The capabilities and promise of artificial intelligence (AI) are driving many countries to invest in hardware to keep up with global competition. Choe Sang-mu, acting president, said on the 17 Feb that the country would expand the supply of 18,000 high-performance GPU graphics processing unit by the first half of next year. At the end of last month, India announced it would provide more than 18,000 gpus to build AI computing facilities for local start-ups and development researchers. Experts interviewed by the Global Times said that the current competition requires countries to have independent, controllable and diversified supply of hardware, as well as to improve the level of coordination between hardware and software.
10,000 gpus by the end of the year
According to a The Hankyoreh report on February 17, Choi Sang-mu chaired a meeting of the“AI Computing Infrastructure Special Committee” on the same day, to have a detailed understanding of the ongoing construction project of the National AI computing center. To boost South Korea’s AI competitiveness, it will expand the processing power of 18,000 high-performance gpus by the first half of next year to support businesses and research institutions, he said. Choe also stressed that the government will give full support in the future to become a true“Three AI powers”. Through a public-private partnership, the government will ensure 10,000 high-performance gpus this year, opening a“National AI computing centre” as soon as possible. The 8,000-GPU equivalent of supercomputer 6 will be built in the first half of next year, with a focus on supporting research institutions.
The South Korean government will ensure the supply of 10,000 high-performance gpus this year through a public-private partnership. Employees at Samsung headquarters in South Korea work in a data center. (visual China)
The South Korean government has not yet decided which GPU products to buy, Reuters reported, citing an official at South Korea’s Ministry of Science, Technology and Information Communications, but details of the budget, the GPU model and the private companies involved will be finalised by September.
South Korea currently ranks sixth in the global AI rankings, performing well in terms of development and infrastructure, the Asia daily reported Monday, however, it lags behind in personnel training, research investment and venture capital. This year, the South Korean government increased its annual AI technology budget by 25 The Korea Herald cent year-on-year, allocating 1.8 trillion won (9.11 billion yuan) to its AI program, Gaijin reported. In addition, the government plans to launch an interagency program aimed at strengthening the country’s artificial intelligence capabilities.
Key hardware for multinational competition
The number of gpus needed for the AI model depends on how advanced the gpus are, the amount of data used to train the model, and the size of the model itself, according to Reuters, and how much time developers want to spend training large models. Nvidia, the US chipmaker, has about 80 per cent of the global GPU market, ahead of rivals Intel and AMD.
As well as South Korea, countries such as India and Saudi Arabia are buying up critical hardware, given the importance of the GPU to AI development. According to a The Hindu report, Indian Minister of Railways, communications and Electronics and information technology, Ashwini Vishnau, said at the end of last month that his country wants to develop a large indigenous AI model, and it will build a large AI computing facility with more than 18,000 gpus. The country announced a $1.2 bn“Indiaai Mission” in March last year, almost half of which will be spent on building computing infrastructure. Moreover, thousands of high-performance Nvidia chips are 2023 for purchase. According to public reports, at least 3,000 Nvidia H100 chips had been purchased.
Some analysts believe that countries that want to develop AI choose to“Stock up” on high-end chips such as GPU, partly out of concern about future difficulties in obtaining these critical hardware. In January, before leaving office, the Biden administration introduced new restrictions on exports of American AI chips and technology. The US divides countries into three tiers, keeps advanced computing power in the US and its allies and looks for ways to block exports to some countries, Radio France Internationale reported. The The Hindu said the US had restricted Indian imports of its GPU. Saudi officials told CNBC in April that they hoped to get Nvidia’s high-performance chips“Within next year”.
Competition among national innovation ecosystems
In his speech, Choi Sang-mu said that competition for dominance in the AI industry had recently become increasingly fierce, with competition shifting from a duel between companies to a state-led“Competition in the innovation ecosystem”. But“What DeepSeek shows us is that even with capital and infrastructure deficiencies, there are performance opportunities that can be surpassed.”
Huang Zhengliang, director of the new industrialization research office of the Industrial Economics Research Institute at the Center for International Economic and technological cooperation of the Ministry of Industry and Information Technology, told the global times on the 18th that competition in the artificial intelligence innovation ecosystem covers many fields, the hardware includes chips, precision instruments and high-performance equipment; the software includes infrastructure supply, talent training and technology transfer efficiency; and the environment, including the system construction, the financing channel as well as the supply demand docking. He said that the competition in the innovation ecosystem has put forward new requirements for the construction and development of hardware industry chains in various countries, “Critical hardware needs to be self-contained and controllable, non-critical hardware needs to have a diversified supply, and hardware and software needs to be more collaborative.”
Liu Shaoshan, director of the Embodied Intelligence Center at the Shenzhen Institute of Artificial Intelligence and robotics, told the global times that countries are no longer seeking the most cost-effective solutions for the sake of AI supply chain security, need to build their own solutions, so more market opportunities. He further explained that in order for a country to gain an advantage in the competition of the innovation ecosystem, on the one hand, the market must be large enough to absorb the innovation, and on the other hand, the innovation capacity must be strong enough to achieve supply.