The rapid advancement of artificial intelligence (AI) technology has sparked an unprecedented growth in the compute power industry, with significant implications not only for AI development but also for the economy at large. On February 7, the stocks associated with the DeepSeek concept experienced remarkable gains, reflecting a burgeoning enthusiasm surrounding this space. Companies such as Inspur Information, which reached the daily trading limit, and Anheng Information saw significant surges in stock prices, indicative of a larger trend within the market.

Recent studies reveal that by 2021, the compute power industry in China, spearheaded by advancements in computing technology, had reached an impressive scale of 2.6 trillion yuan. This ecosystem has had a substantial impact on China's overall economic output, accounting for a direct contribution of 2.2 trillion yuan and an indirect contribution of 8.2 trillion yuan to the nation’s GDP. The significance of compute power can be interpreted through historical milestones: the Central Processing Unit (CPU) ushered humanity into the era of personal computers, mobile chips spurred the mobile internet revolution, and now AI chips are breaking decades-long barriers to AI advancement.

As we delve deeper into the compute power landscape, we witness a fierce arms race focused on AI capabilities. Within the A-share market, agile players have emerged, categorized primarily into two groups. The first of these categories encompasses AI supercomputing centers, which are vital infrastructures supporting AI applications and research. Among the leaders in this sphere is Sugon, a company that has played a pivotal role in establishing China’s supercomputing capabilities.

Sugon's journey dates back to 1993 when it defied international trade restrictions to develop the high-performance computer known as the "Dawn One." Subsequent innovations, particularly the Sugon 1000-6000 series, have culminated in significant accolades for China in the realm of supercomputing. Today, China boasts three prototype supercomputers capable of quintillion operations per second (E-class): the Sunway E-class prototype, the Tianhe E-class prototype, and the Sugon E-class prototype. Each of these is backed by leading research institutions in China, showing the robust support for the nation’s technological ambitions.

Experts suggest that based on the latest computing performance metrics, these three supercomputers could rank within the global top three. Notably, all three utilize "Chinese chips," a significant transition following the U.S. chip embargo in 2015. Among them, Sugon stands out as the only publicly traded firm, showcasing the highest level of commercialization within the field. This has marked Sugon as a bona fide compute power stock in the A-share market.

As major tech and internet companies roll out large language models, there’s an impending explosion in compute resource demand in China. Cloud computing solutions for compute leasing may become the most viable option for meeting this demand. Companies like NVIDIA have launched related services, with monthly rental prices starting at $36,999, illustrating the extent to which businesses are willing to invest in top-tier compute resources.

Sugon has established over 50 cloud computing data centers across China, holding the largest repository of compute resources nationwide. By 2023, it is set to launch a “Compute Resource Trading Platform,” allowing third parties to tap into its compute resource pool. Notably, Sugon has provided AI compute services and infrastructure support for substantial model training for key players like Baidu and Zizhong Taihu, further solidifying its position at the forefront of the AI computing sector.

The second focal group within the compute industry consists of AI chips, which can be construed as the vital “heart” of compute power. These chips serve as the driving force behind AI developments. Hygon Information, a key entity behind Sugon, plays a crucial role in providing the necessary computational vitality essential for this dynamic market. With Sugon as its largest shareholder and with rich R&D backing from the Chinese Academy of Sciences, Hygon has demonstrated substantial growth, with its revenue surging by over 120% year on year in 2022.

The future is promising for Hygon, particularly as the demand for domestic CPUs increases in governmental and financial sectors. Their production of DCU technology, essential for AI advancements, positions them as potential leaders in the industry. Notably, Hygon has exclusive rights to produce x86 architecture chips in China, enabling it to offer an attractive alternative to competitors like Longxin and Shenwei by minimizing client transition costs.

As the nation continues to evolve its capabilities, Hygon emerges as a rare force within the A-share market, constituting a leading player in the domestic CPU and DCU sector, particularly concerning the smart computational requirements vital for AI large model development. Currently leading the field, Hygon’s DCU technology, which supports FP64 double-precision floating-point operations, is unmatched within China.

This year, Hygon unveiled its high-performance DCU, which is akin to NVIDIA’s A100 product, demonstrating around 70% of its performance capabilities. The next-generation chip, Hygon's DSI-2, remains under research and development with promising prospects. The firm is rapidly transitioning into a company that can facilitate AI training applications, adding a competitive edge and addressing a crucial need in the market.

Another emerging competitor is Cambricon, the first publicly-listed AI chip company in China. Their range of products caters to various sectors, including cloud computing and edge computing, and even offers IP licensing, directly competing against global giants like NVIDIA and AMD. Their cloud-based chips are particularly tailored to meet the surging demands of large language models, where companies like NVIDIA currently dominate with their A100 and H100 series.

As a challenger, Cambricon has introduced its Siyuan series chips, which have gained traction among leading players, including Wave and Lenovo. After launching Siyuan 220, it has already surpassed one million units sold, indicating a substantial uptake in its technology.

In 2022, Cambricon made headlines with the launch of the MLU370-X8 accelerator card aimed at boosting AI training tasks, showcasing its ongoing commitment to this booming field. With rapid iterations underway for the Siyuan 590 chips, Cambricon stands poised for an enhanced performance trajectory that could solidify its stake in the industry significantly.

However, like any rapidly evolving sector, risks of volatility exist. If there aren't tangible breakthroughs in profitability for their AI chips, stock prices that have surged could subsequently plummet. The essential backdrop is that companies providing foundational compute resources are likely to benefit greatly in this AI-driven journey, reflecting a fierce competitive landscape that only just began to unfold within China's compute power domain.