The headline in ASM International’s third-quarter report was a revenue record: €1.159 billion, at the top of its own guidance range and up 44% from a year earlier at constant currency. The number that mattered was the guidance.
ASM raised its 2027 revenue outlook to more than €5.7 billion from the more than €4.6 billion it set in July, an increase of about 24% in three months. That also pulls forward the 2030 target it had set for itself, reaching it three years early. The company authorized a buyback of up to €300 million, and its U.S.-listed shares rose 6.7% on Tuesday.
Third-quarter gross margin held at 52.0% and adjusted operating margin at 33.4%. Adjusted net income rose €109 million from a year earlier to €315.4 million. Revenue and profit both came in ahead of expectations, which is why the guidance raise was read as a statement about the order book rather than as a good quarter.
What ASM sells is the machinery that lays down films one atomic layer at a time, along with the epitaxy tools that grow the crystal structures underneath. The economics of that business depend on step count. Every additional film, every additional interface and every additional layer in a stack is another pass through a tool that costs what a tool costs. Wafer starts across the industry grow modestly in a good year. The number of deposition steps per wafer at the leading edge does not.
That gap is the business model, and this quarter shows what it looks like when it widens. Deposition has moved from a supporting role to a defining one as logic has shifted to gate-all-around transistors and as power delivery migrates to the back of the wafer. Each of those changes adds films that must be placed with atomic precision, and each adds tool purchases per fab rather than per wafer. When capital spending rises, ASM’s orders rise faster. The reverse is also true, which is the part investors forget in the third year of a cycle.
The memory side of the order book is HBM-related, meaning tools are being bought for capacity that will be sold into accelerators. Stacking DRAM dies multiplies the number of deposition steps per bit, and that capacity is being added in new buildings, so the tools are ordered before the bits exist. HBM orders are therefore a leading indicator of the memory makers’ confidence, and just as reliably a leading indicator of the next correction in memory.
The detail worth circling is the “first material contribution” from investment in 1.4nm. A customer is spending on a node that will not ship in volume for years. Equipment orders at that stage are the strongest available evidence that a roadmap is being funded rather than discussed, and they are also the orders most likely to be deferred if a customer’s end demand disappoints.
Margins tell a similar story. A 52% gross margin and a 33.4% adjusted operating margin are unusual for a maker of capital equipment, and they reflect a product line with few substitutes at the leading edge. Pricing power of that kind does not come from scale. It comes from being specified into a process flow, where changing suppliers means re-qualifying a step and risking a yield hit that costs more than the tool.
The buyback is modest next to the market value of the company, and its significance is more about what it rules out than what it returns. A company raising long-range guidance by a quarter and simultaneously buying back stock is telling investors it does not see an acquisition large enough to absorb the cash, and that it would rather retire shares than hold a balance sheet for a deal that may not come.
The guidance raise invites an obvious question the company did not answer. If 2027 now equals what 2030 used to mean, what does 2030 mean? Management offered a buyback instead of a new long-range number. The conservative reading is that visibility does not extend that far and nobody wants to publish a target they cannot see. The less conservative reading is that the company intends to re-base again, and would rather do it in a year when the news is good.
Risks are not hard to find. Customer concentration is the first: a small number of logic and foundry buyers plus three DRAM makers account for most of the revenue, and a pause in one customer’s fab schedule shows up within a quarter. The memory cycle is the second, because the same leading indicator that predicts order growth predicts its end. Export controls are the third, and they have already redrawn which customers can receive which tools.
The equipment business has a long history of guiding to a peak and calling it a base. ASM’s argument, that deposition intensity is structural and rises with every node and every stack, is the reason to think this cycle is different. The next two quarters will show whether €5.7 billion was an order book or an extrapolation.
