ASML Raises 2026 Outlook for the Second Time, Sets Ceiling at €45 Billion in the Year of AI
With €9.3 billion in revenue in Q2 and gross margin above the top end, ASML raised its 2026 guidance to €43-45 billion and indicated that AI capex has not yet peaked.
The Numbers That Forced the Revision
ASML closed the second quarter with net revenue of €9.3 billion and net income of €2.9 billion, achieving a gross margin of 54.0%, with both metrics exceeding the top of its own guidance. CEO Christophe Fouquet raised the revenue guidance for 2026 to a range of €43 billion to €45 billion, up from the previous projection of €36 billion to €40 billion. The expected gross margin has been adjusted from 51-53% to 54-56%. This marks the second upward revision for 2026, a signal that is difficult to interpret seasonally.
Fouquet was direct in his comments accompanying the results: "Investments in AI and the continuous advancement of AI technologies are driving demand for advanced logic and memory chips." The company also announced plans to add 30% capacity in low-NA EUV for 2027, based on an approximate base of 65 systems in 2026, with a further 30% expansion being considered for 2028. In immersion DUV scanners, the expansion for 2027 is also set at 30% over a base of around 130 systems.
What the Second Revision Means in the Supply Chain
ASML is, in practice, the only global supplier of EUV lithography, making it a leading indicator for any acceleration of capex upstream in AI. When the company raises its guidance twice in the same year, clients have already signed orders that would not be justified without stable multi-year demand from the silicon side. Translated: TSMC, Samsung Foundry, SK Hynix, Micron, and Intel have provided ASML with the contractual assurance that Nvidia, Google, Amazon, Microsoft, Meta, and now Anthropic have delivered to them.
This reorders the debate regarding bubbles. The thesis of the "AI capex bubble" gained traction in the first half of the year based on spending-side numbers, with hyperscaler capex projected to be close to $400 billion for 2026 according to Dell'Oro Group estimates. What ASML demonstrates is the supply side: they have already sold, are already delivering, and are already growing. Demand-side bubbles may not shrink until 2027 precisely because downstream supply is constrained by lead times of 18 to 24 months on critical equipment. The logic is straightforward: the floor of AI capex is defined by how many EUV machines the Veldhoven factory can package each year.
Where the Money Lands Outside of the Netherlands
Geographical analysis is worthwhile. Taiwan remains the largest buyer from ASML, with TSMC absorbing most EUV capacity for N3 and N2 production, which currently serves Nvidia, AMD, MediaTek, and Apple's in-house line. South Korea is represented by Samsung Foundry for 2nm and SK Hynix for HBM4, the latter being critical for Nvidia's H200 and B200 accelerators. The United States focuses demand from Intel for node 18A in Ohio and Arizona, as well as Micron for HBM memory in Boise and New York State. Japan is moving up the queue with Rapidus in Chitose and the TSMC Kumamoto fab.
For the Brazilian market, the impact is indirect but measurable. No advanced manufacturing stage runs in Brazil, but the price of GPUs imported by banks, regional hyperscalers, and HPC centers from mining companies, Petrobras, and federal universities is tied to the scarcity cycle determined by this same EUV capacity. As long as ASML expands slowly, H200 and B200 servers will continue to be rationed by Nvidia, and the CFO of any datacenter in Fortaleza or Camaçari will continue to target 2027 for actual expansion.
What Still Weighs Against
There is an opposing reading that deserves mention. ASML's growth depends on whether clients truly need 30% more EUV capacity in 2027, and several voices in European equity research note that a downward revision in capex by any of the three major hyperscalers in response to weaker AI demand could destroy the current guidance math in a single quarter. The economy is reflexive: ASML raises guidance because hyperscalers order because they project AI clients who have yet to pay the full bill. The first link to break pulls the entire chain down.
Nevertheless, today the evidence leans towards confirmation. Second guidance revision this year, gross margin rising, capacity being expanded across two product lines, and the CEO explicitly mentioning AI in the company's official material. Anyone betting against this must point out which of the three major players will cut capex first, and no candidates appear to be on the horizon.