If today’s AI bubble bursts, it won’t be a pretty sight

On 8 October, the BoE issued a warning that could have come straight from Alan Greenspan’s “irrational exuberance” era, cautioning that AI company valuations are dangerously “stretched.” That’s central banker speak for: ‘If you squint at the fundamentals, you might eventually see the bottom.’

Meanwhile, JPMorgan’s CEO Dimon, never one to sugarcoat a storm, has indicated that the probability of a bust is probably three times what has been priced in. Naturally, the market ignores this good advice and continues to levitate in a speculative updraft.

But the concern isn’t just about lofty valuations. It’s the increasingly creative ways they are being justified. Nvidia’s financial relationship with OpenAI—where it invests in OpenAI, which then uses that money to buy Nvidia’s GPUs—is a classic example of circular revenue. It’s reminiscent of the dot-com era, when startups bought ads from one another and called it ‘growth.’ Today’s moves are fancier and involve more silicon.

Oracle’s recent earnings added weight to the growing scepticism. Its AI profits fell well below expectations and the stock dropped 7% on 7 October. The reaction suggests that while investors are still enamoured of AI, their tolerance for earning shortfalls is thinning. The music may still be playing, but the crowd may be looking for exits.

Blankfein, former CEO of Goldman Sachs, predicts a market correction within 12 to 24 months. That’s investment banker speak for ‘soon enough to matter but not panic.’ The IMF has warned that AI-driven capital investment overheats markets and distorts returns. For an institution that usually reserves its harsh language for developing economies, that’s a big deal. Even cautious referees are reaching for the yellow card.

So, where does this leave the thoughtful investor trying to participate in the AI boom without getting trampled when the mood shifts? First, let’s acknowledge that this AI wave is real. The core technology isn’t vapourware. Unlike the late 90s, when companies with little more than a website and a Super Bowl ad drew billions in valuation, today’s AI giants have actual products, revenues and in some cases, profits.

Nvidia makes physical GPUs—expensive hot-selling hardware needed to train AI models. OpenAI’s AI tools, for all their opacity, are being used by hundreds of millions of people globally.

However, the line between a ‘growth’ stock and an ‘overpriced’ one is unclear. Expectations define it, and those are sky-high. When Oracle’s numbers came in soft, investors didn’t stop to ask whether its long-term strategy was intact. They sold. That’s a hallmark of a speculative market: Everyone’s in—until everyone’s out.

Investors should look for companies with diversified revenue streams. Microsoft’s AI push, for instance, is significant, but it’s supported by Azure cloud infrastructure, MS Office subscriptions and enterprise software that remain sticky. Amazon is deploying AI in logistics and customer-facing services, but AI isn’t its whole business. These firms can pivot, rebalance or ride out a downturn. You’d want exposure to AI without betting your entire portfolio on it.

Don’t forget that many long-term returns from AI may not come through owning AI companies themselves, but by owning firms that use AI well. A decade from now, the real beneficiaries might be manufacturers that automated their operations, logistics companies that optimized their supply chains or insurers that improved risk modelling. The internet’s biggest business winners weren’t always dotcoms, but brick-and-mortar firms that figured out how to harness it. So the smart money might lie in watching how AI is used, not just hyped.

Emotions must be kept in check too. Bull runs fuel optimism, which makes everyone feel like a genius. When you see a stock trading at 70 times forward earnings and find yourself justifying it with “but it’s the future,” take a breath. Even transformative technologies have messy, uneven paths to profitability. Remember that tulips, railroad stocks, radio companies and internet startups were all the ‘future,’ too, until the market turned realistic.

That doesn’t mean you abandon AI. You should approach it with the same scepticism and discipline you’d apply to any other asset. Invest in firms with real earnings, stable leadership and a clear understanding of how AI fits into a broader strategy. Diversify your bets, watch for hype and be cautious if a company treats revenue like a magic trick.

AI will reshape entire industries. But tech revolutions don’t always produce good quarterly earnings. Sometimes, they produce higher productivity across the board—and that value shows up quietly. So, admire the AI revolution. Track it, learn about it, even invest in it. But don’t build your retirement plan around it just yet. Understand the tech, assess business models and ignore the noise. While algorithms may now predict weather patterns, how proteins fold and your next shopping impulse, they still can’t save investors from themselves.

The warning signs are already here. You don’t need to panic, but you do need to prepare. There’s no shame in sitting out the AI frenzy—or at least sitting near the exit. Just make sure that when the music stops, you’re not left on the dance floor.

The author is co-founder of Siana Capital, a venture fund manager.