[Weaponized AI, K-Defense Compass]② “A Defense AI Budget Less Than the Price of One KF-21... Preparing for an AI War with Only 90 Billion Won”
Interview with Kiil Choi, Director of the Korea Defense Industry Research Institute
Urgent Need for Defense AI Data Center
Earliest Completion Projected for 2030
No Army Budget Allocated for GPUs
Frequent Rotation Hinders
The United States is rapidly integrating civilian artificial intelligence (AI) technology into the defense sector, spearheaded by the so-called "SHARPE ecosystem," which centers on six defense startups such as Anduril, Palantir, and Shield AI. The government builds a data and AI platform, incorporates the latest civilian company models, and opens massive amounts of defense-collected information to the private sector for training AI using real-world data. In contrast, South Korea does not even have a defense AI data center, which could be considered the starting point of such an ecosystem. Its completion is forecast for around 2030. Meanwhile, the military is unable to properly secure the necessary GPUs (high-performance graphics processing units), and defense contractors have not been provided with a single piece of defense data, despite such data being a core element for AI development.
In a recent interview with The Asia Business Daily, Kiil Choi, Director of the Korea Defense Industry Institute, remarked, "Our defense AI still has a long way to go." Choi is an advisor to Army Headquarters and defense contractors on AI issues and is listed as a defense AI specialist at Seoul AI Foundation. During his time at the Blue House (Presidential Office) National Security Office's Cyber Intelligence Secretariat, he handled quantum computing and AI, and for nearly a decade, he has led research projects and published papers on defense AI as a leading expert in this field.
"Out of the 66 Trillion Won Defense Budget, Only 90 Billion Won Goes to AI"
The first point Choi raised was the budget. He said, "This year, the government set an expenditure budget of about 700 trillion won, of which 66 trillion won, or about 10 percent, is assigned to defense. Out of that 66 trillion won, only roughly 90 billion won is allocated to defense AI." He added, "The unit price of a KF-21 fighter jet is over 100 billion won. It is equivalent to pushing forward with defense AI with less than the cost of a single fighter jet," he pointed out.
A lack of budget translates directly into infrastructure gaps. Choi commented, "To implement defense AI, we need a defense AI data center, but we have none at this time. The Ministry of Defense and government are pushing for its construction, but the earliest possible completion will be around 2030. So how do they propose to advance defense AI until then?" he asked.
The situation regarding GPU procurement is even graver. He said, "The Army could not even allocate a budget for GPUs, so it doesn't have enough funds to buy just one. The Agency for Defense Development (ADD) owns about 60 GPUs, yet when we requested to borrow just over 10, ADD responded that they could not lend them out due to their own shortage."
Choi also recounted a story told to him by a defense company CEO at the end of April. He stated, "When asked about the progress of defense AI adoption, the CEO pointed out two issues. The first is that there is no clear definition of the type of AI the military actually wants." He explained that, in order for defense contractors to develop the appropriate platforms, the military as the end user must present clear requirements, but such criteria do not currently exist.
The second issue he identified was data sharing. Choi said, "Data is the core of AI. Advancing defense AI requires sharing defense data with private sector defense companies. But so far, not a single piece of defense data has been shared." He continued, "If they refuse to share data citing security reasons, there will be no progress. How does the US do it? Even companies like Palantir and Anthropic are private, but they collaborate under tight control and clear boundaries regarding what can be shared. It is important to reference such overseas cases to establish clear guidelines for data sharing."
"We Have 5,000 Drone Companies, But No Ecosystem"
He went on to point out that, unlike the United States' SHARPE example—where the private sector forms an organically connected defense AI ecosystem—South Korea lacks such a system. "They say we have more than 5,000 drone companies in Korea, yet annual production is only about 200,000 to 250,000 units. By comparison, DJI, a single Chinese company, produces 20 million units per year on its own. Having many companies does not mean success; forming an integrated ecosystem that brings them together is a much more important task," Choi stressed.
Manpower problems are another stumbling block. He said, "While the military claims to be training AI specialists who can deal with the unique demands of national defense through commissioned education programs, in reality, after finishing such training, personnel must move to a different post due to the principle of rotation. If they don't complete essential posts, they can't even be promoted." He added, "In fact, an Army Headquarters defense AI administrator was recently notified to move to a new post less than a year after assuming the role. For specialized, expertise-critical roles like defense AI, the overall personnel management policy must be adjusted to allow for long-term placements."
Choi cited interoperability between different weapons systems and rapid acquisition systems to quickly transfer commercial technologies to defense as the central issues facing the defense industry. "The current rapid acquisition regime evolved from the earlier concept technology demonstration (CTD) program, which itself was benchmarked on the US MTA (Middle Tier Acquisition) system. However, to properly implement such systems, the original objectives and methodologies need to be fully understood and analyzed before import, but the current approach does not even come close to imitation—resulting in repeated trial and error," he said.
Hot Picks Today
"Two Spend 500,000 Won in a Flash, Yet Wait Two Hours?" ... 'Dreamland' Shunned by Kids and Foreign Tourists
- The Era of Getting 200 Million Won After 30 Years Is Over... A Japanese Bank's "Perpetual Department Head" Who Just Clocked In Cries Out
- Spain Hit by Its Worst Tragedy... Lee Kangin, Moreno and Real Madrid Join Soccer World in Mourning
- They Said, “Make It Your Home for Life” - Then Suddenly Kicked 200 Passengers Off... What Happened to the Around-the-World Cruise?
- "130 Rice Bowls Sold Out in a Day"... 'Unfiltered Face' Adds 8 Million Won to Night Market Sales: China's Bungeunwoo
He also mentioned the "Genesis Mission," which launched last December after being signed by President Trump. Just as this initiative focuses state-led investment in cutting-edge AI and quantum technologies, he stressed the urgent need for clear national priorities and fiscal support in Korea. He also suggested utilizing the Korea Development Bank’s National Growth Fund as a source of funding. "Given that concentrating funds on a single company may trigger fairness concerns, it would be worth considering having companies such as LIG Defense & Aerospace, Hanwha Systems, and Hyundai Rotem form a consortium, establish a special purpose corporation (SPC), and funnel investment there," he proposed. In closing the interview, Choi emphasized, "The priority must be to quickly reflect these issues in the medium-term national plan, and speed up the establishment of the AI data center."
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.