With many recent news headlines dishing out growing trepidation about a possible stock market bubble, it’s hard to believe that just six months ago, investors were instead fearful that Washington’s policy missteps might push the economy into recession. This dramatic shift in sentiment highlights both the resilience of the U.S. economy and the extraordinary power of spending on artificial intelligence infrastructure. With GDP growth running close to 4% for the past two quarters, this year’s earlier worries about tariffs and softening labor markets have now given way to concerns that dot-com type ghosts might reappear and spoil the party.

While it’s true that today’s market looks expensive relative to recent history, it’s worth recalling just how extreme valuations became during the dot-com era. In 2000, the leading technology companies—Microsoft, Cisco Systems, Intel, Oracle, IBM, Lucent, and Nortel Networks—traded at an average forward P/E ratio of 52x. By comparison, today’s “Magnificent Seven”—Nvidia, Microsoft, Apple, Alphabet, Meta Platforms, Amazon, and Tesla—trade at about half that level, with an average forward P/E near 27x. From this perspective, current valuations are elevated but not reminiscent of the excesses of 2000.

Importantly, today’s market leaders are far more mature and profitable than those frothy predecessors a generation ago. In 2000, the combined market capitalization of the top growth companies was roughly $2.4 trillion. Today, the Magnificent Seven together command an astonishing $20.5 trillion, underscoring how dramatically these firms have expanded in scale, earnings power, and global influence.

Much of their price performance has been driven by sustained, powerful profit growth rather than by speculative hopes for future earnings. Plus, despite current valuations being above long-term averages, returns on equity are also far higher, suggesting that richer multiples may be justified. Moreover, the billions now being poured into AI infrastructure are largely being funded with cash rather than borrowed capital. Even if returns on these investments ultimately fall short of expectations, investors might simply face disappointment—but not distress. And while the magnitude of AI spending can appear staggering, it remains modest in historical context: AI-related investment is still under 1% of U.S. GDP, compared with peaks of 1.5–2% during past innovation waves such as electrification in the 1920s or the dot-com expansion of the late 1990s. In other words, the current cycle may still have room to run.

And yet, there might be one place where goblins are hiding in the closet. The private market is showing signs of extreme exuberance, and this is causing legitimate fright for those who lived and invested through the dot-com era. In 2000, the term “unicorn” hadn’t even been coined, and nearly every billion-dollar company was publicly traded. Since then, though, a combination of factors has dramatically expanded the private market: the rise of large private-equity pools, the added cost and complexity of going public after Sarbanes-Oxley was introduced in 2002, the influx of capital from pension and sovereign funds, and the extended and unprecedented era of near-zero interest rates from 2008 to 2021. As a result, the number of private-market unicorns has soared in recent years to roughly 1,200–1,600 globally, representing as much as an estimated $5 trillion in total value. This is one of the most significant structural shifts in modern capital markets.

Today’s AI-focused private-market giants—OpenAI, xAI, Anthropic, Scale AI, and Perplexity—now collectively approach $1 trillion in valuation. That’s less than half the market value of the dot-com darlings in 2000, but still quite striking given that none of these companies are profitable. Goldman Sachs recently noted that bubbles tend to form when both valuations and stock prices together rise faster than the future cash flows that those firms are likely to generate. While AI overall could indeed yield meaningful gains in productivity and profits, our view of the private-market leaders in this space remains cautious.

Part of that caution stems from the increasingly circular nature of business relationships among major technology players and a resurgence of vendor financing. For instance, in September, Nvidia announced a $100 billion investment in OpenAI, for funds earmarked for building AI data centers using Nvidia’s chips. In exchange, Nvidia received an equity stake in OpenAI. During the dot-com era, telecom companies engaged in similar self-reinforcing arrangements—though then it was primarily through debt rather than equity—and those debt loads eventually became unsustainable, leading to widespread bankruptcies and years of overcapacity. Because today’s AI investments are largely funded by public companies with minimal leverage, that specific risk is limited for now. Still, if unprofitable private firms continue to pursue aggressive and increasingly creative financing strategies, the potential for imbalance will rise. Having said all of this, selective exposure to the private market through experienced and disciplined managers can be warranted, particularly in a period of such profound technological change.

There are, of course, other important differences between now and the late 1990s. Telecom companies of that era were not only financially overextended but, in some cases, tainted by outright fraud—with Enron and WorldCom being the most infamous examples. The Federal Reserve also significantly played a role in inflating and then pricking that bubble: it lowered interest rates during a period of near-5% GDP growth, only to tighten policy as the economy began to slow, triggering the collapse of a highly leveraged system. By contrast, if we view the launch of ChatGPT in 2022 as the effective start of the modern AI boom, the Fed was actually raising interest rates during that early phase and has only recently begun to ease again. That makes today’s general environment fundamentally different—though no less fascinating—as innovation, liquidity, and human optimism once again converge.

In our view, artificial intelligence has the potential to be even more transformational than the internet. However, even in times of remarkable innovation, there can be fits and starts and surprises—and sometimes even corrections, although, again, we do not believe we are on the verge of a bubble bursting—and hence, it’s important to remember the enduring value of diversification and discipline in a long-term investment strategy. As AI adoption broadens, we expect its benefits to extend well beyond the technology sector, potentially driving profitability and returns across a wide range of industries, and internationally too. Indeed, over time, those gains could rival or even exceed the value captured by the initial companies building today’s AI infrastructure.