The AI Trade Roars Back to Record Highs

The AI trade was supposed to be broken. Six weeks ago investors fled the story, convinced that spending had outrun demand and that the great buildout would end in tears. Instead the market has answered with a run of fresh records, dragging the Dow and the S&P 500 to all-time highs and pulling Micron back above a trillion dollars. The question now is not whether the AI trade survived the summer, but whether anything can stop it.

By Blumefield | August 10, 2026

The Rebound That Silenced the Skeptics

Markets rarely reverse this fast. In the middle of July, the mood around artificial intelligence stocks was close to funereal. A wave of warnings about excess data center capacity, a brutal sell-off in memory names, and a Chinese open model that undercut Western pricing had convinced traders that the AI trade had finally met its reckoning. The Philadelphia Semiconductor Index shed a fifth of its value, and the phrase "compute glut" moved from analyst notes into dinner conversation.

Four trading sessions in early August erased more than half of that decline. By the close on August 9, the S&P 500 sat at another record, one of roughly two dozen all-time highs the index has printed this year, while the Dow joined it at fresh peaks. Micron climbed almost 8 percent in a single day and returned to a trillion-dollar valuation, and the semiconductor index jumped more than 6 percent as Marvell, SanDisk and SK Hynix led the charge.

The reversal was not built on sentiment alone. It was built on numbers, delivered in the space of a fortnight by the companies at the center of the AI trade. When Microsoft, Amazon and Alphabet each confirmed that capital spending was still accelerating rather than cooling, the bear case lost its foundation. The buildout, it turned out, was not slowing. It was compounding.

What the July Panic Got Wrong

The summer scare rested on a plausible fear. If hyperscalers were pouring hundreds of billions into chips and buildings, and if the revenue from AI products lagged that spending, then somewhere a painful correction had to arrive. That logic drove the drawdown, and for a few weeks it looked prophetic.

The earnings season demolished it. Rather than trimming forecasts, the largest cloud operators raised them, pointing to enterprise adoption that was pulling AI capacity into use faster than it could be installed. Combined capital expenditure commitments across the big platforms pushed well past two hundred billion dollars for the year, and executives framed that figure as a floor rather than a ceiling. The demand signal that skeptics said was missing was hiding in plain sight, inside backlog numbers that stretched years into the future.

The hard industry data told the same story. The Semiconductor Industry Association reported that global chip sales reached 403.3 billion dollars in the second quarter, a 35.1 percent jump from the first quarter, with June sales alone up more than 123 percent year on year. That is not the shape of a market drowning in unsold inventory. It is the shape of an industry that cannot build fast enough, and it gave the rally the empirical backbone it lacked in July. You can read the association's full quarterly release through the Semiconductor Industry Association.

A Market Balanced on Two Companies

There is a less comfortable truth beneath the celebration. The AI trade has become extraordinarily concentrated, and the record highs rest on a very small number of shoulders. Goldman Sachs estimates that Nvidia and Micron alone will account for roughly a third of all S&P 500 earnings growth this year, an astonishing share for two companies to carry in an index of five hundred.

Widen the lens to the full AI infrastructure complex, and the concentration grows starker still. The bank projects that companies tied to the buildout, from chipmakers to hardware vendors to the utilities racing to power new campuses, will drive close to half of the index's earnings expansion in both 2026 and 2027. When performance leans this heavily on a handful of names, the line between a healthy bull market and a fragile one becomes hard to see.

That fragility cuts both ways. The same concentration that makes the AI trade look precarious is also what has powered the year's returns, lifting the S&P more than 13 percent on a total-return basis through early August, one of the strongest starts in more than three decades. Investors are being asked to hold two ideas at once. The rally is real and grounded in genuine earnings, and it is also unusually dependent on a cluster of companies that would all suffer together if the buildout ever stalls.

The Memory Supercycle Underneath It All

If Nvidia has been the face of the AI trade, memory has quietly become its engine room. Micron's surge past a trillion dollars reflects a shift that few appreciated a year ago, that the scarcest resource in modern AI systems is not raw processing but high-bandwidth memory, the stacked chips that feed data to accelerators fast enough to keep them busy.

The economics are striking. A recent study from the industry association and Deloitte found that semiconductors account for roughly 95 percent of the value of an AI server rack, and estimated that annual revenue from chips deployed in AI data centers could exceed 1.2 trillion dollars by 2028, a nearly tenfold increase in four years. Memory sits at the heart of that expansion, and the pricing power it has handed to Micron, SK Hynix and Samsung is the clearest evidence yet that the boom runs on far more than graphics processors. The underlying report is available from the Semiconductor Industry Association.

For companies like Micron, the supercycle has rewritten the investment case. A business long treated as a cyclical also-ran, prone to boom and bust, is now valued as a structural winner of the AI era. That reappraisal is precisely what made the July sell-off so violent and the August recovery so sharp. When the market doubts memory, it doubts the whole edifice. When it believes, the AI trade flies.

Whether the AI Trade Can Keep Climbing

The bulls have momentum and hard data on their side, but the risks have not vanished, they have simply been deferred. Goldman itself flagged the danger that lurks beyond the horizon, the rising tide of depreciation from all that capital spending. Every dollar poured into chips and buildings eventually becomes a cost on the income statement, and by 2027 that drag could bite into the very profit growth now propelling the AI trade higher.

There are nearer risks too. A market this concentrated is exposed to any single disappointment from Nvidia or Micron, and the geopolitical backdrop remains volatile, with export controls, supply chain shocks and Middle East tensions all capable of rattling sentiment in an afternoon. The compute glut fear that gripped July was overdone, but it was not baseless, and it could return the moment capacity growth visibly outruns demand.

For now, the weight of evidence favors the optimists. Record chip sales, accelerating hyperscaler spending and a memory shortage with no quick fix all point in the same direction, and the market has repriced accordingly. The lesson of the past six weeks is that the AI trade is more durable than its critics believed and more concentrated than its champions like to admit. Both things can be true, and investors who forget either half will eventually be reminded. For continuing coverage of the buildout and the markets it is reshaping, follow Blumefield.