For the third time since artificial intelligence reshaped the investment landscape in late 2022, Asian technology stocks have shed a quarter to a third of their value — and for the third time, the question of whether the cycle has exhausted itself hangs in the air. J.P. Morgan, surveying the wreckage with measured calm, argues that price and reality have diverged: the underlying architecture of AI demand, from frontier model development to hyperscaler infrastructure commitments, remains structurally sound through at least 2027. What the market is pricing as an ending, the bank reads as a pause
J.P. Morgan: Asia Tech Sell-Off Won't Derail AI Investment Cycle
The market is pricing in a downturn that is unlikely to occur.
Mark
Why does J.P. Morgan think this sell-off is different from a real bear market?
Mimi
Because the things that made AI spending necessary haven't changed. The models are still improving, companies still need to run them, and the hyperscalers haven't signaled they're going to stop building.
Mark
But 25 to 30 percent is a real correction. Doesn't that suggest something fundamental has broken?
Mimi
It's the third time this has happened since 2022. The market gets nervous, prices fall, but the underlying demand persists. The bank sees this as fear, not evidence.
Mark
What about memory chips? That seems like a real problem if Nvidia and AMD are using less of them.
Mimi
It is a problem for the memory narrative—the story that AI demand would be insatiable. But the fundamentals are still solid. Supply will remain tight for years. It's more that the market got ahead of itself on how much memory would be needed.
Mark
So where should investors actually be looking?
Mimi
Equipment makers and packaging companies. The infrastructure has to be built, and those are the picks and shovels. The real constraint coming isn't chips—it's power.
Mark
Power? That's a different kind of problem entirely.
Mimi
Exactly. You can make more chips, but you need electricity to run them. In eighteen months, that might be what stops the whole thing from scaling further.
Il Polso
- Asian tech stocks have collapsed 25–30% in a single summer, dragging chip shares down with them and forcing investors to confront whether the AI investment thesis has finally broken.
- The sell-off carries the weight of repetition — this is the third major drawdown since the AI cycle began — amplifying fears that each recovery is weaker and each correction a step closer to a reckoning.
- J.P. Morgan is pushing back hard, pointing to uninterrupted hyperscaler spending commitments, accelerating frontier model development, and real-world agentic AI adoption as evidence that the fundamentals have not moved with the prices.
- The clearest opportunities, the bank argues, lie in semiconductor equipment makers and advanced packaging technologies, while memory stocks face a more tangled story — sound in fundamentals but wounded by Nvidia and AMD quietly reducing memory content in next-generation products.
- Beneath the near-term debate, a slower and more consequential constraint is taking shape: within two years, the availability of power — not chips — may become the true ceiling on how far AI infrastructure can grow.
For the third time since artificial intelligence reshaped the investment landscape in late 2022, Asian technology stocks have shed a quarter to a third of their value — and for the third time, the question of whether the cycle has exhausted itself hangs in the air. J.P. Morgan, surveying the wreckage with measured calm, argues that price and reality have diverged: the underlying architecture of AI demand, from frontier model development to hyperscaler infrastructure commitments, remains structurally sound through at least 2027. What the market is pricing as an ending, the bank reads as a pause — though it notes that the next constraint on this era of computation may not be silicon, but the electricity required to power it.
It has been a punishing summer for Asian technology stocks, which have lost between a quarter and a third of their value — the third such collapse since artificial intelligence became the defining investment story of the decade. Chip stocks have fallen alongside them, and the question investors are now asking aloud is whether the AI boom has finally reached its limit.
J.P. Morgan's answer, issued in a research note this week, is an unambiguous no. The bank's analysts concede the severity of the selling but insist the underlying drivers of the AI cycle are undisturbed. Frontier models continue to improve on a rolling basis. Inference demand — the computational work of running those models in the real world — remains strong. Agentic AI, capable of acting autonomously, is finding genuine commercial footing. And the hyperscalers — Amazon, Google, Microsoft, Meta — are showing no inclination to slow their infrastructure buildout, with J.P. Morgan expecting that commitment to hold through 2027, financed by equity and debt markets if necessary.
