A sharp divergence between the Dow and Nasdaq shows the artificial-intelligence trade is maturing, forcing investors to separate durable earnings power from inflated expectations.
The most important market signal this week was not that artificial intelligence stocks stumbled. It was that the broader market did not fall with them. That distinction matters because it suggests investors are no longer treating AI as a single, unstoppable force that lifts every chipmaker, cloud provider and software name in its orbit. The trade is becoming more selective, more demanding and, ultimately, healthier.
The latest test came as Broadcom (AVGO) sold off sharply despite continuing to benefit from AI-related demand. The market’s reaction was less about a collapse in the company’s fundamentals than about the gap between strong results and even stronger expectations. That gap has become the defining risk for many technology leaders. When valuations assume years of near-flawless execution, even a good quarter can feel disappointing. Meanwhile, the Dow Jones Industrial Average surged to a record high while the Nasdaq lagged, a rotation that showed investors still want equities, just not at any price.
That is a useful correction in sentiment. For much of the past two years, the AI boom has compressed nuance. Investors have often treated data-center spending, semiconductor demand, cloud infrastructure, enterprise software and electric-grid investment as one large theme. In practice, they are different businesses with different margins, balance-sheet risks and competitive pressures. Nvidia (NVDA) remains central to the infrastructure buildout, but the presence of one extraordinary winner does not guarantee that every company touching AI will earn extraordinary returns.
The issue is not whether AI spending will continue. It almost certainly will. Major technology companies are still racing to secure computing capacity, and AI has become a strategic priority for corporations and governments. Morgan Stanley has estimated that nearly $3 trillion of AI-related infrastructure investment could move through the global economy by 2028, with most of that spending still ahead. That is a powerful long-term backdrop, but it also raises a harder question for investors: who captures the return on that spending, and who merely funds it?
Capital intensity is the uncomfortable side of the AI story. Building data centers, buying chips, securing power, training models and hiring scarce engineering talent require enormous upfront commitments. The largest platforms can absorb those costs because their balance sheets are deep and their core businesses throw off cash. Smaller companies, or companies with weaker pricing power, may find that AI raises revenue while pressuring free cash flow. In that environment, sales growth alone is no longer enough. Investors should care about operating leverage, customer concentration, depreciation schedules and the durability of demand once the initial buildout slows.
This is why the market’s reaction to Broadcom matters beyond one stock. A company can be well run, strategically placed and exposed to a powerful theme while still being vulnerable to an overextended multiple. That is not a contradiction. It is valuation discipline returning to a market that had temporarily forgotten it. A stock can represent a good business and a poor risk-reward trade at the same time.
The broader market’s resilience also suggests that investors are rediscovering the value of diversification. On Thursday, gains were broad enough to lift hundreds of S&P 500 constituents even as semiconductor shares weakened. Financials, health care and smaller companies attracted fresh buying, helped by lower bond yields and easing oil prices. That matters because a bull market dependent on a handful of mega-cap technology stocks is fragile. A market that can rotate leadership without collapsing is sturdier.
Still, the rotation should not be mistaken for an all-clear signal. The S&P 500 remains up meaningfully for the year, and valuations in several growth sectors leave little margin for error. The next catalyst is labor-market data, with investors watching whether job growth is cooling enough to support future rate cuts without signaling a deeper economic slowdown. Futures weakened ahead of the report, reflecting the market’s dependence on a narrow path: inflation must keep moderating, employment must soften only gradually, and earnings must continue to justify elevated prices.
That path is possible, but it is not guaranteed. The Federal Reserve still has limited room to reassure equity investors if inflation proves sticky. Lower oil prices help, but wage growth, services inflation and tariff effects can complicate the outlook. If rates stay higher for longer, long-duration growth stocks remain exposed because much of their value rests on profits expected far in the future. That does not invalidate the AI thesis, but it does make price discipline more important.
For portfolio managers and retail investors alike, the lesson is straightforward: own AI exposure, but do not outsource judgment to the theme. The better approach is to distinguish between companies selling scarce inputs, companies integrating AI into existing products, and companies merely attaching AI language to ordinary business models. The first group can earn exceptional margins when supply is tight. The second can create durable value if AI improves productivity or customer retention. The third group is where enthusiasm is most likely to destroy capital.
Nvidia remains the clearest example of the first group, though its valuation already reflects enormous success. Microsoft (MSFT), Alphabet (GOOGL) and Amazon.com (AMZN) represent the second group, where AI may deepen competitive moats but also demands heavy capital expenditure. Broadcom sits somewhere between infrastructure enabler and expectation-risk test case. That complexity is precisely why blanket optimism is no longer sufficient.
A maturing bull market does not require the AI trade to end. It requires investors to stop pretending every AI-adjacent company deserves the same premium. The strongest markets are not the ones where every narrative is accepted. They are the ones where capital becomes more discriminating.
That is what this week’s split between the Dow and Nasdaq appears to show. Investors are not abandoning growth. They are demanding proof. After a long rally fueled by promise, that may be exactly what the market needs.
