The AI market is no longer treating every company like they’re all on the same level. Things started coming apart in the final stretch of 2025, when tech stocks were being tossed around like dice.

Sell-offs, wild rallies, overpriced deals, and mountains of debt raised the alarm on whether this AI wave is turning into something more fragile, like a bubble. And now, it’s not just about who’s talking about AI, it’s about who’s getting paid and who’s footing the bill, according to CNBC.

Stephen Yiu, who runs money over at Blue Whale Growth Fund, said this market has never cared much about separating winners from the ones just throwing money around.“Every company seems to be winning,” Stephen said.“But it’s very important to differentiate,” he added, warning that this is exactly what investors might finally start doing.

He pointed out that people, especially retail investors using ETFs, don’t bother looking under the hood. Whether it’s a startup, a firm bleeding cash to fund AI infrastructure, or someone just collecting the checks, everyone’s been lumped into the same pile. That may not last.

Big Tech dumps billions into AI infrastructure as models evolve

Stephen broke things down into three clear categories: private AI startups, public companies spending heavily on AI, and firms supplying the infrastructure. The first camp, companies like OpenAI and Anthropic, got $176.5 billion in funding in just the first nine months of 2025, based on PitchBook data.

Meanwhile, giants like Amazon, Microsoft, and Meta are cutting massive checks to infrastructure vendors like Nvidia and Broadcom.

Stephen’s fund doesn’t just look at hype. They judge value by checking how much free cash a company has after capital spending, compared to its stock price. The problem? Most of the Magnificent 7 stocks are now “trading at a significant premium,” Stephen said.

They’ve gone deep into AI spending, and their numbers look bloated as a result. He said he wouldn’t touch AI spenders right now, even if he believes in the future of the technology. His focus is on the firms that are getting paid, not the ones blowing money to chase future returns.

Julien Lafargue, who leads market strategy at Barclays Private Bank and Wealth Management, said all this froth isn’t everywhere. It’s concentrated in “specific segments rather than across the broader market.” And the real trouble? It’s the companies riding the AI buzz with no revenue to show for it.

Julien singled out “some quantum computing-related companies” as examples where the hype train has left the station with no earnings in sight. “Investor positioning seems driven more by optimism than by tangible results,” he said. “Differentiation is key.”

Rising asset costs complicate business models for AI spenders

The AI market shake-up also exposes how the biggest names are changing. Big Tech, once proud of being asset-light, is now buying land, building data centers, and gobbling up GPUs.

Companies like Google and Meta are no longer software firms, they’re hyperscalers with physical overhead. That shift doesn’t just eat cash. It changes how investors should look at them entirely.

Dorian Carrell, who oversees multi-asset income at Schroders, said old methods of valuing these companies don’t cut it anymore. “We’re not saying it’s not going to work,” Dorian said. “But we are saying, should you pay such a high multiple with such high growth expectations baked in?”

To keep the AI build-out running, tech firms went to the debt markets this year. Meta and Amazon both tapped that route, but according to Ben Barringer, tech research head at Quilter Cheviot, they’re still in a net cash position.

That’s very different from companies barely holding it together. “The private debt markets will be very interesting next year,” Dorian added.

Stephen warned that unless AI-driven revenue outpaces what’s being spent, profit margins will tighten. And investors will start asking harder questions.

Infrastructure and hardware don’t last forever. They wear down. The cost of that wear and tear isn’t yet visible in profit and loss statements. “It’s not part of the P&L yet,” Stephen said. “Next year onwards, gradually, it will confound the numbers. So, there’s going to be more and more differentiation.”

If you're reading this, you’re already ahead. Stay there with our newsletter.