If Big Tech Cuts AI Spending, What Happens to Chip Stocks?
AI's long-term growth and chipmakers' near-term profits may not move together. A Bank of Korea downside scenario for 2027 raises a question for US equity investors too.
How should investors read a 140 trillion won figure?
According to Reuters on Oct. 27, the Bank of Korea analyzed a downside scenario in which major US technology companies cut AI infrastructure spending in 2027. The analysis was in a written response submitted to South Korea's National Assembly.
The central bank estimated that South Korea's chip exports could come in $50 billion to $100 billion below its baseline forecast. At an assumed exchange rate of 1,400 won per dollar, that is roughly 70 trillion to 140 trillion won.
The baseline matters here.
This does not mean chip exports would fall by 140 trillion won from today's level. It means exports could be that much lower than in the baseline, where AI investment keeps expanding. Actual exports could still grow and yet fall short of earlier expectations.
Under pessimistic conditions that include a 10% drop in chip export prices, the Bank of Korea said South Korea's 2027 GDP growth could be about 0.4 percentage point lower. That does not mean growth would fall to 0.4%. Nor is it a forecast of how 2027 will turn out.
Why can chipmakers struggle while AI keeps growing?
More people using AI services is a separate matter from ordering new chips for data centers.
If companies run the GPU servers they have already built more efficiently, or extend the life of existing equipment, usage can rise while growth in new server orders slows.
For chipmakers, that gap matters. In memory, average selling price (ASP) drives revenue and profit as much as unit volume does.
For example, if memory shipments rise 10% but prices fall 10%, revenue declines by about 1%. Higher volume does not guarantee higher revenue. In chip manufacturing, where fixed costs are heavy, falling prices can squeeze margins even harder.
How memory revenue works (a simple hypothetical, starting from revenue of 100)
| Case | Revenue |
|---|---|
| Baseline | 100 |
| Volume +10%, price unchanged | 110 |
| Volume +10%, price -10% | 99 |
Revenue = shipments x average selling price. This is an illustrative calculation, not company results or a Bank of Korea estimate.
What should US investors separate?
Companies in the same AI industry feel a spending slowdown differently.
| Company | Key variables | Possible impact |
|---|---|---|
| Micron (MU) | HBM and DRAM prices and shipments | Lower memory prices and margin pressure |
| Nvidia (NVDA) | GPU orders, customer capital spending | Slower growth in new orders |
| Big Tech cloud companies | Capex, AI revenue, cash flow | Possible relief from the investment burden, though slowing AI revenue is also a risk |
Micron investors in particular need to separate the growth potential of HBM (high-bandwidth memory) from the cyclicality of commodity DRAM. Even if high-value HBM demand holds up, rising supply of general-purpose memory could pressure profitability across the company.
Hyperscalers, by contrast, could see free cash flow (FCF) improve if capex growth slows. That is a positive only if AI revenue and competitiveness hold.
For long-term investors, three things need to be kept apart: the growth of the AI industry, the earnings growth of chipmakers, and the expectations already priced into shares. They do not always move at the same speed.
Insight Times Editorial Desk
