When does prudence become crowd risk in disguise? UK pension schemes paring back US equities on AI bubble fears look sensible at first pass. The Bank of England has warned on stretched tech valuations, macro heads have noted misplaced certainty, and several managers have proposed equal-weight or small-cap shifts to dilute the Magnificent Seven’s grip on the S&P 500. Yet the dominant risk here is not whether AI is overvalued. The risk is that a synchronized move to “safety” creates new single points of failure in liquidity, governance, and policy. The 2022 LDI episode was a reminder: systems fail where they are thinnest, not where they’re loudest.
Market concentration is a feedback machine. Cap-weighted indices send new money to yesterday’s winners, which raises weights and draws in more allocators. That is a classic coordination game: stay as long as others stay. Defection is rational only if enough players defect together. Bridgewater’s leadership has flagged that risks in the AI-led rally may be underpriced. The Bank of England has said tech valuations look rich. UK pension funds, long overweight US markets, are now stepping back. But coordinated defection creates its own fragility. If many large funds try to escape the same door, factor exposures compress elsewhere. Value, equal-weight, and small-cap baskets become a new crowd. Monoculture is fragile whether the crop is mega-cap AI or mid-cap industrials.
Equal-weight funds reduce single-name concentration, but they import other risks. They raise turnover and transaction costs. They tilt toward smaller and more indebted companies that are more sensitive to refinancing cycles and credit spreads. They carry higher liquidity risk in a selloff. Fidelity and others suggest equal-weight or small-cap tilts to blunt AI exposure. Fair. But this is a trade, not a virtue signal. You are swapping momentum and mega-cap crowding for liquidity and leverage risk. In stress, equal-weight indices can gap lower when dealers widen spreads and active cash steps aside. During the dot-com unwind, many new-economy names fell 80 to 100 percent. In 2008, small caps that were not in a bubble still lagged because liquidity vanished. Which path is more acceptable for a pension scheme matched to long liabilities?
Liquidity is capacity over time. Plenty of vehicles offer daily dealing, but the underlying markets do not magically scale because governance committees met on Tuesday. If a critical mass of UK schemes rotate away from US big tech at once, where does that money go, at what size, and through which pipes? UK domestic equities today represent a small slice of scheme assets. Pushing 10 percent of assets toward local markets by policy nudge tightens the corridor further. The system’s weak point is order-book depth and collateral liquidity, not a price-to-sales multiple in Silicon Valley. If dispersion rises and volatility spikes, the incremental seller of a mid-cap ETF will discover the difference between quoted spread and executable size. The 2022 gilt selloff was not a story of creditworthiness; it was a story of duration, leverage, and forced collateral calls. Rotation risk can rhyme with that.
Encouraging pension funds to allocate more to domestic markets sounds like diversification away from US tech. It also concentrates exposure to domestic policy and macro shocks. Japan has lived with this for decades as institutions toggled between domestic and foreign risk to support policy goals. The euro area’s sovereign bank loop showed how quickly “supporting the home market” can turn into correlated drawdowns when fiscal or currency stress hits. A 10 percent domestic equity allocation target may reduce US single-name risk but deepen the link between national funding needs and retirement outcomes. That is a different concentration. You diversify from fire to smoke if the house is the same.
The AI bubble debate often misses the pension question. Dot-com taught that being right about the future is not the same as being right about price. Railroads, electricity, and the internet were transformative, and investors repeatedly overpaid. Today’s winners throw off cash, unlike 1999. But AI economics require heavy capex and uncertain unit returns. The base rate for revolutionary technologies is a long S-curve with violent mid-cycle setbacks. For a pension scheme, the hazard is not permanent impairment at a handful of mega-caps; it is the sequence of returns and the collateral path along the way. A 30 percent drawdown early in a de-risking cycle can do more damage than the same drawdown later, because it interacts with funding ratios, hedging overlays, and rebalancing rules. Path dependence, not point estimates, determines solvency comfort.
You do not need to time an AI peak to protect a fund. You need to remove single points of failure. That means putting hard caps on issuer and sector weights at the total portfolio level, not just within a sleeve. It means diversifying by risk driver and liquidity tier, not market label. It means testing currency hedges under scenarios where the dollar weakens at the same time mega-cap tech sells off, forcing losses on both equity and hedge overlays. It means setting pre-committed rebalancing bands that force buying of what is falling and selling of what is rising, before the committee narrative shifts. It means accepting that some carry must be paid for convexity and that optionality is insurance, not a trade. Antifragility is designed ex ante; it is not a response memo after an FOMC day.
Passive is not riskless, and overlays are not free. Securities lending adds counterparty exposure. Synthetic exposures and currency forwards embed margin and roll risks. Factor sleeves layered on top of passive cores create implicit leverage when correlations jump. In a drawdown, what matters is who needs cash tomorrow morning and what they must sell to get it. The unseen fragility in many pension portfolios is the mismatch between daily-dealing vehicles and weekly or monthly liquidity of the assets underneath, especially in small caps or private markets. Rotation out of US mega-caps toward less liquid names may look diversified on a pie chart and more brittle in a tape bomb. Check the plumbing before changing the furniture.
Pulling back from an AI-fueled rally can be sensible. Doing it because everyone else is doing it is how risk migrates. A system becomes fragile when it optimizes on a single story, whether that story is American exceptionalism in tech or patriotic home bias. The remedy is not to pick the right story. It is to build portfolios with multiple ways to be wrong without breaking. That means humility in forecasts, discipline in process, and redundancy in liquidity. Markets punish monotony. The point is not to avoid bubbles. It is to ensure your fund does not need the exit door at the same time as everyone else.
