Following NVIDIA's footsteps, China's "computing power scalpers" are coming on strong

06/28 2024 558

Produced by: Bullet Finance

Art Editor: Qian Qian

Reviewed by: Song Wen

AI is reshaping the entire world. The booming development in related fields has not only ushered in the "intelligent" era for thousands of industries but also opened up new economic growth points. Technology newcomers such as Yuezhi Anmian, Zhipu, MiniMax, and Baichuan Intelligence have achieved remarkable results in their generative AI layout. This harvest is not only reflected in financial statements but also in driving the prosperity of related industrial chains. Coupled with continuous policy benefits, Chinese AI companies have broader development space.

Among the various AI subdivisions, the commercial application of AIGC (Artificial Intelligence Generated Content) is almost at the forefront of the industry, automatically generating text, images, videos, and code, greatly improving the efficiency and quality of content production, reducing costs, and creating possibilities for innovative profit models for enterprises.

On the other hand, as technology matures, AIGC is beginning to empower more complex scenarios, such as multimodal content generation, interactive agents, and even applications in various scenarios related to people's daily lives.

The imagination space for AIGC is expanding. In this magnificent technological revolution, it has attracted batch after batch of entrants, and many listed companies have successively deployed AIGC technology. Among them, Shan Gao Holdings, with its keen industry insight and forward-looking strategic layout, has successfully entered the emerging concept track of "computing power + electricity" by strategically investing nearly $300 million in 21Vianet.

Recently, there have been rumors in the industry about its ambition to deploy AI chips. As soon as the news came out, Shan Gao Holdings, which already had the halo of the AIGC concept, quickly became the focus of attention in the capital market.

1. Laying the foundation for AI infrastructure with computing power

When discussing the cutting-edge field of AIGC, algorithms, computing power, and data are the three essential elements that cannot be bypassed. The former two are like the "double helix" of the intelligent era, while data is like a "gas station." The three rely on each other to jointly promote the vigorous development and widespread application of AIGC technology.

Algorithms are the "engine" of the AIGC high-speed train, determining how fast and far it can run. If AIGC were a magic book, it would not record "spells" but lines of code. These code instructions allow computers to learn to "think" like humans, from text, images to music, videos, and more. And they are far smarter, more capable, and faster than the human brain.

These codes are the core. How the code is designed, whether it is single-threaded or today's deep learning models, depends on the ability of the algorithm owner. Therefore, AI companies are committed to refining their algorithms, and the industry has welcomed one technological revolution after another. For example, last March, Open AI's GPT series and DALL-E models both drove the industry forward with groundbreaking algorithmic innovations.

Then there's computing power. With a robust brain, you also need a "body" that can execute it. Otherwise, algorithms will be like water without a source. In recent years, with the rapid development of graphics processors and dedicated AI chips, the improvement of computing power has made it possible to train larger and more complex multimodal large language models. Companies like NVIDIA's TPU and Huawei's Ascend series chips are contributing to reducing training costs and accelerating model iteration times.

As one of the three essential elements of AI, data is even more difficult to obtain than algorithms and computing power. Without continuous high-quality data "feeding," the commercial application of artificial intelligence will also become empty talk.

However, most players choose to start with algorithms and computing power. After all, compared to data, which requires industry accumulation and time, these two investments can directly complete the layout with "money power."

Therefore, after the industry made breakthroughs in AIGC last year, laying out computing power, seeking better learning efficiency, and targeting the commercial needs of target users has become the first breakthrough point for players to focus on.

In grasping this trend, Shan Gao Holdings is quite representative. As an important domestic investment, financing, and emerging industry holding platform under Shandong Expressway Group, its series of mergers and acquisitions since the second half of 2021 have demonstrated its strategic goals of focusing on new energy and new infrastructure. In 2022, Shan Gao Holdings merged with the Hong Kong-listed company North Control Clean Energy Group (now renamed "Shan Gao New Energy Group") and entered the new energy industry with a high starting point. In 2023, Shan Gao Holdings keenly captured the new trend of AI and quickly acted, completing a strategic investment by the end of that year and becoming the largest institutional shareholder of 21Vianet.

As one of the largest independent third-party data center operators in China, 21Vianet, founded in 1996, has nearly 30 years of development history, mainly providing hosting network interconnection services such as data center services, cloud services, hybrid IT services, and CDN services. Public information shows that 21Vianet currently operates over 50 data centers in more than 30 cities across the country, with over 87,300 cabinets serving more than 7,000 enterprise customers.

After Shan Gao Holdings' investment, 21Vianet has also ushered in a new development situation, with not only financial conditions being alleviated but also business expansion entering a new phase. In the first quarter of this year, 21Vianet's net revenue from data center business was 1.28 billion yuan, an increase of 5.2% compared to the same period in 2023. According to a report from Qianzhan.com in mid-June, 21Vianet recently collaborated with NVIDIA to build a new generation of data centers focused on artificial intelligence.

