Has the AI boom turned into a bubble?

Investors are becoming increasingly uneasy that we may be in the middle of a stock market bubble driven by soaring enthusiasm for artificial intelligence, or AI. Do investors have good reason to be nervous?

Over the past three months, Google searches for “AI bubble” have surged, reflecting growing investor anxiety that prices of many key AI stocks have risen too far, too fast. 

These worries are being driven by two key factors. 

First, stock valuations—especially among large-cap growth companies—have risen to levels reminiscent of the late 1990s. That period, of course, culminated in a tech bubble driven by surging expectations for the commercial possibilities of the internet.

Second, the hyperscalers—the five mega-cap tech giants that are investing heavily in AI infrastructure— have dramatically expanded their capital spending plans. This has raised questions about whether these companies can maintain profitability while making such substantial outlays. 

But context is crucial: Much of the valuation expansion in AI winners has been matched by equally strong earnings growth. Current profitability levels go a long way toward supporting these valuations—assuming they can be maintained. 

This is a very different backdrop from the late 1990s, when valuations rose sharply without a corresponding improvement in return on equity. By contrast, the current cycle has been marked by a long, steady rise in return on equity, or ROE, which has supported a gradual rise in valuations. 

This marks an important distinction between the internet bubble and the current enthusiasm for AI. In the late 1990s, investors were betting that emerging internet companies eventually would become profitable once the technology matured. Today’s investors are betting that already highly profitable companies will remain so as AI adoption accelerates. This is a fundamentally different proposition.

Investor concerns about the current scale of AI capital expenditures are somewhat more grounded but again should be viewed in the appropriate context. 

Projected AI-related capex indeed has climbed rapidly. In aggregate, hyperscaler capex budgets now exceed their projected free cash flow—meaning these companies increasingly will need to rely on debt to fund their buildouts. That is new territory for this group. But it does not necessarily signal an unsustainable investment boom. 

With the notable exception of Oracle, the five major hyperscalers all carry very low debt loads relative to other large-cap companies. 

As of late November, the debt-to-equity ratio for the Russell 1000 Index (weighted by market cap) was 83%, but Amazon, Microsoft, Alphabet, and Meta each carried less than half that amount. 

Meanwhile, there is little evidence that credit markets are sounding alarm bells. Spreads on credit default swaps for the hyperscalers—once again excluding Oracle—remain broadly in line with the overall investment-grade corporate bond market, as measured by the CDX North American Investment Grade Index.

While AI is a transformative technology, the rapid gains in AI-exposed equities have sparked understandable concerns about valuation and capex sustainability. However, while these risks are real, the underlying fundamentals look far more balanced than during the peak of the late-1990s tech bubble.

Given this backdrop, the T. Rowe Price Asset Allocation Committee is maintaining a neutral stance between U.S. growth and value equities. AI’s economic potential is far too large to ignore, but it would also be unwise to build equity portfolios solely around this theme. We believe a balanced style position should allow investors to participate in AI’s promise while avoiding overexposure if the current enthusiasm proves excessive.