The focus of government-initiated artificial intelligence (AI) projects is shifting from construction of buildings and complexes to operational projects that build services on existing infrastructure. According to the Public Procurement Service, the budget for AI-related projects issued in construction form has decreased by over 90% in the past year, while the scale of national projects has grown to dominate the overall procurement size.
As of August 24, 2026, the total number of construction projects related to 'artificial intelligence' issued through the Public Procurement Service was five, amounting to approximately 500 million won. This represents a 94% decrease in budget compared to the same period last year, which saw nine projects totaling 7.88 billion won.
Last year, a single project for relocating the AI Education Center of the Gyeongsangbuk-do Office of Education alone received 5.27 billion won, and when including associated mechanical, electrical, information communication, and fire safety works, the total expenditure approached 6 billion won. This year, large-scale new construction projects have disappeared, leaving only small ancillary works such as the installation of electric meters at the AI complex, which cost 270 million won.
Services and platform projects have filled the void. New initiatives have emerged that operate administrative services on already established infrastructure, such as the 6.6 billion won government-wide AI common infrastructure project and the 1.8 billion won AI agent-based review support system project, both commissioned by the Korea Intelligent Information Society Agency.
Local governments, which previously limited AI basic plan development to regions like Gyeonggi Province and Gyeongsangnam-do, have expanded this initiative nationwide, including cities like Seoul, Busan, Suncheon, Yeosu, Chuncheon, Goryeong, Geumsan, and Pohang. This shift indicates a transition from the infrastructure-building phase to actual utilization.
Despite the growth in the procurement market, small and medium-sized IT companies continue to face tough realities. Most projects are concentrated among large enterprise consortiums.
From January 1 to August 24, 2026, the total amount of projects issued under the keyword 'artificial intelligence' through the Public Procurement Service reached 93.9 billion won, a 50.6% increase from 63.4 billion won in the same period last year. However, when combined with the budgets of three major projects awarded to a few large enterprise consortiums, this amount approaches 30 times the total procurement amount.
The largest disparity is seen in the National AI Computing Center. This project, awarded to the Samsung SDS consortium through a sole bid, has a total project cost exceeding 2.5 trillion won. Initial funding of 116 billion won has been allocated from the Ministry of Science and ICT and a national policy bank. The initial budget for this single project alone surpasses the total budget for AI projects from the Public Procurement Service over eight months.
The consortium includes three Samsung affiliates—Samsung Electronics, Samsung C&T, and Samsung SDS—along with Naver Cloud, Kakao, and KT. The independent AI foundation model (Doppamo) project has a total budget of 213.6 billion won, which is 2.3 times the total procurement amount. Five elite teams, including Naver Cloud, Upstage, SK Telecom, NC AI, and LG AI Research Institute, have been selected to receive comprehensive support for GPU data talent.
The 'Everyone's AI' project, aimed at creating a free AI chatbot for all citizens, will utilize 512 NVIDIA B200 GPUs owned by the government. When converted to market rental rates, this amounts to approximately 40.5 billion won worth of government infrastructure being allocated to two or three companies developing public services. This project also requires that more than half of the independent AI models be utilized, raising concerns that only a few large enterprises overlapping with the Doppamo teams will benefit.
Through the Public Procurement Service, the scale of projects that small and medium-sized IT companies can secure in a year is dwarfed by just three projects from large enterprise consortiums. Furthermore, these projects are conducted not through competitive bidding by the Public Procurement Service but through government-led solicitations and special purpose corporation (SPC) investments, limiting the avenues for small businesses to participate.
A lack of coordination among government departments is also seen as a variable. Among new projects this year, there have been instances where ten different ministries simultaneously entered the same commercialization support project. Of the 20 new projects exempt from preliminary feasibility studies, 14 are AI-related. As the procurement structure shifts from construction to services and expands from a few local governments to nationwide, the verification process for the large-scale budgets allocated to large enterprises has become relatively lax.
A representative from a mid-sized IT company stated, “In fact, there have been many cases where the government has integrated existing projects from small and medium-sized enterprises and handed them over to large companies under the guise of AI transition. The market is increasingly centered around large enterprises or flowing towards American capital companies.”
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.