HSBC Sees 37% Odds AI Markets Are Now Driven by Overspending Fear

HSBC's most likely AI regime is overspending fear, not demand collapse

HSBC sees the market's highest-probability AI regime as one defined by spending anxiety, with Asia already showing the strain.

HSBC's model now puts the highest probability on a "hyperscaler overspend" regime, in which AI infrastructure and semiconductor names can still outperform while the capex spenders lag. That is not proof that AI demand is breaking. It suggests investors are increasingly focused on who is footing the bill.

Crowded positions can become more fragile once that kind of fear takes hold. Bank of America's July survey found 82% viewed semiconductors as the market's most crowded trade. When many investors are exposed to the same setup, price can start to reflect who might be least willing or able to keep spending, rather than just long-run demand.

Recent chip-stock action shows how quickly sentiment can wobble. The SOX climbed 11% last month, then fell more than 11% over the following two weeks. After such a strong run, investors appear more sensitive to wallet pain than to the longer AI spend story.

Asia has been absorbing that shift early. HSBC closed its "overweight" stance on emerging markets equities after EM Asian equities slipped more than 2%, with South Korea's benchmark closing 5.35% lower even after Samsung forecast a 19-fold jump in second-quarter operating profit. The takeaway is not that AI demand has been disproven. It is that fears about hyperscaler spending can hit suppliers before the broader demand story cracks.

The debate is shifting from how much is being spent to how fast that spending is slowing

The core shift is straightforward: investors are moving from excitement about AI demand to anxiety about who is paying for it. The latest evidence points less to a collapse in spending and more to a change in what the market is willing to reward.

Absolute spending remains large

UBS expects hyperscaler capital expenditure to rise 76% this year to $673 billion. Last quarter, Big Tech spending still climbed 19% QoQ. Even Samsung forecast a 19-fold jump in second-quarter operating profit. Those figures suggest this is not a demand collapse.

What matters more for valuation is the next change in pace. Once investors start focusing on deceleration, softer capex growth can carry more weight than the fact that spending is still large.

Why suppliers may feel a slowdown first

If hyperscaler spending growth cools from explosive to merely fast, the payers' vendors are likely to feel it before the broader AI story weakens. UBS expects hyperscaler spending growth to slow to 25% in 2027 and just 6% in 2028. The absolute level can remain historic while the market rerates the pace.

That helps explain why investors may keep treating chips and networking equipment as the cleanest AI proxies even as the cycle matures. When growth slows, the concern is not only earnings timing. It is also whether the boom will prove as durable as hoped.

HSBC's framework points to more selective rotation

HSBC's framework says the "hyperscaler overspend" regime carries a 37% probability and sees data center and semiconductor names outperform "at the expense of the capex spenders". In practical terms, that looks less like a single AI trade and more like a market rotating among competing winners as spending becomes more selective.

Bull case: rotation rather than collapse

The bull case is that hyperscaler budgets keep expanding, but more slowly. In that scenario, investors may stop paying any price for infrastructure exposure and start rewarding the buyers themselves. That fits the emerging tilt toward software companies and major cloud providers, Goldman Sachs' view on AI platform stocks and productivity beneficiaries, and Morgan Stanley's call that investors may turn toward AI "hyperscalers" as semis cool.

Bear case: slower spending starts to pressure earnings

The bear case is stronger than simple rotation. If slower budgets begin to squeeze revenue visibility, margins, or earnings durability across the supplier chain, the rerating could widen beyond semis. That would be the point at which fear of hyperscaler wallet pain starts to spread beyond the payers' first-round suppliers.

What to watch now

• Whether positioning continues shifting from semis toward major cloud providers
• Whether the market keeps favoring AI platform and productivity-linked names
• Whether Asia continues selling through strong operating forecasts on fear of AI over-spending

If those signals hold, the key question is no longer whether AI spending is large. It is who can keep earning the premium as the pace of growth slows.

Positioning while the market argues about AI returns

The practical question is not whether AI spend is still huge. It is how to position while investors debate whether returns are arriving fast enough to justify the spending.

Base-case tilt

Stay constructive on AI, but lean more toward the balance-sheet owners than just the gear sellers. UBS expects hyperscaler capex growth to slow to 25% in 2027 and then just 6% in 2028. That does not end the AI build-out; it changes where the market may look for value.

What would confirm the base case

Look for signs that the market is rewarding large spenders for their scale while still favoring software and platform exposure over pure infrastructure leverage.

What would invalidate it

The cleaner invalidation is not weaker demand on its own. It is clear evidence that hyperscaler budgets are being restrained before the market fully works through a slowdown in growth.