South Korea launches decade-long AI and chip plan: reported at $880–900 billion
In June 2026, South Korean President Lee Jae Myung announced a public-private investment plan for AI and chips that, according to Cryptobriefing, is worth roughly $880–900 billion (about 1,350 trillion won) over ten years.
President Lee Jae Myung announced in June 2026 a public-private investment plan for AI and chips that, according to Cryptobriefing, is worth approximately $880–900 billion (about 1,350 trillion won) over ten years. The figures, however, are unsettled: BigGo Finance reports a different total of around 2,000 trillion won (roughly $1.5 trillion) for a program with essentially the same three pillars, and AIMag has not seen any primary government document. This is a review of what the plan actually commits to, pillar by pillar — and where the verifiable milestones end and the targets begin.
According to Cryptobriefing, the plan was launched on June 29, 2026, by President Lee, who described it as a "national survival strategy" — a framing term reported through the source, not a quotation AIMag has verified directly. The program rests, according to both sources, on three flagships: memory chips, AI data centers, and physical AI with robotics. Cryptobriefing characterizes it as one of the largest national technology commitments on record; that characterization, too, should be read as the source's, not as an established superlative.
What was announced, and when
The dating is itself an example of the limits of the source base. Cryptobriefing gives June 29, 2026, as the announcement date, while BigGo Finance says only that the government formalized the megaprojects "in June," without a precise day. The June 29 date is reported here with reference to Cryptobriefing; the exact date cannot be confirmed from the available documentation.
The plan also has a backstory. BigGo Finance describes the president launching his concept of merging AI and energy in a speech at Smart Energy Plus (SEP) 2025, on October 15, 2025, at COEX in Seoul. The vision was thus formulated in the fall of 2025 and formalized as megaprojects the following year. AIMag has not previously covered the program; this article is therefore a first treatment, not news from the announcement day.
Pillar 1: Memory chips in Gwangju
The heaviest of the three pillars concerns memory chips. According to Cryptobriefing, Samsung and SK Hynix are committing a combined roughly $518 billion (800 trillion won) to new memory fabs in the Gwangju region. In addition, about $52 billion is to be set aside, according to the same source, for an HBM (High Bandwidth Memory) packaging hub — the memory technology that dominates supply to AI accelerators.
It is important to stress what kind of figure this is. These are company commitments as reported by one secondary outlet, not confirmed government budget lines. How binding they are, and how the costs are split between the state and corporate balance sheets, does not emerge from the documentation. Whether the companies actually build out in Gwangju at this level, and on what timeline, must be clarified by future reporting.
Pillar 2: AI data centers and the power question
The second pillar is AI data centers, with a program of roughly $356 billion (550 trillion won) according to Cryptobriefing, with SK Group, GS Group and Naver as the reported participants. The capacity targets are concrete: 8.4 GW by 2029 and more than 18 GW by 2035.
The two targets say something about the time horizon: a more-than-doubling of the capacity target in six years. But the figures also expose a documented knowledge gap. The financing of the power buildout this implies — power plants, grids and connections — is not described in the sources, nor is how the program framework is divided between state and private actors. With SK Group, GS Group and Naver named as the carriers, much of the capital likely sits with the companies, but that is an assumption that should be kept separate from what is actually documented.
Pillar 3: Physical AI and robotics
The third pillar differs from the first two in consisting primarily of government targets rather than heavy investment figures. According to Cryptobriefing, Seoul wants South Korea to rank among the world's three leading AI robot powers by 2030, and the country plans to train 10,000 specialists in physical AI by the same year.
These are targets, not funded budget lines. The source material contains no breakdown of amounts, program structure or sub-milestones for the robotics pillar beyond the 2030 targets. This part of the plan should therefore be read as a level of ambition, not committed funds.
How the program is organized
The overall structure, according to Cryptobriefing's analysis, is a public-private partnership rather than a pure state spending bill: the state sets direction and targets, while corporate balance sheets are to contribute much of the capital. This characterization comes from the source's analysis, not from an official government document.
That has a practical consequence: the figures reported for Samsung, SK Hynix, SK Group, GS Group and Naver are private investments stimulated by political direction — not items in a state budget. Overall delivery therefore depends on the companies actually following through, which in turn depends on market developments, capital costs and technology trends over decades.
The numbers problem: 1,350 or 2,000 trillion won?
The clearest tension in the source material is the total. Cryptobriefing reports roughly 1,350 trillion won ($880–900 billion) over ten years. BigGo Finance reports roughly 2,000 trillion won (about $1.5 trillion) for what it describes as the same three megaprojects over ten years.
The figures cannot be reconciled from the available documentation. Possible explanations — differing definitions of what is included, differing timeframes, or differing treatment of public versus private capital — are real but unresolved. AIMag therefore presents both figures with source attribution, without taking a position on which is correct. An official government announcement or a budget decision would settle precisely this: the actual program framework and its composition.
What cannot yet be verified
There is no primary government documentation in the available material — no press release from the ministry, no official plan summary, no decision by the National Assembly. All figures, dates and company commitments thus rest on two secondary sources which additionally disagree on the headline number, and which cannot be treated as independent confirmation of each other.
This affects three points in particular:
- The actual total and the split between state and private capital.
- The binding character of the company commitments from Samsung, SK Hynix, SK Group, GS Group and Naver.
- How the power buildout implied by the 8.4 GW and 18 GW targets is to be financed and integrated into the grid.
What to watch
Despite these unresolved points, the program is concrete enough to be tracked over time. Three things will say more about how real it is than the announcement itself.
First: official government documents. Confirmation from the relevant South Korean ministry, with a final program framework and timeline, would clarify both the total and the program's organization.
Second: concretization in the data center pillar. 8.4 GW by 2029 is a hard milestone. If financing, grid connections and actual construction do not become visible in the coming years, that is a clear signal that the plan lies more at the level of ambition than execution.
Third: progress toward the 2030 targets in physical AI. The top-three placement in AI robotics and the 10,000 trained specialists can be measured against actual activity in education and industry in the years ahead.
Whichever total turns out to be correct, the plan — according to the source's own characterization — is among the largest national technology commitments on record. What actually gets built and financed, and what remains political target-setting, the next couple of years will show.