South Korean Deputy Prime Minister: Artificial Intelligence Should Be Viewed as Basic Infrastructure Like Electricity, Affordable for Every Citizen Without Cost Concerns
South Korean Deputy Prime Minister and Minister of Science and ICT, Lee Kyung-hoo, recently stated that artificial intelligence should be regarded as basic infrastructure like electricity, and every citizen should be able to use it affordably without worrying about costs. President Lee Jae-myung previously mentioned that this technology has brought South Korea to a significant historical turning point, questioning whether to continue as a potential follower or to become a leader seizing infinite opportunities.
To address this question, the government selected three consortiums led by SK Telecom, KT, and Kakao last month to implement the "Universal AI" plan, providing approximately 51.6 million citizens with free, unlimited access to general chatbots and personal AI assistants, which are expected to handle government and daily tasks. The Ministry of Science selected these three from six bidders, planning to sign contracts within the year, conduct tests in the fall, and officially launch by the end of the year.
This year, the state will provide a total of 512 NVIDIA B200 GPUs to the three companies, subsidizing national operating costs starting in 2027; at least half of the inference must utilize domestic foundational models. SK Telecom promotes its self-developed A.X K2 (approximately 688 billion parameters) and will offer applications, web, phone, and SMS services; KT will integrate telecommunications and IPTV into the same dialogue; Kakao will use KakaoTalk as an entry point, adding features for close friends, finance, and government agency representation. Foreign models can only be used when deemed "minimally necessary" by the ministry and will not receive subsidies.
Lee Kyung-hoo wrote on social media that despite concerns about catastrophic consequences from uncontrolled technology, South Korea must become "one of the top three countries in AI": "There is currently no room to slow down development. We will accelerate our pace, but will do so more safely." Lee Jae-myung outlined three paths: an irreplaceable supply chain for storage chips, faster nationwide rollout than other countries, and distributing AI benefits to every resident. About two-thirds of South Koreans have tried AI, with tens of millions paying for ChatGPT, and the government aims to transition some of these users to a domestic free tier.
In market mechanisms, the buyers are all residents exempt from subscription fees, while the sellers are the three telecom and internet consortiums receiving national GPUs and financial subsidies. Funds will flow from the budget and chip quotas to domestic model inference rather than to foreign subscriptions like OpenAI. Beneficiaries include SK Telecom, KT, Kakao, and their model partners, while foreign chatbots that have established paid habits will be under pressure, as will the technical route requiring more than half of the models to be domestic to receive subsidies. The event is driven by positioning AI as a public good at the level of electricity, rather than a price war among individual companies.
Source: Public Information
ABAB AI Insight
After Lee Jae-myung made AI a national policy issue, Lee Kyung-hoo is responsible for realizing the slogan of "one assistant per person." South Korea is not aiming to create a GPT-level closed-source model first, but instead is leveraging telecom entry points, KakaoTalk daily active users, and financial subsidies to increase the inference volume of domestic models. The 512 B200 GPUs are relatively sparse compared to the national population; the real leverage is the requirement that over 50% must use sovereign models, turning daily conversations into training and deployment data weighted towards domestic models.
The capital path is fiscal spending to buy traffic. GPUs will be provided this year, with operating costs entering the budget starting in 2027, allowing companies to exchange existing pipelines for national orders. Resources will shift from residents' ChatGPT monthly fees to the inference bills of SK, KT, and Kakao. The motivation is to reduce dependence on the sudden shutdown of the two major platforms in the U.S. and China, while also creating a new application layer narrative beyond the storage provided by Samsung and SK Hynix. The strategy is to first expand usage, then discuss exporting the "public AI" template to similarly anxious middle-income countries.
Similar cases include France promoting domestic assistants, the UAE treating large models as national projects, and China integrating large models into government services. The industry is in a period of expanding sovereign AI: chips and electricity remain constrained, and control has been incorporated into the Ministry of Science's model allocation terms, yet the transition from free chat to auditable government representation has not been completed.
Structural judgment pertains to the transfer of pricing power. The pricing power of personal AI shifts from a $20 subscription to national budgets and domestic model quotas; the implication of the electricity metaphor is that usage will no longer be limited by laboratory constraints, but by fiscal and GPU inventory limits. The mechanism is: when entry points are locked by telecommunications and national-level chat software, the free tier will first consume paid habits, then force foreign models to become "minimally necessary" supplements in South Korea.
ABAB News · Cognitive Laws
Technologies written as electricity transfer the charging rights to the state.
The true quota for free assistants is that domestic models must account for half.
Nations fearful of shutdowns will first grant entry to their own telecommunications.