ASML raises forecasts and prices amid AI chip demand surge

ASML's ability to raise prices mid-cycle, while its largest customer publicly resists, is only possible because of an absolute manufacturing monopoly. No other company produces commercial EUV lithography machines ^[1]. This means chipmakers have no alternative supplier to threaten ASML with - their only negotiating lever is deferral, which is exactly what TSMC is doing with High-NA EUV. But deferral has limits: TSMC cannot defer indefinitely without ceding process node leadership to Samsung and Intel, both of which are already adopting High-NA EUV ^[2]. The pricing dynamic playing out today - Chinese customers accepting 10% DUV price hikes while TSMC holds out on EUV - reveals a tiered leverage structure. Customers with no alternatives (China, locked out of newer systems by export controls) capitulate first. Customers with scale and strategic alternatives (TSMC) extract the longest resistance window. But as CFO Roger Dassen noted, the booked-out 2027 capacity means the window for TSMC's resistance is narrowing: every quarter of deferred adoption is a quarter where Samsung and Intel deepen their High-NA experience curves ^[3].

The magnitude of ASML's second guidance raise is striking. Moving from a €36-40 billion midpoint (€38 billion) to a €43-45 billion midpoint (€44 billion) represents a 16% upward revision in a single quarter ^[4]. Q2 2026 net sales of €9.33 billion beat analyst estimates by 6%, while net income of €2.92 billion beat by 11%. Gross margin guidance was simultaneously raised from 51-53% to 54-56%, signaling that the revenue beat is not coming at the expense of profitability - a common concern with capacity expansion ^[5]. The capacity expansion math is similarly aggressive: starting from a 2026 base of ~65 Low-NA EUV units and ~130 DUV immersion units, adding 30% annually compounds quickly. By 2028, ASML could be shipping roughly 110 EUV and 220 DUV immersion systems per year if the 2028 expansion is confirmed. Memory alone accounts for a significant share of this demand surge, with CEO Fouquet flagging 75% memory revenue growth in 2026 ^[4]. Applied Materials, a downstream bellwether, also raised its 2026 semiconductor equipment growth forecast to more than 30%, further corroborating that the AI-driven demand signal is broad-based rather than ASML-specific ^[6].

The High-NA EUV divide is the most consequential strategic split in semiconductor manufacturing right now. At ~$400 million per unit - roughly double the current-generation EUV price - High-NA systems represent a step-change in both capability and capital commitment ^[1]. Intel has moved first, shipping its 'Panther Lake' chips as the first High-NA EUV logic products, turning the technology from a lab curiosity into a demonstrated production tool. Samsung has committed to approximately 20 systems for sub-10nm nodes, and SK Hynix is advancing adoption for memory ^[2]. TSMC's position is the outlier. Kevin Zhang, TSMC's Deputy Co-COO, described current Low-NA EUV as 'exceptionally effective in pushing current EUV capabilities while maintaining aggressive scaling' - a careful framing that avoids conceding ground on node leadership while justifying cost discipline. TSMC has deferred High-NA for its A13 node through at least 2029 ^[2]. The strategic risk for TSMC is that Intel and Samsung accumulate yield learning and process integration experience on High-NA EUV for three-plus years before TSMC enters. In lithography, early yield curves are notoriously hard to replicate quickly. TSMC is betting that Low-NA can keep pace long enough that the gap will be small when it finally adopts - but that is a bet on supplier roadmap stability, not on demonstrated capability.

The geographic composition of ASML's Q2 2026 revenue reveals the structural forces reshaping the semiconductor supply chain. South Korea represented 43-45% of Q2 2026 net sales, driven by Samsung and SK Hynix's aggressive AI memory buildout ^[2]. This concentration reflects the HBM (high-bandwidth memory) demand wave feeding AI accelerator production - every Nvidia GPU needs stacked memory, and Korean chipmakers are the primary suppliers. China, meanwhile, is expected to represent ~20% of ASML's 2026 total sales - meaningful, but constrained by export controls that block shipment of EUV systems entirely and restrict newer DUV tools. Chinese customers' acceptance of 10% DUV price increases reflects their trapped position: they cannot access EUV alternatives and must take what is available ^[3]. TSMC, historically the anchor of ASML's revenue at 56% of the global EUV installed base, is becoming relatively less dominant as Korean memory demand surges and as ASML's customer mix diversifies. This geographic rebalancing gives ASML additional negotiating leverage with any single customer - including TSMC - because no one counterparty can credibly threaten ASML's growth trajectory by withholding orders.

Wall Street's reaction to ASML's results has been constructive: BofA raised its price target to $2,345 and JPMorgan maintains an Overweight rating with a $2,200 target ^[6]. Analyst Didier Scemama at BofA expected ASML's 2027 order book to be fully sold out by the time Q2 2026 results were reported - a prediction the earnings confirmed. The retail investor community is more divided. Discussions across investing-focused communities reflect a tension between two readings: bulls point to the monopoly, the booked-out capacity, and ASML's role as a bellwether for AI infrastructure spending - a framing reinforced by financial media coverage that treats ASML's results as a proxy for the broader AI capex cycle. Bears argue that a ~55x P/E already prices in an extended growth runway, raising the risk of a 'sell the news' reaction even on strong results. The longer-duration question - whether ASML's moat holds for 10-20 years - has also drawn substantive community debate, surfacing risks that institutional notes tend to underweight: e-beam lithography and nano-imprint lithography (NIL) as potential long-term alternatives, and the geopolitical concentration risk inherent in a single Dutch manufacturing site. The emerging community view is that a decade of competitive protection is near-certain given the complexity of EUV supply chains, but that two decades involves meaningful technology and geopolitical uncertainty. What is not debated is the immediate picture: with capacity fully booked through 2027, AI infrastructure spending accelerating, and no credible EUV alternative, ASML's near-term pricing power is as strong as it has ever been.