Investors rotate from AI chips to hyperscalers on capex slowdown
A sharp projected deceleration in cloud computing capital expenditure is prompting fund managers to pivot from crowded semiconductor stocks toward hyperscalers, testing the sustainability of AI hardware valuations.
Global investors are shifting capital away from semiconductor manufacturers and toward the cloud giants that buy their chips, driven by mounting evidence that AI infrastructure spending will peak sooner than expected. UBS estimates project hyperscaler capital expenditure will surge 76% this year to $673 billion. However, that growth trajectory is forecast to decelerate abruptly to 25% in 2027 and a mere 6% in 2028.
This looming slowdown presents a critical valuation test for chipmakers that have ridden the AI wave to record highs. The Philadelphia Semiconductor Index has doubled over the past year, even after shedding nearly 18% since its June peak. By comparison, the equal-weighted S&P 500 is up just 11% and Europe's STOXX 600 has gained 8%. Bank of America's July survey found 82% of fund managers consider semiconductors the most crowded trade globally, with zero respondents holding bearish positions.
Financing squeeze emerges
The mechanics behind the AI build-out are also facing heightened scrutiny. Having exhausted internal cash flows, major technology companies are increasingly turning to debt markets to fund expansion. Apollo data indicates investor appetite for this debt is cooling, with bond cover ratios dropping below two times in July from nearly five times in February. The Bank for International Settlements warned in June that subpar AI returns could choke off financing entirely, risking a prolonged downturn.
Physical limits constrain growth
Capital allocation is not the only friction point for data center expansion. Roughly 70% of planned US projects face local opposition tied to heavy electricity and water consumption. New York this week instituted a one-year moratorium on large new data center construction, a regulatory precedent that could disrupt deployment timelines across the industry.
Divergent strategies emerge
Portfolio managers are positioning for a transition where cloud companies benefit from AI deployment rather than just supplying the infrastructure. Empirical Research points to a stark disconnect between slowing capital expenditure and bullish hardware revenue forecasts, suggesting analysts will eventually have to cut their numbers. Despite the risks, DWS maintains an overweight in semiconductors after taking some profits, while adding exposure to industrial and electrical equipment makers.
Broad conviction in the underlying theme remains intact for now. Chip-focused funds pulled in a record $10 billion in net inflows through May, according to Morningstar. Yet the market is clearly maturing, with investors increasingly pairing infrastructure bets with positions in enterprise software, healthcare and finance to capture AI adoption downstream.
