NextFin News - GAM chief investment officer Paul Markham says the global selloff in chip stocks is not yet a buying opportunity, and his warning goes beyond a short-term trading call. The question now is whether the market is repricing a temporary flush in a crowded trade or a more durable reset in how investors value the AI build-out, the companies that supply it, and the customers that pay for it.
The immediate problem is positioning. Markham said there are “so many people on the same side of the boat” that over-ownership leaves the sector vulnerable when sentiment turns. That matters because semiconductors have become one of the clearest expressions of the artificial-intelligence trade, with investors leaning on the idea that rising data-center spending, tight supply in advanced chips and years of AI demand growth can support both revenue and valuation expansion at the same time.
When that combination works, it is powerful. Revenue rises, margins widen and multiples can stay elevated because the market believes growth has a long runway. When it stops working, the same mechanism reverses quickly. A sector that was being valued as a structural winner can start to trade like a crowded factor exposure, where even a modest change in confidence can compress the multiple before earnings themselves weaken.
That is why the selloff is being read as more than a routine dip. Investors are no longer just asking whether AI spending continues. They are asking whether that spending still translates into enough profit growth to justify the prices attached to chipmakers and their suppliers. In other words, the market is moving from a story about demand to a test of economics.
The distinction is important because a cyclical correction and a structural repricing are not the same thing. A cyclical selloff usually comes from inventory swings, order pauses or short-term positioning pressure. Those moves often reverse once supply adjusts and buyers return. A structural reset is more stubborn. It happens when the market starts to question the durability of the earnings model itself, or the amount of capital required to sustain it. Markham’s warning sits closer to that second category, even if the current move still has a strong cyclical component.
What The Selloff Is Pricing
Markham’s view is not that AI demand has disappeared. It is that investors may have pulled forward too much good news into semiconductor valuations. The mechanism is straightforward. Hyperscalers and other large buyers have been pouring capital into data centers, accelerators, memory and networking gear. That spending lifts supplier revenue expectations, which can support richer multiples. But if the market starts to believe the payback is slower, thinner or less certain than expected, the whole chain weakens.
That is why the selloff is not just about the level of chip prices. It is about the duration of growth that those prices imply. A high valuation is easier to defend when the market believes a theme still has years of unsated demand ahead of it. It becomes harder when investors start asking whether the spending curve is normalizing, whether customers are becoming more disciplined and whether the next leg of AI investment will produce the same incremental profitability as the last one.
The second-order effect matters too. Semiconductor weakness does not stay confined to chip names. It feeds back into the broader AI stack: cloud builders, server vendors, networking suppliers, memory makers and even the hyperscalers whose capital-spending plans have become the key demand signal for the sector. If investors decide that AI infrastructure is still growing but doing so with weaker economic payback, then the market must reprice not only the suppliers but also the customers that have been using AI spending to justify their own budgets.
That is the real transmission chain. The first-order move is a fall in chip prices. The second-order move is a reset in confidence about AI return on capital. The third-order move is a broader change in how investors value the whole AI ecosystem, from equipment makers to platform companies. The market is not just debating whether semis are cheap. It is debating whether the entire trade has gotten ahead of the economics that are supposed to support it.
“There are so many people on the same side of the boat that there is always going to be a situation where there is over-ownership,” Markham told Bloomberg TV.
That is the core point. The selloff is not merely testing whether AI is real. It is testing whether the market has already discounted too much of the future value that AI spending could generate. If the answer is yes, then the dip is a lower price on the same old assumptions — not yet a compelling invitation to buy them.
Why This Is Not Yet A Classic Dip-Buying Setup
The strongest bullish counterargument is that this is still only a sentiment-driven pullback inside a secular AI uptrend. That view has force. Data-center build-outs remain large, chip demand has not vanished and the sector is still supported by the long lead times and capacity constraints that have characterized the AI infrastructure cycle. If the current slide is mainly the result of profit-taking and crowded positioning, then lower prices can eventually restore the asymmetry that dip buyers want.
There is also a historical case for caution in both directions. Semiconductor rallies have often looked vulnerable just before a fresh round of positive earnings revisions or capex guidance, and they often rebound once the next set of numbers validates the long-term story. That pattern is one reason many investors reflexively treat weakness in chips as temporary. The sector has repeatedly punished complacency on the way up and rewarded patience on the way down.
But Markham’s warning suggests this correction may be more complicated than a routine washout. The cyclical element is real: chips can fall hard when the same crowded investors decide to reduce risk at once. Yet the structural question is different. The market is increasingly focused on whether AI investment can justify the valuations assigned to the beneficiaries. If spending is still growing but monetization is less obvious, then the selloff becomes a repricing of the business model, not just a reset in sentiment.
The distinction matters because markets recover faster from inventory and positioning shocks than from structural doubts about profitability. A cyclical dip clears when supply adjusts and buyers return. A structural repricing clears only when the market gets evidence that the old assumptions were too cautious or that the new economics can sustain richer valuations. Until that happens, the market is not buying future earnings at a discount. It is arguing over the size and reliability of those earnings in the first place.
That means the burden of proof has shifted. Dip buyers now need more than the fact that AI spending is large. They need evidence that the next round of investment still carries strong margin leverage, that customer budgets remain committed and that inventories or lead times are not masking a slower demand path underneath. Without that evidence, a lower price is just a lower price.
What Would Prove The Bear Case Wrong
The counter-thesis is not weak. It is that the market is overreacting to a move driven by positioning, not by fundamentals, and that the long-term AI build-out still has room to run. If hyperscale capital spending remains elevated, if chip makers keep reporting firm order books and if the sector’s leaders show that revenue growth still converts into cash flow, then the selloff will eventually look like another mid-cycle washout rather than the start of a deeper regime change.
The clearest signal that would support that view would be sustained evidence of spending resilience and margin durability. In practice, that means the next round of company updates would need to show no material slowdown in data-center demand, no meaningful compression in gross margins and no broad-based cut to 2026 capex plans from the largest AI buyers. If those conditions hold, the market can argue that the rout was a temporary de-risking event rather than a change in the earnings framework.
But the falsifier is equally clear. If customers start flattening their AI capex plans, if inventories stop tightening or if leading chip companies begin cutting margin assumptions for multiple reporting cycles, then the selloff would be telling investors something more durable: AI demand is still real, but it is no longer enough to support the old valuation framework.
That is why this is still better described as a test than a bottom. The short term is a crowding and positioning unwind. The medium term is a challenge to earnings expectations. The long term is whether AI spending becomes a steady profit engine or merely a capital-intensive race that forces multiples lower.
The base case is a choppy consolidation in which chip stocks remain volatile as investors wait for the next round of spending and margin evidence. The upside case is a faster rebound if guidance re-accelerates and the market decides the pullback was mostly technical. The downside case is a deeper derating if the next wave of customer commentary shows that AI capex is still rising but its payoff is becoming less certain.
For now, Markham’s message is not that the chip trade is broken. It is that the market may still be working through the difference between a temporary dip and a valuation reset. Until that gap closes, the selloff looks less like an invitation than a question.
Explore more exclusive insights at nextfin.ai.
