On November 1, 2026, The Korea Times reported that South Korea's exports "jumped 93.7 percent to a record $115.3 billion in October". Semiconductor exports "skyrocketed 259 percent from a year earlier to $56.4 billion in October", which the report attributes to strong global chip demand and investment by large technology companies in AI infrastructure. Cumulative exports for January to October "exceeded $900 billion for the first time", with "the ministry projecting that the country's annual exports could reach $1 trillion as early as November".
The budget speech a day later
On November 2, President Lee Jae Myung presented next year's budget to the National Assembly. The Korea Herald reports that the government "will inject 21.3 trillion won into building a semiconductor manufacturing hub and data centers", together with "developing physical AI and promoting AI" under its three megaprojects. A separate 4.6 trillion won will be used to "acquire a sovereign AI model", and 13.8 trillion won will go to "10 strategic technology sectors". The government is "proposing 820.9 trillion won ($614.5 billion) in total government spending next year". These are proposals that the Assembly still has to pass.
The arithmetic
The following is our arithmetic from the reported figures, not an official statistic. Chips were $56.4 billion of $115.3 billion, or about 49 percent of October's exports. A 259 percent rise means the same month a year earlier was about $15.7 billion (56.4 divided by 3.59). In one year, the country's largest export category has more than tripled on demand for AI hardware. In the same week, the government proposed large allocations for AI capacity at home.
What remains uncertain. Neither article separates memory chips or high-bandwidth memory from other semiconductors, so the figures do not show how much of the rise is accelerator memory. The Herald article does not say how the sovereign model will be "acquired" or what hardware the data center allocation buys. Monthly trade figures are often revised.
Analysis: For teams that buy inference rather than build chips, when one major exporter's chip sales more than triple in a year, demand for AI hardware is clearly still rising fast, and there is no reason in these figures to expect GPU capacity to get cheaper soon. For Belgavi's Book Store, Geetha could plan next year's help-desk capacity in two tiers. The large model would handle the hard CASE tickets, and a small classifier that routes routine return and order-status questions would keep working if GPU capacity gets tight. Planning for a shortage costs little when there isn't one.
Sources and evidence
The Korea Times reported on November 1 that October exports rose 93.7 percent to a record $115.3 billion, with semiconductors up 259 percent to $56.4 billion (about 49 percent of the total, by our arithmetic). On November 2, President Lee's budget speech proposed 21.3 trillion won for a chip hub, data centers and AI, 4.6 trillion won to acquire a sovereign AI model and 13.8 trillion won for 10 strategic technologies, within 820.9 trillion won of total spending.