How T. Rowe Price is actively navigating the AI bubble

As seen in the chart below, however, it’s not hard to find some similarities.

T. Rowe Price Portfolio Specialist Sam Ruiz says that what the graphs show is “very similar behaviour” between the tech bubble and the current AI landscape.

“Part of what we're doing here is navigating this AI cycle, where we think we should be more prudent, focus more on valuation, don't take risks in companies that aren't yet profitable. Part of that might mean giving up some absolute return on the upside, but we think it makes us a lot more defensive on the downside,” he says.

“We do believe that we’re in a bubble.”

So, should investors dump all of their stock in AI companies? Not exactly, but Ruiz argues that understanding the current market conditions is key to portfolio construction.

“We believe it's more important to identify that you're in a bubble and identify a framework of what will dynamically change your risk appetite and your positioning as that bubble matures,” he says.

“It’s much better in our view than taking a binary view of whether to sell out entirely or whether to back up a truck and still have enough of this risk in the portfolio. Both of those can be incredibly portfolio defining, and the volatility can be enormous.”

All that matters is capex

It’s no secret that the hyperscalers are ramping up their spend on AI compute capacity. This massive capex from the likes of Microsoft, Google, Amazon, Meta and Oracle has been the driving force behind the massive gains for semiconductor companies.

While the last few weeks have been rough for chipmakers - the Philadelphia Semiconductor Index is down around 25% from its peak last month - the amount of money flowing from the hyperscalers is almost comically large.

“The 2026 estimate was that these five companies would spend US$510 billion. Roll forward to January, that had already increased 26% to US$640 billion. We now expect that's probably going to be closer to US$800-$900 billion for 2026,” Ruiz says.

“So an enormous jump, and that could even grow further to US$1.5-$1.6 trillion in 2027 spent by five companies, basically all on AI capacity. If we roll back to 2020, only US$100 billion dollars. 

"That's changing the composition of the market, it's changing leadership of the market, and it's changing where profits are driven in the market because of this belief that this AI is going to generate a real return. That is the biggest debate right now.”

While there are a whole range of other factors - regionally, sector level, top down, bottom up - according to Ruiz: 

“The only thing that really matters right now is capex.”

Given the scale, it would be fair to think this massive build out would be reaching a breaking point, especially given the spending is increasingly coming from debt, the signals from the AI players is the opposite.

“During the quarter, we have met leaders of OpenAI, Anthropic, Meta, Amazon, Google, and all of them are telling us we would spend more if we could spend more,” Ruiz adds.

“Either if the supply chain could give us more components, we would be buying more, and capex would be higher. Or if we thought the market would tolerate more debt from us, we would be spending more.”

“That signals to us that they see returns for AI. It signals to us that AI capex will be here for longer. It signals to us that AI hardware stocks have more to benefit,” Ruiz says.

When does AI hit a breaking point?

According to Ruiz, the areas of the AI trade that are concerning mostly centres on leverage and geopolitics. As evidenced in wild swings seen in the South Korean KOSPI Index, which is extremely concentrated in semiconductor players SK Hynix and Samsung, leverage is exacerbating movements in the market.

“We're seeing short options volumes significantly increase. We're seeing a lot more levered products coming to market and getting demand. Some of the regulators in various countries are having to limit the multiple or the turns of leverage,” Ruiz explains.

“Right now, it's a one directional bet. That worries us and we are seeing the market become more volatile in response to how this leverage is playing in.”

On the geopolitical front, the instability around the US-Iran war and “China on top of that” has increased market risk. The ceasefire at the end of June helped reduce that risk, but increased hostilities over subsequent weeks has driven this higher.

Throw everything together and you might expect T. Rowe Price to be looking for the exits on the AI trade. However, that’s not the call just yet.

“The net takeaway from this is, for here and now, we think that the economy is growing, the market's growing, earnings expectations are being revised higher. We do see that AI is actually still in the sweet spot with more to spend, more leverage for these companies can be taken up,” Ruiz says.

“There will be a point where that gets too high, or that fork in the road moment of the monetisation. We're not seeing that yet, so we are still overweight AI hardware, we have a slight overweight to risk and beta in the portfolio, but we're being very cautious around how we position there, not on the speculative, profitless side of it. Also a lot more critical on valuation.

“We are going to lag when you see those 10% gap higher months where there's extremely high beta, momentum on, risk on, but we're still going to participate quite well and be more defensive on the downside.”