South Korea is making a push to strengthen its AI infrastructure by acquiring 10,000 high-performance graphics processing units (GPUs) this year. Revealed by Acting President Choi Sang-mok, the initiative is part of a broader strategy to strengthen national AI infrastructure. Choi stated that AI competition is evolving beyond company-level rivalry, turning into a contest between national innovation ecosystems. The South Korean government is working with the private sector to secure the necessary GPUs, with the goal of launching services at the national AI computing centre earlier than originally planned.
GPUs play a critical role in AI model development, with requirements depending on processing power, dataset size, model complexity, and training duration. South Korea has not yet determined which specific GPU models it intends to procure, but according to the Ministry of Science and ICT, details such as budget allocation and private-sector partners will be finalised by September.
The plan follows recent regulatory changes by the US, which last month imposed new restrictions on AI chip exports. The measures categorise countries into different tiers, with South Korea among 18 nations exempt from the limitations. Around 120 other countries face restrictions, while China, Iran, and Russia are completely barred from acquiring advanced AI chips.
The AI sector has seen an increased demand for GPUs, with Nvidia holding an estimated 80% share of the global market, well ahead of competitors Intel and AMD. Nvidia’s chips are widely used for AI applications, including generative models and high-performance computing. Meanwhile, AI firms are working to diversify their hardware supply. Microsoft-backed OpenAI, which relies on GPUs to train its AI systems such as ChatGPT, is developing its own AI chip. The company is expected to finalise its first in-house semiconductor design in the coming months and plans to manufacture it through Taiwan Semiconductor Manufacturing. Alternative approaches to AI chip development are also emerging. Chinese startup DeepSeek is focusing on computational efficiency rather than relying on high-powered hardware, aiming to close the performance gap between Chinese AI processors and US-made counterparts.
AI investments expand in Europe and the US
Last week, France announced a €109bn investment to expand its AI sector. The funding, unveiled by President Emmanuel Macron at the Paris AI Summit, includes a €20bn commitment from Canadian investment firm Brookfield and potential funding of up to €50bn from the UAE. More than 60 European companies, including Airbus, ASML, Siemens, and Deutsche Bank, have also launched the EU AI Champions Initiative. The programme, introduced at the AI Action Summit in Paris, aims to accelerate AI development across Europe. Discussions at the summit included European Commission President Ursula von der Leyen and heads of state from 17 European nations. The initiative consists of a €150bn investment commitment from over 20 international investors, such as Blackstone, KKR, EQT, and Warburg Pincus, with funding allocated for AI projects in sectors like healthcare, energy, defence, and manufacturing over the next five years.
In the US, a $500bn private-sector AI infrastructure project, known as Stargate, was announced last month. The initiative, backed by OpenAI, SoftBank, and Oracle, represents a significant investment in AI technology development. The project follows the repeal of a 2023 executive order issued under the previous administration, which sought to regulate AI-related national security risks. Companies involved in Stargate include Microsoft, Nvidia, and Arm, which will work on expanding AI data centre capabilities.
Read more: Trump administration unveils $500bn Stargate initiative to dominate AI infrastructure
