The Great AI Dilemma: How to Invest When Extinction Warnings Meet Market Highs

AI might be unlike any technology that has reshaped the global economy. That also means it brings risks unlike anything investors have experienced before. (CHARLY TRIBALLEAU / AFP via Getty Images)

Key Points

• Anthropic’s IPO registration document raises the possibility that AI could wipe out humanity, thrusting risk management into the spotlight.
• Investors can manage AI risk by raising cash, changing portfolio weightings, diversifying AI exposure, or using a barbell approach.
• Nvidia announced the creation of the Nvidia Open Agent Safety Platform, which is a form of industry-led self-regulation.

To invest or not to invest. That’s the AI conundrum.

AI is the engine of this stock market, but Anthropic’s own IPO registration document raises the possibility that AI could, well, wipe out humanity. So should investors sell before a Matrix-style AI disaster, which might not come, headed off by stock-crushing regulation? Or should they go all in on AI, fatalistically believing there is little they can do about how AI will develop or impact the global economy?

Selling can seem like the right choice, or even the moral choice, but there is a cost to underperforming the stock market, especially if this time is not different and a more benign AI outcome, with relatively safe AI agents enhancing productivity, becomes reality.

However things develop, the current conundrum has thrust risk management into the spotlight.

The Barron’s Daily

Free Weekly Newsletter

A morning briefing on what you need to know in the day ahead, including exclusive commentary from Barron’s and MarketWatch writers.

Email addressSubscribe

I agree to the Terms of Use, Privacy Notice and Cookie Notice.

I would like to receive updates and special offers from Dow Jones and affiliates.

Successful risk management strategies range from the mundane to the complex. The easiest option is to sell some stocks and raise cash. Investors can also change portfolio weightings. One way to de-emphasize the largest AI-fueled tech stocks is to take money from the market capitalization-weighted SPDR S&P 500 ETF and buy the Invesco S&P 500 Equal Weight ETF. Investors can also diversify their AI exposure, including everything from copper to cooling to chips to computing. There is also the barbell approach, which means going heavily into AI-related stocks such as SpaceX, with its orbital valuation and moonshot ambitions to build a huge AI computer in space, as well as non-AI stocks (are there any?) such as consumer staples. Goldman Sachs has also suggested investing in heavy-asset-low-obsolescence (HALO) stocks that will be around no matter what happens to AI.

All approaches have pluses and minuses. Boeing is a HALO stock. Only two companies make large planes. Still, Boeing has had its own issues tied to production and production quality. The stock is down more than 50% from its all-time high in 2019. As for staples, they have “poor fundamental momentum and are far from cheap, so…better to hold cash,” says Ironsides Macroeconomics’ Barry Knapp.

The equal weight strategy might work. CappThesis founder and market technician Frank Cappelleri says outperformance of large-cap tech stocks has left the S&P 500 “overbought” relative to the equal-weight index. Cappelleri isn’t making a fundamental call on valuations. He’s looking at stock charts and market history to understand what comes next. Overbought essentially means a lot of good news is reflected in a stock or an index. It’s a timing tool to help know when to reduce exposure.

Other risk management strategies can include selling call options on portfolio positions to raise cash. Options strategies tend to be more complicated and executed by more seasoned investors.

What’s best depends on each investor’s life stage and risk tolerance. Morbidly, truly horrific outcomes don’t really need to be discounted. People won’t worry about the Dow Jones Industrial Average holding 50,000 in an apocalypse.

What’s clear is that some sort of regulation is likely coming, and many steps have been proposed. An expert-led FDA-like panel to evaluate AI model safety and effectiveness is one potential leg of a stable AI-regulation table. Nuclear arms-style global cooperation is another. A few executives, including Microsoft co-founder Bill Gates, have suggested AI token taxes to raise the cost of banal queries and protect human labor.

No one wants the industry to default to a National Transportation Safety Board-style model, where safety investigators make recommendations after something bad happens. Still, an entity to monitor AI models after their release isn’t a bad idea. Neither is establishing penalties for AI model misuse.

Investors might believe that the likelihood of political cooperation is lower than the likelihood of AI disaster. Of course, the industry doesn’t want the gravy train to end. On Monday, Nvidia announced the creation of the “Nvidia Open Agent Safety Platform,” which is a form of industry-led self-regulation.

Loading tweet…

Self-regulation—driven by self-interest—can be part of the solution. It probably shouldn’t be the only part. As for investors, they shouldn’t rely on politicians or the industry to deal with their portfolio risks.

AI isn’t the only risk. Interest rates, oil prices, and geopolitical tensions also matter. AI is just the biggest one and will probably remain so for years.

Write to Al Root at allen.root@barrons.com.

Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8