It is understood that this data center relies on 21Vianet's years of deep cultivation in the data center field, integrating NVIDIA's cutting-edge technology and ecological advantages in the field of artificial intelligence, jointly breaking through the limitations of traditional data center infrastructure, deploying an accelerated computing platform tailored to the needs of the AI industry, and building a brand-new efficient, intelligent, green, and sustainable computing power ecosystem.

Transitioning from traditional IDC to AIDC requires a high threshold, involving more complex technical and operational challenges. For example, traditional IDC service providers are similar to developers providing rough housing, a standardized process that involves demand research, construction, shelving, operation, and maintenance. However, AIDC requires more capabilities, similar to providing furnished apartment-style services, balancing factors such as PUE (data center energy efficiency indicator) and computing power cluster utilization, significantly raising the operational threshold.

As a result, in just over two years, Shan Gao Holdings has completed rapid layouts in the fields of green power and computing power through controlling mergers and acquisitions and strategic investments. From the secondary market perspective, both green power and computing power are hot tracks. Especially after completing the computing power layout, the AIGC concept instantly ignited the capital market.

In the view of "Interface News · Bullet Finance," Shan Gao Holdings' move not only aligns with the country's general trend of building a nationwide integrated computing power network but also gradually constructs a comprehensive system encompassing computing power supply, green energy integration, and even core hardware technology for itself. From this perspective, the long-term value in the AIGC field cannot be underestimated.

2. Laying out chips silently and digging deep into the moat of computing power

Since AIGC has broad prospects and computing power is crucial for success, how should players lay out?

In fact, the computing power layout is more like a light show, requiring the placement of sufficiently powerful lighthouses (data centers) at key positions and ensuring efficient transmission channels (network connections) between these lighthouses. In the entire link, chips are like high-performance bulbs in lighthouses, crucial components for demonstrating computing power.

Powerful chips can allow computers to understand and execute complex tasks faster, such as the current hot AI large model training. At the same time, chips not only require high performance but also consider energy consumption, which can significantly reduce operating costs.

In addition, with the diversification of application scenarios, general-purpose chips can no longer meet all needs, and customized and specialized chip designs have emerged. Furthermore, modern computing systems often adopt heterogeneous computing architectures, where different types of chips work together to achieve optimal computing effects. This requires chips to not only have strong individual performance but also good interoperability and low-latency communication capabilities. At the same time, there are also higher requirements in terms of security, reliability, and ecosystem construction.

Therefore, chips play a central role in the computing power layout. This is also one of the reasons why technology giants such as NVIDIA and AMD have been continuously optimistic about them in recent years. The goals of these international technology giants in continuously introducing innovative chips are clear: to enhance AI processing capabilities, thereby consolidating and expanding their leadership positions in the market, and also reflecting the global urgent need to enhance artificial intelligence processing capabilities.

Powerful companies like Meta, Microsoft, and Google have invested heavily in chips. Meta's annual report revealed that capital expenditures in 2024 may reach up to $40 billion; Microsoft's capital expenditures are $50 billion, of which R&D expenditures are $27.2 billion; Amazon is $14 billion. And the flow of these capital all points to GPU chips.

In China, major manufacturers are also paying increasing attention to chips. In the past few years, people still vividly remember being "strangled" in the chip field. This passive situation has also made the industry realize the importance of achieving independent controllability in key technology areas. Excessive reliance on imported high-end chips can not only lead to supply chain instability but also affect national security and industrial development. For example, in the automotive field, chip shortages have significantly impacted the production and technological innovation of domestic cars.

To address this challenge, China has been vigorously promoting the independent development of the chip industry, such as increasing R&D investment, cultivating professional talent, encouraging technological innovation, and supporting domestic substitution. These measures aim to fundamentally solve the "strangling" problem and enhance the country's overall technological strength and economic competitiveness.

In recent years, some domestic companies have followed this general trend in niche tracks, focusing their attention on more fundamental chips. In this process, a group of representative players have emerged.

Returning to Shan Gao Holdings, if it wants to seize more opportunities in the AIGC era, in addition to obtaining computing power bases through mergers and acquisitions, it is more likely to focus on the upstream of the computing power supply chain.

The recent investment layout of 21Vianet also indirectly confirms this trend. In May, 21Vianet led a multi-million dollar Pre-A round of financing for the computing power scheduling platform MatrixOrigin, further improving its layout of a full-chain artificial intelligence ecosystem. It is disclosed that 21Vianet will continue to innovate around GPU infrastructure and cooperate deeply with MatrixOrigin on the neolink.ai platform to create a Chinese version of CoreWeave.

