The Fund Manager’s AI Infrastructure Call

For investors questioning whether the feverish rally in AI infrastructure stocks has finally peaked, one of the best-performing small-cap fund managers has a clear answer: not yet. The manager, whose fund returned 52% over the past 12 months and beat 99% of its category over three years, told Business Insider that the companies in his portfolio show no signs of a slowdown.

The trade has been a turbocharger for returns this year. The iShares Semiconductor ETF (SOXX) surged 72% during 2026 as hyperscale cloud providers poured billions into data-centre buildouts. But since 22 June, the ETF has fallen 16% amid growing fears that these hyperscalers may start reining in capital expenditure if they cannot demonstrate clear returns on their AI investments.

The fund manager’s track record lends weight to his optimism. He noted that the median stock he has sold since the fund’s inception went on to underperform its benchmark by 12 percentage points in the following 12 months. In other words, his sell discipline has historically been a reliable signal that momentum is about to fade. That he is not selling now suggests he sees more room to run.

Why the AI Infrastructure Trade May Not Be Over

The Manager’s Sell Discipline and Why It Matters Now

The fund’s history of well-timed exits gives the manager’s current stance unusual credibility. He does not sell merely because a position has gained; he sells only when the underlying business trajectory weakens. The fact that he is holding on to his AI infrastructure exposure—and revealing two specific names he likes—implies he believes the fundamental demand picture remains intact. His approach suggests that the recent sector-wide pullback is driven by sentiment, not by deteriorating orders or margins at the companies he holds.

What the Hyperscaler Capex Fears Are Really Saying

The 16% slide in SOXX since June reflects a market that is suddenly pricing in the possibility that Amazon, Microsoft and Google will dial back their infrastructure spending. Yet none of the hyperscalers have guided capex lower; the anxiety stems from a narrative that AI monetisation is taking longer than expected and that model efficiency gains may reduce hardware demand. The manager’s continued bullishness indicates that, at least from his vantage point, the order books and earnings guidance of his chosen companies do not yet support that narrative.

Where the AI Infrastructure Trade Could Still Have Upside

The broader AI buildout is still in its early stages. Even if hyperscalers pause certain projects, the industry is shifting towards more custom silicon and specialised networking, creating new pockets of growth. The manager may be positioning in companies that supply these next-generation components. The fact that he released two specific stock names during the interview suggests he wants to highlight where the remaining value lies—a move he likely would not make if he thought the sector was about to roll over.

Key Signals for Investors Tracking the AI Infrastructure Space

  • Track quarterly capex guidance from the major hyperscalers (Amazon, Microsoft, Google) as a leading indicator for AI infrastructure demand; any downward revision would undermine the bull case.
  • Watch the manager’s future filings and comments for signs that he is beginning to trim his AI positions—his sell timing has historically preceded underperformance by a wide margin.
  • Monitor SOXX’s technical levels around the 22 June high. A sustained break above that level with volume would suggest the pullback was a correction in an ongoing uptrend, while a failure to reclaim could signal deeper concerns.

Risk & Opportunity Assessment

Commercial RiskHighAI infrastructure stocks are heavily dependent on continued hyperscaler capex; any reduction in spending plans would likely trigger sharp price declines, as the 16% SOXX drop since June already illustrates.
Competitive RiskMediumThe fund manager’s strong performance could attract copycat strategies, but his distinct and historically successful sell discipline provides a competitive moat that is difficult to replicate.
Regulatory RiskLowNo imminent regulatory threats exist for AI infrastructure in most major markets, though potential export controls on advanced chips remain a tail risk.
Reputation RiskMediumIf the manager’s current call proves incorrect and the AI trade reverses sharply, his reputation for timely selling—which is a core part of his fund’s pitch—would be damaged.
Technology DisruptionHighRapid advances in AI model efficiency could reduce the volume of hardware required for training and inference, undermining the long-term demand thesis for AI infrastructure.
Commercial OpportunityHighThe AI infrastructure buildout remains in its early phases, and the manager’s conviction suggests his selected companies are poised to benefit from sustained investment, particularly as industries adopt custom silicon and specialised networking.