On February 7, the DeepSeek concept continued to surge, driving up related sectors focused on computing power. Among AI stocks, Inspur Information hit its daily limit, with Anheng Information and Meiri Hudong also experiencing significant gains.

According to institutional data, China's computing power industry, centered around computers, reached a scale of 2.6 trillion yuan in 2021, directly and indirectly driving economic outputs valued at 2.2 trillion yuan and 8.2 trillion yuan, respectively.

CPUs ushered in the PC era, mobile chips sparked the mobile internet wave, and AI chips have shattered the decades-long bottleneck in computing power within the AI industry.

Today, the computing power economy has emerged, and the race for AI computing power has been ongoing for some time.

The A-share market is no stranger to elite players in this field, which can be broadly divided into two categories.

First, AI supercomputing centers. These can be considered the core infrastructure of AI, with Sugon leading the pack.

In 1993, Sugon broke through international barriers to develop the high-performance computer "Dawn 1"; its self-developed Dawn series, from Dawn 1000 to Dawn 6000, helped China's supercomputing achieve multiple milestones.

Currently, China operates three exascale supercomputing prototypes: the Sunway E-scale prototype, the Tianhe E-scale prototype, and the Sugon E-scale prototype, supported by the Jiangnan Institute of Computing Technology, the National University of Defense Technology, and Sugon, respectively.

Some analysts believe that, based on the latest computing performance, China's three supercomputers rank among the top three globally. Furthermore, following the US chip embargo in 2015, all three computers are equipped with "Chinese chips".

Among the three supercomputers, Sugon is the only listed company and the most commercialized enterprise. It is therefore not an exaggeration to say that Sugon is the most genuine computing power stock in the A-share market.

As leading internet and technology companies roll out large language models, domestic demand for computing resources is set to explode. Cloud-based computing power rental will become the optimal solution for computing power supply, with NVIDIA offering a related service starting at $36,999 per month.

Sugon has previously deployed over 50 cloud computing data centers nationwide, controlling the largest computing power resources in the country. In 2023, it officially launched the "Computing Power Trading Platform," allowing third parties to access its computing power resource pool. Additionally, the company has assisted in the training of large models such as Baidu's ERNIE Bot and has provided AI computing power services, AI servers, and other infrastructure.

Second, AI chips. These can be considered the heart of computing power and the "spark" igniting the long march of AI.

Hygon, backed by Sugon, is a national enterprise providing momentum for computing power resources with "Chinese chips".

Sugon is the largest shareholder of Hygon, which is supported by the Chinese Academy of Sciences. Hygon boasts deep research and development capabilities. With Sugon as its major shareholder and customer, Hygon does not need to worry about sales.

Starting in 2021, the company's performance has shown explosive growth. According to its performance report, in 2022, Hygon's revenue was 5.125 billion yuan, an increase of over 1.2 times year-on-year; net profit attributable to shareholders was 802 million yuan, a year-on-year increase of 145.18%.

In the future, the company's CPU products will benefit from domestic substitutions and continue to receive support from government, financial, and other corporate procurements. Meanwhile, the DCU industry, fueled by the explosion of the AI industry, will become a rapidly rising new growth point.

Moreover, Hygon holds AMD's x86 architecture technology license, making it the only domestic enterprise producing x86 chips. Based on the mature x86 system, compared to enterprises like Loongson, Shenwei, and Phytium, Hygon can reduce customer switching costs and enjoys ecological advantages.

Hygon is a scarce leader in the domestic CPU+DCU (accelerated computing chip) industry within the A-share market. Generally speaking, computing power can be divided into three main parts: basic computing power, intelligent computing power, and supercomputing power. What AI large models need is intelligent computing power.

As a product focused on general-purpose computing that solely provides AI computing power, Hygon has already made significant gains. Currently, Hygon's DCU DeepCompute 1 is the only one in China that can support FP64 double-precision floating-point operations.

Compared to NVIDIA's A100 product, Hygon's DeepCompute 1 performs at approximately 70% of its level. According to information on the investor interaction platform, Hygon's DeepCompute 2 is in the research and development stage, with all work progressing smoothly.

Another company with "hard power" is Cambricon, the first AI chip stock on the STAR Market. The company is one of the few globally that fully masters AI chip technology.

Cambricon's product matrix includes cloud products, edge products, IP licensing, and software. The first two are comparable to NVIDIA and AMD, while IP licensing corresponds to the UK's ARM.

Cloud chips are currently the most sought-after source of kinetic energy for large language models. In this field, NVIDIA's A100/H100 series occupy the top positions.

As a competitor, Cambricon has launched the Thinker series, with Thinker 290 and Thinker 370 having entered multiple leading customers such as Inspur, Lenovo, and Alibaba Cloud. Since the launch of Thinker 220, cumulative sales have exceeded one million units.

In March 2022, Cambricon officially released the new training accelerator card "MLU370-X8," equipped with Thinker 370, primarily for AI training tasks.

The Thinker 590 chip is undergoing rapid iteration, with a significant improvement in floating-point computing power compared to the 290. The company's cloud product performance is expected to further improve.

However, if its AI chips fail to achieve a true performance explosion in the future, after a surge in share price, there will be risks of a sharp decline.

As the industry's "water seller" and infrastructure provider, underlying computing power companies undoubtedly benefit greatly. In this journey of AI, the picture of Chinese computing power enterprises competing in myriad ways has only just begun.

Disclaimer

This article involves content related to listed companies, which is the author's personal analysis and judgment based on information disclosed by listed companies in accordance with their legal obligations (including but not limited to temporary announcements, periodic reports, and official interaction platforms, etc.). The information or opinions in the article do not constitute any investment or other business advice. Market Value Watch assumes no responsibility for any actions taken as a result of adopting this article.

——END——