Goldman Sachs said in a report centered on an “AI super surplus” that the export boom in AI chips is driving current account balances in Korea and Taiwan to historic levels. The bank expects Taiwan’s current account surplus to rise above 20% of GDP this year, while Korea’s could move past 10%, both marking record highs.
The report makes a direct macro call. Demand tied to AI infrastructure is not only lifting growth, it may also force monetary tightening as stronger external balances feed into currency strength and reduce room for central banks to stay accommodative.
AI infrastructure demand is reshaping trade balances
Goldman Sachs labels the trend an “AI super surplus.” The bank links the shift to global spending on data centers, servers, compute cards, and memory chips. Companies named as major beneficiaries include Samsung and SK Hynix in Korea, and TSMC in Taiwan.
Analyst Andrew Tilton said this is the strongest technology cycle on record for Korea and Taiwan. Even though both economies are heavily exposed to energy imports from the Middle East, Goldman argues that the expansion in trade surpluses is large enough to offset that cost burden.
On growth, the bank expects Taiwan’s 2026 GDP growth to come in at close to 10%, up from 8.7% last year. Korea is projected to rebound from 1% last year to 2.5%.
Stronger surpluses may translate into rate hike pressure
Goldman’s view is that expanding current account surpluses will support stronger local currencies and add pressure on central banks to raise rates. Its forecast calls for the Bank of Korea to deliver one 25 basis point hike in the third quarter and another in the fourth quarter. For Taiwan’s central bank, Goldman expects one 12.5 basis point move in the second quarter and another in the fourth quarter.
Higher policy rates would raise borrowing costs and can weigh on household consumption and business investment. In export-heavy economies, currency appreciation may also make goods less competitive overseas. That tension sits at the center of Goldman’s thesis: stronger AI exports improve headline external balances, but they can also tighten financial conditions.
Tilton said the trend is likely to continue, though geopolitical risks and energy supply remain open variables. Goldman’s base case assumes the global AI infrastructure buildout lasts at least until before 2027, keeping chip export demand in Korea and Taiwan elevated.
Supply chain concentration is widening Asia’s divergence
The report also points to a “K-shaped” divergence across Asia. In Goldman’s framing, Korea and Taiwan are moving higher on the back of AI chip exports, while other economies in the region that rely more on traditional manufacturing and conventional export sectors face a different growth path.
The source material, citing Bloomberg, says the AI supply chain is highly concentrated. The world’s most advanced logic chips are produced almost entirely by TSMC, while high-bandwidth memory, or HBM, is dominated by SK Hynix and Samsung. Other Asian economies do not occupy the same strategic position in that chain.
That concentration means the global wave of AI compute demand is not being spread evenly across Asia’s exporters. It is being captured by a narrow group of economies with established foundry and packaging advantages, and Goldman’s report suggests that company-level chip exports are now large enough to shape macro data and rate expectations.