architecture, the company still faces risks associated with its technological pathwayThe AI computation infrastructure heavily relies on the support of high-performance chipsAlthough TPU and other ASICs are specifically designed for machine learning and AI applications, the ecosystem surrounding these chips lacks the maturity compared to that of GPUs.

Looking at financial results from the first three quarters of this year, while several computing power chip firms reported increased revenues, not all achieved profitability