How AI Could Wreck Your 401(k)

Artificial intelligence: boom, bubble or bust? Probably a mix of all three. Whatever comes of it, it increases investment risk. It suggests that a lot of retirees and near-retirees need to rethink their portfolios.

The problem is not just that much of the stock market’s exuberant valuation hangs on data centers, via companies like Amazon, Meta Platforms and Nvidia. It’s that AI might prove to be so powerful that it destabilizes a large fraction of the economy. After that: unemployment, recession, market crash. Maybe.

The putative crash is far from certain. But the mere possibility should influence your thinking.

Now look at your retirement portfolio. If it’s in a target-date fund, you do nothing. The stock allocation is being automatically reduced as you age.

If you are young, you can stand to do nothing. If the market crashes, it will probably recover before you spend the money.

But if you are over 55? And setting allocations on your own? You may have veered off course. That’s because stocks have raced well ahead of bonds.

If you had a traditional 60/40 portfolio a decade ago, and made no adjustments or additions, you now have a lopsided 84/16 portfolio. More precisely: A 60% stake in the Vanguard Total Stock Market fund, paired with a 40% stake in the Vanguard Total Bond Market fund, would now be 84% invested in stocks and 16% in fixed income.

If either inertia or bullishness has caused your portfolio to tilt, you have plenty of company. Vanguard Group has the evidence, taken from the numbers in its 401(k) accounts. Its last survey, for December 2024, revealed that half of savers over 55 who manage their own allocations had more than 70% of their money in equities. Given last year’s results (a 17% return on stocks, 7% on bonds), it’s likely that these people are in even riskier territory now.

Crash? It has already come to software vendors. Adobe and Salesforce rent software to corporations, software that could be undercut by AI code-creating agents that compete with human programmers. Their stocks are down 40% and more over the past year. Block is laying off 4,000 workers.

A wider crash is hypothesized by Citrini Research. In a doomsday vision published on February 22, two analysts there posit a world two years hence in which artificial intelligence has first destroyed coding jobs, then almost any kind of white-collar work. People in insurance, travel booking, finance, real estate, customer service or anything that involves human intermediation get the ax. Productivity goes up but so does unemployment. Laid-off product managers take jobs as DoorDash deliverers. Personal income tax revenue collapses. Mortgages go unpaid. Economic chaos ensues.

It’s quite a fantasia, ending, whimsically, with the modest retraction: “Some of these scenarios won’t materialize.” Still, this kind of talk gives you a shudder.

Amazon, Meta, Alphabet, Microsoft and Oracle are collectively planning to sink $700 billion into AI this year. We can hope that this capital outlay will not result in a lowering of living standards. There is a precedent. Farm machinery disrupted jobs but did not make the country poorer.

A less hysterical review of artificial intelligence comes from Mark Zandi and fellow economists at Moody’s, the bond rater. In a recent report they optimistically assign a 40% probability to this positive outcome: AI causes some job losses, boosts productivity, keeps corporate profits on their upward march and leads in the end to prosperity.

The Moody’s team puts a 25% probability (these percentages are nothing more than hunches) on a pessimistic sequence: AI disappoints, its providers get far less revenue from it than they expect, their stock prices plunge and take the tech-heavy S&P 500 down, spenders feeling suddenly poorer cut their spending and we get a recession.

Other outcomes, per Moody’s: 20% probability of a job market upheaval leaving some skilled workers much better off and a lot of workers worse off; 15% probability of a gradual productivity gain of the sort seen in the early days of the internet.

Even if you ascribe low probabilities to the grim trajectories, it would be wise to rethink portfolio allocations. The main reason for owning risky stocks is that, historically, they have done much better than bonds. But how well do past results point to future ones?

It is well known that, over the past century, U.S. stocks have delivered a real return (return net of inflation) of better than 7% annually. Less well known is what made this happen. The return was driven by earnings yields (earnings divided by stock price) averaging 7%.

The relationship is straightforward. Acme’s shares, trading at $100, earn $7. The $7 can go out as a dividend that can be used by the holder to invest in more shares, or be used by Acme to buy other corporate assets with 7% earnings yields, or be used by Acme to buy in Acme’s own shares. Via any of these three routes Acme delivers a 7% compound annual return.

This oversimplifies a bit. In the past century corporations would have had to reinvest a portion of earnings in order to keep their earning power constant in real terms. That pushes the potential return below 7%. Offsetting this: A rise in price/earnings ratios, pushing the historical return on stocks back above 7%.

A 7% earnings yield equates to a P/E of 14. The S&P 500 index is now trading at 28 times 2025 earnings, double the historical norm. That tells you to expect future returns only half the 7% seen in the past.

A 3.5% expected return is better than what you get on Treasury bonds, but not enough better to justify the risk of having all your money in stocks. If you are 84% in equities and nearing retirement, you might want to reallocate some of the portfolio to Treasury Inflation Protected Securities (TIPS).

Schwab has an exchange-traded TIPS fund at an annual fee of 0.03%. Vanguard and Fidelity have open-end funds at an annual 0.05%. All three funds have maturities near 7 years, with real yields averaging 1.5%. If your 401(k) provider doesn’t let you into cheap funds, have your employer fire the 401(k) provider.

If you are investing in a self-directed IRA and have more than $100,000 to put into each position, buy equal quantities of TIPS due in 5, 10, 20 and 30 years. Average maturity: 16 years. Average real yield: 1.9%.