No Safe Haven: Why the Next Crash Could Hurt More

 

AI hype is soaring, but history screams warning: when bubbles burst, fortunes vanish fast, and the old safe havens might fail you this time.

On February 8th 2026, as America inhaled wings and beer and waited for the halftime show, an AI chatbot named Claude flashed across the Super Bowl screen. I didn’t see innovation. I saw déjà vu. I saw ghosts from 2000, when 17 dotcom firms bought 30-second ads during the same game, each torching millions for a shot at immortality. Weeks later, the Nasdaq began its long fall. The party lights went out. Portfolios bled.

History does not repeat, but it rhymes—and sometimes it screams.

Now the hype machine is back, only this time it runs on neural networks and trillion-parameter dreams. Alphabet, Amazon, Meta and Microsoft have pledged a combined $660bn on AI in the coming year. That is not pocket change. That is empire money. A year ago, Wall Street would have cheered. Today, investors flinch. Microsoft’s stock is down 16% since its spending plans landed. Meta popped, then wobbled. Confidence feels thinner than it did.

Everyone knows stocks are expensive. The S&P 500’s cyclically adjusted price-to-earnings ratio has hovered well above its long-run average of about 17. When valuations stretch, expected returns shrink. That is not ideology. That is math. When you pay too much for earnings, you bake in disappointment. The higher you climb, the thinner the air.

So I ask myself the question that keeps money managers up at 3 a.m.: how do I hedge a bubble when everything looks bubbly? The obvious answer is to sell. Cash out. Walk away. But I have watched this movie before. During the late 1990s, the Nasdaq rose nearly 12-fold in the 5 years to March 2000. Along the way, it suffered at least a dozen corrections of 10% or more. Any one of those dips could have scared a cautious investor into selling. Anyone who bought at the start of 1995 and simply held on would have doubled their money even after the crash. Timing a mania is like trying to catch a falling knife while riding a bull.

Professional managers often cannot just sit in cash. Their mandates chain them to equities. Clients pay them to be invested, not to hide under the desk. Even for individuals, selling because stocks “look pricey” can mean missing the last explosive leg up. Bubbles do not pop politely. They stretch further than reason allows. The market can stay irrational longer than you can stay solvent.

So I look for refuge. In the 1990s, bonds did the job. From early 1995 to the Nasdaq’s peak in March 2000, a Bloomberg index of American Treasuries rose nearly 50%. When stocks crashed, the same Treasury index rose another 30% as central banks slashed rates. Bonds were a shock absorber. Stocks fell, bonds rose. The classic 60/40 portfolio looked like genius. But 2022 rewrote the script. Inflation surged to 9.1% in June 2022, the highest in 40 years. The Federal Reserve hiked rates aggressively. Stocks fell. Bonds fell. The S&P 500 dropped 19% that year. Long-term Treasuries suffered one of their worst drawdowns in decades. The supposed hedge failed. Both sides of the portfolio burned at once.

If inflation resurges again, bonds may not save us. If governments’ debts—now exceeding 100% of GDP in many rich countries—spark fears of fiscal strain, both stocks and bonds could land in the same blast radius. We saw a glimpse of that when policy shocks rattled markets and Treasuries briefly lost their safe-haven glow. When trust cracks, correlation spikes.

Gold? It has swung wildly. Bitcoin? It calls itself digital gold, but in 2022 it fell more than 60%. When liquidity dries up, even the rebels beg for mercy.

That leaves derivatives. Options. The financial world’s insurance contracts. Buy a put option on the S&P 500 and you gain the right to sell at a pre-set strike price. Own the stock and the put, and you cap your losses. Today, a 1-year put limiting losses to 10% costs about 3.6% of the amount insured. Pay 3.6% and you sleep better.

But insurance is never free. Analysts at Goldman Sachs studied how such strategies would have fared from 1996 to 2002. A series of 1-year puts limiting losses to 10% produced roughly the same annualised return as unhedged stocks, but with less volatility. Not bad. A series of 1-month puts limiting losses to 4% did worse, because the repeated premiums piled up like credit-card interest. Protection can quietly eat performance.

I think of it like paying for flood insurance every month when the river never rises—until it does. You resent the premiums, then you bless them.

The problem is timing and cost. If volatility is already high, options get expensive. If everyone smells danger, insurance sellers raise prices. Buying puts after the crash starts is like shopping for fire extinguishers while your kitchen burns.

So I circle back to an idea that feels almost insulting in its simplicity: hedge equities with equities. During the dotcom bust, filtered baskets of steadier stocks outperformed. The S&P 500 low-volatility index, which holds the 100 least volatile stocks in the main index, proved a surprisingly effective diversifier. A 50/50 split between it and the broader S&P 500 from 1996 to 2002 generated nearly twice the annualised excess return over cash compared with the S&P 500 alone. Dividend aristocrats—companies that have increased dividends for at least 25 consecutive years—also delivered similar resilience. Quality stocks with high returns on equity and low debt did the same.

In other words, the best hedge was not fleeing the market but tilting within it.

That feels unsatisfying when AI stocks dominate headlines and valuations. Nvidia’s market capitalisation has at times surpassed $2trn. A handful of mega-cap tech firms account for a large share of the S&P 500’s gains. Concentration risk is real. When leadership narrows, fragility grows.

But if I step back, I see that not all stocks are the same. A cash-generating consumer-staples firm with stable earnings is not a venture-backed AI dream burning capital. A healthcare company with steady demand is not a speculative chip designer priced for perfection. If AI spending disappoints, if $660bn fails to produce commensurate profits, the high-fliers will suffer first.

The dotcom crash offers a brutal lesson. Cisco and Intel survived but traded sideways for years. Pets.com vanished. The Nasdaq fell nearly 78% from peak to trough. Yet diversified, profitable companies endured. Investors who owned quality and held on were scarred but not destroyed.

So how do I hedge a bubble, AI edition? I accept that there is no perfect shield. Bonds may not rescue me. Gold may whipsaw. Bitcoin may implode. Options cost money and discipline. Selling everything may mean missing the last manic surge.

Instead, I think in layers. I trim exposure to the most euphoric names. I tilt toward quality, dividends, lower volatility. I consider selective option protection, knowing it will drag on returns. I keep some cash—not because I am scared, but because liquidity is power when markets panic.

Above all, I remind myself that bubbles are stories we tell each other until the bill arrives. In 2000, it was eyeballs and clicks. In 2007, it was house prices that “never fall nationwide.” In 2026, it is AI that will change everything. Maybe it will. But prices can outrun reality.

Protecting a portfolio from a crash looks harder than ever because the old playbook is fraying. Stocks and bonds can fall together. Safe havens wobble. Insurance is costly. The hedge is imperfect. But I would rather carry an imperfect shield than march naked into a storm. In markets, survival is victory. Returns are optional. Staying in the game is not.

 

On a different but equally important note, readers who enjoy thoughtful analysis may also find the titles in my  “Brief Book Series” worth exploring. You can also read them here on Google Play: Brief BookSeries.