In the evolving landscape of artificial intelligence, platforms like ChatGPT have surged into the public spotlight, drawing an unprecedented volume of users and escalating the demand for computational powerThis influx has prompted OpenAI, the creators of ChatGPT, to temporarily halt new registrations, citing an overwhelming strain on their computing resourcesThe event has not only affected the technology sector but has also rippled through stock markets, particularly in Asia, where shares related to AI computing have quickly become hot commodities.

The circumstances surrounding the demand for ChatGPT exemplify a broader trend within the AI industry: the quest for computational dominanceThose who control the resources necessary for advanced AI functionalities will play a pivotal role in shaping the future of the digital economyThis includes not only the algorithms and data that power AI but the extensive hardware that makes this technology operationalThe sudden halting of subscriptions has highlighted the precarious nature of scalability in AI, especially as companies race to innovate amid fierce competition.

To understand the connection between hardware capabilities and AI’s advancement, we can consider the recent developments at NVIDIA, a leader in graphics processingSix years ago, NVIDIA's CEO, Jensen Huang, delivered OpenAI a supercomputer powered by the A100 chip, which was instrumental in creating ChatGPT’s underlying architectureFast forward to a recent conference in March, where Huang revealed an upgraded model called the HGX A100, boasting ten times the speed and significantly reducing processing costs for large language modelsNVIDIA’s dominant position in the AI chip market has become increasingly evident, as evidenced by an 84% rise in stock value this year, catapulting the company's market capitalization to around $667.8 billion.

The dramatic rise in the stock price of NVIDIA has placed it at a valuation nearly five times greater than that of Intel, showcasing a pivotal shift in the industry where computing power dictates market value

Advertisements

The A100 chips, priced at approximately $15,000 each, represent a substantial investment for businesses looking to commercialize AI technologiesIt has been estimated that processing a GPT-3.5 model, which contains 180 billion parameters, requires around 20,000 such chips, leading to an astronomical expenditure of up to $450 million for companies intending to harness AI's commercial potential.

Looking globally, the implications are significantReports indicate that only a handful of Chinese companies possess access to over 10,000 GPUs, with possibly just one holding a full stock of A100 chipsMost companies in China are relegated to mid-tier GPU models, which limits their capabilities in the burgeoning AI marketThe urgency for enhanced computational power aligns with Huang’s assertion that the “iPhone moment” for AI has arrived, emphasizing the dire need for China to maintain a competitive edge in the global landscape.

The concept of "computational economy," which was first introduced by researcher Zhang Yunqian in 2018, underscores the importance of computing power as a national assetIn recent years, initiatives to boost computational capacity in China have gained traction, most notably with the launch of the "Eastern Data, Western Computing" project aimed at national-wide optimization of computing resourcesAnalysts predict that within a decade, we may see the establishment of a state-backed organization akin to China’s current power grid, which would manage the nation’s computational infrastructure.

A pivotal statistic illustrates this urgent need for AI capabilities: by 2021, China's computational industry reached a staggering 2.6 trillion yuan, generating direct and indirect economic contributions in excess of 10 trillion yuanThis economic driver firmly places AI computation at the forefront of national strategy, leading to an increasingly competitive landscape as companies strive to become China's answer to NVIDIA.

As we venture deeper into the trajectory of AI development, the race for computational supremacy has intensified

Advertisements

The market is witnessing a technological arms race, not just among major tech firms but at a national levelConcerns about supply chain security have intensified following the U.SBureau of Industry and Security’s regulations, which restrict advanced chip exports to China, notably targeting the A100 chips that are critical for AI development.

With the prohibitions on proprietary chips like the A100 and H100, Chinese firms are left to explore alternatives such as the newly developed A800 and H800. Several leading Chinese technology companies have reportedly placed substantial orders for these substitute modelsHowever, these alternatives only serve as diminished versions of the A100, with considerably less computational power, indicating a severe gap that needs to be filled.

As the AI battle intensifies, companies like Baidu are leading the charge in deploying their large language models and other competitors like 360, Alibaba, Huawei, and Tencent are joining the ranksThe influx of these competitors inevitably positions the foundational computational resource companies to reap significant benefitsWithin the domestic market, there are two major categories within the AI sector: AI supercomputing centers and AI chip manufacturers.

The AI supercomputing centers serve as the backbone of the AI infrastructureA prime example is Sugon Information Industry, which has been breaking barriers in high-performance computing since the early 1990s and has provided substantial contributions to China's supercomputing milestonesThe country currently maintains several leading E-class supercomputers that rank among the top globally, continuously utilizing domestic chips to shoulder operations.

Another critical player is the domestic chip manufacturer that stands as the "heart" of computational powerCompanies like Haiguang Information Technology, which operates under the auspices of Sugon, have made remarkable strides in CPU and DCU developmentThey have begun to witness explosive growth since 2021, demonstrating the rapidly rising demand for domestic chips

Advertisements

Advertisements

Advertisements