Major pension funds are making a relative bet, not abandoning US stocks: they still hold them, but hold less than a global benchmark would prescribe. That reduces dependence on a few AI-linked giants while preserving a stake if US companies keep leading returns. The cost is that diversification can underperform the benchmark for as long as those giants continue to rise. 36

• Australian Retirement Trust has reduced its US equity weight relative to the MSCI World index this year and moved it below its own strategic allocation. Its manager described AI and broader US valuations as somewhat elevated, while saying he did not expect a repeat of the dot-com crash. ART has also bought Japanese and European shares and questioned whether heavy reliance on passive index tracking leaves it too exposed to a concentrated market. 789
• La Caisse and the People’s Pension likewise hold US equities below their respective global-benchmark weights. The available reporting establishes those positions, but provides insufficient evidence here to attribute a precise recent trade or a distinct explanation to either fund. 3
• The risk they are managing is not simply that AI will fail. When a handful of megacap stocks drive much of an index, high prices leave less room for earnings to disappoint; a market-cap-weighted passive fund automatically gives those stocks greater weight as their prices climb. That makes an apparently broad US index less diversified than its number of holdings suggests. 159
• The counterweight matters. ATP’s caution against treating this as an obvious exit underscores the risk of cutting a major source of returns too far or too soon. Marsh’s survey points to a broader shift in intentions: about a third of institutional investors planned to reduce US equity exposure over the following year, roughly twice the prior year’s share—not evidence that a third had already sold. 12

The trade-off is therefore between concentration risk within the benchmark and tracking risk outside it. An underweight can cushion a reversal in expensive AI-linked stocks, but it can also lag badly if their earnings justify today’s prices and the rally continues. 56

Major pension funds are making a relative bet, not abandoning US stocks: they still hold them, but hold less than a global benchmark would prescribe. That reduces dependence on a few AI-linked giants while preserving a stake if US companies keep leading returns. The cost is that diversification can underperform the benchmark for as long as those giants continue to rise. 36

• Australian Retirement Trust has reduced its US equity weight relative to the MSCI World index this year and moved it below its own strategic allocation. Its manager described AI and broader US valuations as somewhat elevated, while saying he did not expect a repeat of the dot-com crash. ART has also bought Japanese and European shares and questioned whether heavy reliance on passive index tracking leaves it too exposed to a concentrated market. 789
• La Caisse and the People’s Pension likewise hold US equities below their respective global-benchmark weights. The available reporting establishes those positions, but provides insufficient evidence here to attribute a precise recent trade or a distinct explanation to either fund. 3
• The risk they are managing is not simply that AI will fail. When a handful of megacap stocks drive much of an index, high prices leave less room for earnings to disappoint; a market-cap-weighted passive fund automatically gives those stocks greater weight as their prices climb. That makes an apparently broad US index less diversified than its number of holdings suggests. 159
• The counterweight matters. ATP’s caution against treating this as an obvious exit underscores the risk of cutting a major source of returns too far or too soon. Marsh’s survey points to a broader shift in intentions: about a third of institutional investors planned to reduce US equity exposure over the following year, roughly twice the prior year’s share—not evidence that a third had already sold. 12

The trade-off is therefore between concentration risk within the benchmark and tracking risk outside it. An underweight can cushion a reversal in expensive AI-linked stocks, but it can also lag badly if their earnings justify today’s prices and the rally continues. 56