Within the semiconductor supply chain, the bank sees the clearest upside in equipment manufacturers, whose tools will be in high demand as wafer fabrication spending accelerates. Advanced packaging technologies — particularly 2.5D configurations and TSMC's coming 3D stacking investment cycle — are also flagged as strong performers, with IC substrates standing out among components.
Memory tells a more complicated story. Supply and demand look favorable for the next two to three years, yet recent decisions by Nvidia and AMD to reduce memory content in their next-generation AI products have unsettled the narrative that AI demand is indifferent to cost. J.P. Morgan calls the current market story on memory 'problematic' even while affirming the fundamentals, and expects a partial recovery over the next six months — though not a return to May's peaks.
Looking further out, the bank identifies a constraint that may prove more consequential than any chip shortage: power. As semiconductor capacity expands over the next eighteen to twenty-four months, the energy required to run AI systems at scale could become the binding limit on how far this infrastructure era can actually go.
The stock market has been brutal to Asian technology companies this summer. Shares have tumbled between a quarter and a third of their value—the third time this has happened since artificial intelligence became the dominant investment thesis in late 2022. Chip stocks have followed them down. For many investors, the question has become unavoidable: Is the AI boom finally running out of steam?
J.P. Morgan's answer, delivered in a research note this week, is no. The bank's analysts acknowledge the selling pressure but argue that beneath the price action, the fundamentals that have driven the AI cycle remain intact. They see no evidence of a slowdown in the next six to twelve months. The worry, they suggest, is overdone—a market pricing in a downturn that is unlikely to materialize.
The reasoning rests on several pillars. Frontier AI models continue to improve every few months. Demand for AI inference—the computational work of running trained models—remains robust across both proprietary systems and open-source alternatives. Profitability is spreading through the AI ecosystem as agentic AI, which can perform tasks autonomously, gains real-world traction. Most crucially, the hyperscalers—Amazon, Google, Microsoft, Meta—show no signs of pulling back on their aggressive spending on AI infrastructure. J.P. Morgan expects them to maintain these investments through 2027, tapping equity and debt markets as needed to fund the expansion.
The bank's analysis of the semiconductor supply chain reveals where the real opportunity lies. Equipment manufacturers—the companies that build the tools to make chips—are positioned as the strongest performers over the next year as wafer fabrication spending accelerates. Packaging and testing companies should see sharp growth as the industry moves toward 2.5D packaging, which places multiple chips side by side, and as Taiwan Semiconductor Manufacturing Company begins its 3D packaging investment cycle, where chips are stacked vertically to improve speed and efficiency. Among components, IC substrates emerge as the most promising.
Memory chips present a more complicated picture. Supply-demand fundamentals remain solid, and the bank expects demand to outpace supply for the next two to three years. Yet recent moves by Nvidia and AMD to reduce memory content in their next-generation AI products have punctured the prevailing narrative that AI-driven memory demand is immune to price sensitivity. J.P. Morgan's analysts call the current market story on memory "problematic, even though the fundamentals are sound." They expect memory stocks could rebound over the next six months but are unlikely to reach the highs seen in May.
As the industry optimizes for efficiency, a new constraint is emerging: interconnect—the technology that links chips together. But the bank's longer view suggests an even more fundamental bottleneck ahead. Within eighteen to twenty-four months, power availability may overtake semiconductor capacity as the primary limiting factor on AI infrastructure expansion. As chip production ramps up, the energy required to run these systems could become the real ceiling on growth.
Citazioni salienti
We do not see any fundamental indicators that signal meaningful weakness in the next 6-12 months— J.P. Morgan analysts
Market narrative on Memory is problematic, even though the fundamentals are sound— J.P. Morgan analysts
Vuoi la storia completa?
Leggi l'originale su cnbc.com ↗
Technology