Singapore's Temasek warns of the ‘biggest risk’ facing markets right now

Temasek CIO Rohit Sipahimalani said an unwinding of the AI trade is the biggest market risk, though he does not see it as imminent and said a setback in 2027 is possible. AI earnings momentum has helped keep the S&P 500 near record territory, even as roughly half of Russell 3000 stocks were at least 20% below their June highs. Temasek remains bullish on AI long term; about half its AI exposure is in public assets, and it would ideally raise that share to 70%–75% for greater flexibility.

Analysis

The key vulnerability is not simply AI valuation: it is the feedback loop between concentrated index leadership and the capex cycle. If customers cannot show adequate returns, spending deferrals could pressure the AI infrastructure chain first—accelerators, networking and data-center buildout—then weaken earnings support for the broader index. Conversely, firms that can demonstrate productivity gains may retain pricing power even as infrastructure vendors re-rate. The breadth signal makes a concentration unwind more consequential than a routine sector correction, but does not establish that an unwind is imminent.

Temasek’s preference for more liquid exposure is best read as a portfolio-construction warning, not a bearish forecast. Public-market price discovery could accelerate while private AI valuations adjust more slowly; any rotation out of private holdings would also be constrained by liquidity and fund terms. Near term, watch earnings commentary on customer ROI and capex plans; over 1–3 months, those disclosures matter more than broad AI sentiment. Over 6–18 months, safety regulation or persistently weak monetization could shift spending toward fewer proven use cases. The thesis is falsified by improving market breadth alongside repeatable customer ROI and sustained AI-related revenue growth. The source provides no valuation or positioning data, so avoid treating this as a timing signal for an outright short.

Key Decisions for Investors

• Do not initiate a broad AI short on this commentary alone. For existing concentrated AI or mega-cap growth exposure, consider a small 3–6 month QQQ put spread as conditional downside insurance; add only if earnings reports show customer ROI disappointment or capex guidance cuts. Define the premium as the maximum loss.
• Monitor semiconductor and data-center supply-chain earnings for order cancellations, lead-time normalization and customer concentration. A downward revision to orders or guidance would strengthen the case to reduce infrastructure beta; continued order growth without customer ROI evidence is a warning, not confirmation.
• Track equal-weight versus cap-weighted U.S. equity performance and Russell 3000 breadth. If breadth improves while AI leaders hold earnings momentum, reduce index-hedge exposure; if breadth deteriorates further and leadership earnings estimates roll over, increase the hedge.
• Treat private AI exposure as a liquidity and valuation-marking risk, not a near-term listed-equity catalyst. Verify any disclosed secondary sales, fund-level liquidity terms and public/private valuation gaps before positioning around a presumed forced sale.