Looking back at CoreWeave's development journey: Since 2019, the company has focused on purchasing enterprise-grade GPU chipsets, building dedicated cloud infrastructure, and adjusting its business around NVIDIA chips. Riding on the wave of AI development, CoreWeave has become a cloud computing service company relying on NVIDIA graphics cards, successively winning major customers such as Microsoft and Google, entering the unicorn camp, and becoming one of the world's largest independent GPU cloud providers, serving customers in various industries such as VFX (visual effects), artificial intelligence, gaming, and healthcare.

Currently, CoreWeave is colloquially known as a "computing power scalper" in the industry, holding tens of thousands of NVIDIA AI graphics cards and successfully completing $2.3 billion in debt financing using NVIDIA chips as collateral, with its valuation soaring to $8 billion in four years. It is clear that the phenomenon space of the Chinese version of CoreWeave is considerable.

Comprehensive with Shan Gao Holdings' actions since its transformation, it can be seen that Shan Gao Holdings' intention to lay out the chip field, especially chip applications related to new energy and data centers, is built on the basis of industrial ecological collaboration. Rather than saying that Shan Gao Holdings is chasing hot tracks, it is more accurate to say that it is building an ecological industrial chain with endogenous power.

Although there is limited direct news about Shan Gao Holdings' layout in chip manufacturing, considering its emphasis on computing power and new energy, as well as its close cooperation with data centers, it can be speculated that its layout may focus more on the application level, such as investing or co-developing high-performance computing chips suitable for data centers, management chips for optimizing computing power usage, or smart chips combining new energy technologies.

If this move is true, it means that Shan Gao Holdings is further extending upstream in the AIGC industry chain, deepening its technological moat by mastering core computing power technology. After all, in the AIGC era driven by both algorithms and computing power, having independent chip technology will undoubtedly bring stronger competitiveness and broader market space to enterprises.

3. Highlighting long-term value and enjoying the dividends of the AIGC era

Through forward-looking layouts in the "computing power + green power" ecosystem and potential exploration in the chip technology field, Shan Gao Holdings has demonstrated a profound insight into and active response to technological development trends. It is not difficult to see that Shan Gao Holdings is gradually transforming from a traditional investment holding company into a technology investment group involved in new energy, computing power infrastructure, and even potential AI chip fields.

Currently, the AI field is booming, with numerous concepts and applications emerging from horizontal general-purpose products to vertically focused specific applications. However, the computing power base remains the key to winning in the AI race.

In April this year, Google and Meta successively launched their latest versions of self-developed chips, with customization as a major focus. Meta, as a social media giant, released the latest version of MTIA specifically designed for the ranking and recommendation system of Meta's social software. Google announced that it is manufacturing a custom CPU based on the ARM architecture called "Axion," which is planned to support services such as YouTube ads on Google Cloud.

The intention behind this move is not difficult to guess. In recent years, the development of new-generation AI has driven a surge in chip demand, coupled with the rapid development of AI applications in various fields, pushing competition in the chip industry to become increasingly fierce. Domestic and foreign technology giants have bet on AI self-developed chips, hoping to seize development opportunities.

For Shan Gao Holdings, if it can make substantial progress in chip technology, whether through independent R&D or strategic investment cooperation, it will greatly enhance its competitiveness in AIGC and related fields.

It is important to note that AIGC is not just a concept but a practical technology that needs to be combined with specific industries and application scenarios. Institutional views also point out that in terms of AI investment opportunities, the core clue lies in profit release before the emergence of blockbuster products or applications, and profit release usually starts from the upstream first according to industrial development laws.

Judging from Shan Gao Holdings' dual layout of "new energy + new infrastructure," it is committed to building a complete industrial chain from energy to computing power. With the combined forces of chip investment and the original industrial layout, both in terms of product and application, it will help Shan Gao Holdings form a unique green competitive advantage in the future, attracting more customers seeking environmentally friendly computing power solutions.

According to the latest forecast from market research firm Gartner, the AI chip market size is expected to grow by 25.6% year-on-year to $67.1 billion in 2024, and it is expected to be more than twice the market size in 2023, reaching $119.4 billion by 2027.

NVIDIA, which dominates 70% of the market share, achieved revenue of $26 billion in the first quarter of this year, a year-on-year increase of 262%, far exceeding the market expectation of $24.7 billion. Among them, the revenue of its core computing power business - data centers - was $22.6 billion, a year-on-year increase of 427%.

Undoubtedly, the AI chip market has broad prospects, but competition is also becoming increasingly fierce, and the "computing power war" has quietly begun. It remains to be seen whether Shan Gao Holdings can seize the advantage of latecomers and bring new imagination to the market.

It is foreseeable that considering that chips are the core of computing power, any investment in this field is likely to transform into independent R&D capabilities for key technologies. Shan Gao Holdings' move will also help it reduce dependence on external suppliers, improve supply chain security,

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.