Some Investors Rotate into Hyperscale Cloud Providers, Betting on Decelerating Capex Growth
Miles Bennett
Active fund managers are trimming semiconductor holdings and adding Microsoft, Amazon and peers, betting that slowing capex growth will ease costs for cloud giants — yet the same logic undercuts the high-growth narrative for the chip supply chain.
What is the logic behind this rotation?
AI chip stocks have surged — the Philadelphia Semiconductor Index more than doubled over the past year — but valuations now sit well above what fundamentals can support.
Some managers are selling chips and buying cloud: cutting semis, adding Microsoft, Amazon, Alphabet and Meta.
This means → they are not betting against AI itself, but that the spenders (hyperscalers) are a better deal than the shovel-sellers (chipmakers) — once capex growth slows, cloud firms get cost relief while chip vendors lose orders.
How fast will capex growth actually decelerate?
UBS forecasts hyperscaler capex rising 76% year-on-year this year to $673 billion.
But growth narrows sharply — 25% in 2027, just 6% in 2028.
In plain terms = hyperscalers are still spending, but the *acceleration* is about to hit the brakes. For chip suppliers, the customer isn't stopping — it's just no longer buying more and more each quarter.
How crowded is the semiconductor trade?
Bank of America's July fund-manager survey: 82% call semis the most crowded trade in the market, and not a single respondent is short.
Chip-themed funds took in a record $10 billion in net inflows through May this year.
This reflects a dangerous consensus: everyone is on the same side. The SOX index has pulled back nearly 18% from its June peak, yet still far outpaces the equal-weight S&P 500's roughly 11% gain over the same period.
Who is already repositioning?
Edmond de Rothschild Asset Management's Alexis Bossard has sharply cut semiconductor exposure, added Amazon, and flagged liquid cooling, cybersecurity and select software names.
LFG+ZEST CIO Alberto Conca has trimmed memory-chip and equipment-maker positions, rotated into hyperscalers and healthcare, and bought put options on individual semi stocks as a hedge.
Conca's logic is blunt: hyperscaler free cash flow is being almost entirely consumed by capex. This means → they will eventually turn conservative on spending growth, and chip-vendor order momentum will slow.
What is going wrong on the financing side?
Hyperscalers have burned through their own cash on early AI buildout and are increasingly turning to external financing.
Apollo chief economist Torsten Slok notes that bond subscription ratios — a gauge of how eager investors are to lend — fell from nearly 5× in February to under 2× in July.
In plain terms = bond-market appetite for Big Tech debt is cooling fast. The Bank for International Settlements warned in June that if AI investment returns disappoint, financing could snap shut, turning the capex boom into a prolonged slump.
Where does the ultimate uncertainty lie?
Empirical Research flags the core mismatch: capex growth is decelerating, yet chip and AI-infrastructure suppliers still carry very high revenue expectations — the two lines are diverging.
This means → either hyperscalers raise their capex trajectory again (suppliers breathe easy), or suppliers must find growth elsewhere — otherwise estimates get cut.
DWS senior portfolio manager Madeleine Ronner takes a more optimistic view: she expects this earnings season's hyperscaler guidance to still support continued investment, and notes that buy-side capex forecasts for 2027 sit well above sell-side consensus. But the market's ultimate question remains: can hyperscalers' downstream software monetization justify the massive hardware spend across the entire supply chain?
Content is for reference only, not financial advice.