Идеи
US stocks are stretched and vulnerable.
US equities are in bubble territory because valuations are stretched across multiple metrics: the Shiller CAPE is around 41 versus 44 in December 1999 and a long-term average near 17, the real equity risk premium is 1.1%, less than half its average since 2010, and the Buffett Indicator is around 240%, far above the level Warren Buffett called risky. With AI investment likely to slow, earnings expectations likely to fall, and equity supply from IPOs and expiring lockups set to increase, the market is vulnerable to a drawdown.
AI supplier earnings growth should decelerate.
The AI investment boom is currently giving a big impetus to the economy and to suppliers of AI hyperscalers, but by 2027 the increase in AI CapEx will likely be smaller than in 2026. Since it is the change in investment that matters for supplier earnings growth, profit growth expectations should come down and margins should be compressed, creating downside for AI infrastructure suppliers.
AI hyperscalers face overcapacity and competition.
AI hyperscalers are at risk of a boom-bust cycle because everyone is spending aggressively in a winner-take-all race, which will likely end in overcapacity and impaired pricing power. Dudley doubts hyperscalers can generate the roughly $2 trillion of revenue needed to justify the investment, and he expects only 2 or 3 of the current 7 or 8 AI providers to remain viable long-term, making the transition painful.
Fiscal risk could spike long-term yields.
Long-term Treasury yields are a key risk because the US has a very large budget deficit that is already weighing on the bond market. If investors lose confidence in US fiscal sustainability, yields could spike higher and become the precipitating event for broader market stress; Treasury buyback announcements are only tactical and do not solve the fiscal problem.