The memory sell-off is the sharpest sign yet that investors are pricing a peak in the AI capex cycle before the earnings data confirms it. That keeps pressure on chip-heavy indexes such as the Kospi and on semiconductor names globally. JPMorgan’s view that earnings are protected for six months gives value buyers a case for stepping in. Goldman’s credit-spread warning, though, suggests the funding question could keep sentiment fragile, especially with tonight’s Nvidia session adding to headline risk . Further widening in AI-linked credit spreads would be a signal that the equity downside isn’t finished.
—
Memory stocks have already halved from their peak on fears that the AI spending boom is cresting, yet JPMorgan says the earnings downturn the market is bracing for is still at least six months away.
Summary:
Goldman Sachs’ AI memory stock basket has fallen 51% from its June peak, its deepest drawdown since ChatGPT launched.
The basket had more than quadrupled in the year to that peak, driven by AI data centre demand and heavy positioning.
Goldman had earlier warned that crowded positioning left memory stocks exposed to further liquidation.
A Goldman strategist says AI cloud provider credit spreads are at their widest this year and that AI capex has likely peaked.
JPMorgan says markets are pricing a memory downcycle as early as the first half of 2027.
JPMorgan expects no earnings cuts over the next six months, citing long-term agreements and prepayments.
A Goldman Sachs basket of AI memory-related stocks has fallen 51% from its June peak, its deepest drawdown since ChatGPT launched in late 2022, Bloomberg reported. The market is moving to price in a downturn that JPMorgan argues has yet to show up in earnings.
The slide marks a dramatic reversal for one of the hottest trades of the AI boom. Goldman’s memory basket had more than quadrupled in the year to its late-June peak, powered by demand for the high-bandwidth memory used in AI data centres and by heavy positioning from fast-money and institutional investors alike. Stocks in the space include Samsung Electronics, SK Hynix, Micron, Kioxia and Western Digital.
Goldman had already warned earlier in the selloff, when the basket was down around a quarter from its high, that crowded positioning in memory stocks left the sector exposed to further liquidation. It noted then that asset managers’ long positions remained close to record levels. The deeper drop suggests much of that unwinding has since taken place.
The bank’s wider read on AI is also more cautious. One of its strategists has noted that credit spreads for the largest AI cloud providers have widened to their highest levels this year, and argued that equity prices have not fully caught up. That implies either that bond markets are too gloomy or that stocks have further to fall. The same strategist said AI capital spending has likely peaked, shifting investor attention from earnings upgrades to financing and returns on investment.
JPMorgan offers a more measured view on the fundamentals. In a recent note on the memory sector, the bank said order visibility is extending. It said the market appears to be anticipating a memory downcycle driven by AI funding and power constraints as early as the first half of 2027. However, it does not expect earnings cuts over the next six months, pointing to long-term supply agreements and customer prepayments as a buffer.
Put together, the two banks frame the debate clearly. Share prices are already discounting a turn in the cycle, while near-term earnings, on JPMorgan’s view, remain protected. Whether the selloff has gone too far depends on how quickly AI investment slows and how it is financed.
The pressure on memory comes alongside broader strain on the AI trade, from questions over the scale of off-balance-sheet commitments at large technology companies to the governance overhang at Nvidia. For investors, the next signposts are hyperscaler spending plans, memory pricing and contract data, and whether credit spreads for AI-linked issuers keep widening.
—
Graphic via Bloomberg:
This article was written by Eamonn Sheridan at investinglive.com.
