South Korea chip exports up 259%: don't budget for cheap hardware

South Korea just posted the clearest price signal in tech. On Sunday, the country's trade ministry reported that South Korea's October chip exports rose 259% to $56.4 billion, and total exports rose 93.7% to $115.3 billion. The ministry tied the chip number to excess demand for DRAM and to rising contract prices for both DRAM and NAND. If your 2027 plan assumes servers, laptops, SSDs, or AI compute get cheaper next year, this is the data that says don't.

What actually happened

Per The Korea Times, which carried the Yonhap report, October exports reached $115.3 billion, up 93.7% from a year earlier, and semiconductor exports reached $56.4 billion, up 259%. Imports rose 36% to $72.78 billion, for a trade surplus of $42.5 billion. Exports for January through October passed $900 billion for the first time.

The Asia Business Daily adds the detail that matters for buyers:

• Chips were 48.9% of all exports, the highest October share on record.
• Chip exports to the US rose 497%, the 10th month in a row of triple-digit growth.
• Computer exports grew by triple digits for the ninth straight month, driven by enterprise SSDs for AI servers.
• The ministry pointed to continued excess demand for DRAM and sustained increases in fixed prices for DRAM and NAND.

That last line is the key. Part of this growth is volume. Part of it is price. Memory makers are selling more chips and charging more for each one.

Why it matters for your business

Memory sits inside every device and every cloud bill you pay. When DRAM and NAND contract prices climb, the cost shows up later in laptop quotes, NAS and server builds, and eventually in cloud instance pricing. AI data centers are buying the same chips you need, and they are buying first.

We've watched this pattern before. Operators who wait for prices to fall end up buying the same hardware six months later at a higher price. Our read for small teams:

• Buy known needs now. If a laptop refresh or a storage upgrade is already on the 2027 list, price it this quarter.
• Don't build a P&L on falling AI prices. Per-token prices can drop while the hardware under them gets more expensive. Budget AI as a real, recurring line item.
• Spend less compute per task. Route easy work to small models, cache repeat answers, and trim context. That saves money at any memory price.

Sources: The Korea Times (Yonhap), The Asia Business Daily.