TSMC ASML Record Earnings Q1 2026 | AI Chip Demand Soars

ASML, the Dutch lithography monopolist that manufactures the extreme ultraviolet machines without which no advanced chip can be produced, posted first-quarter net sales of €8.8 billion and net income of €2.8 billion, with a gross margin of 53% that came in at the high end of guidance. Earnings per share of $8.37 topped the $7.72 consensus by a wide margin.

The company lifted its 2026 revenue forecast to a range of €36 billion to €40 billion, up from prior guidance of €34 billion to €39 billion. “Demand for chips is outpacing supply. In response, our customers are accelerating their capacity expansion plans for 2026 and beyond,” CEO Christophe Fouquet said in a statement. The word “accelerating” is the key signal: ASML's order book is a 24-month leading indicator for total chip production capacity, and it is growing.

Hours after ASML's report, TSMC delivered the confirmation that the semiconductor upcycle is not a mirage. The world's largest contract chipmaker posted first-quarter net income of NT$572.48 billion, roughly $18.12 billion, a 58% jump from a year earlier that handily beat the FactSet consensus of NT$536.99 billion. Revenue climbed to NT$1.134 trillion ($35.9 billion), with high-performance computing, the segment that captures AI and 5G work, contributing 61% of total sales.

Advanced chips manufactured at 7 nanometers and below accounted for roughly 74% of total wafer revenue. The 3nm node, which serves Apple, Nvidia, AMD, and Qualcomm, is essentially sold out through 2027. TSMC raised its 2026 capital expenditure outlook to the high end of its $52 billion to $56 billion range to build additional “GigaFabs” in Taiwan and Arizona.

CEO Wei's comment about “agentic AI” is the most consequential forward-looking statement in either earnings report. The shift from generative AI, where a model responds to a single query, to agentic AI, where autonomous systems execute multi-step workflows without human intervention, represents a structural step-up in silicon demand. Each agentic task consumes orders of magnitude more tokens than a simple chatbot interaction, and those tokens translate directly into GPU cycles, which translate into wafer starts at TSMC.

This divergence, software companies under existential threat while their hardware suppliers post record profits, is the defining market dynamic of 2026. It explains why Salesforce and Snowflake have underperformed the S&P 500 by double digits year-to-date while the Philadelphia Semiconductor Index has climbed for ten consecutive trading sessions.

The dual beats from ASML and TSMC triggered a massive wave of foreign investment into Seoul's chip giants. Foreign investors have poured over 4.83 trillion won in net purchases into Samsung Electronics and SK Hynix since April 1, according to Korea Exchange data. SK Hynix hit an all-time high earlier this week on insatiable demand for High Bandwidth Memory (HBM) used in AI GPUs, while Samsung reported a record operating profit of 57.2 trillion KRW in Q1.

The Korean rally reflects a structural shift in how memory is valued. HBM is no longer a commodity product sold at spot prices. It is a capacity-constrained, custom-engineered component ordered 6 to 12 months in advance, exactly like TSMC's leading-edge logic wafers. This “foundry-like” pricing model gives memory makers pricing power they have never had before, and investors are repricing them accordingly.

With SK Hynix set to report its own Q1 earnings on April 23, the semiconductor cycle's momentum faces its next immediate test. A strong HBM revenue print from Hynix would complete the trifecta, confirming that demand is robust across the entire stack: lithography equipment (ASML), logic fabrication (TSMC), and memory (Hynix). Needham has already raised its price target on TSMC following the results, and multiple sell-side desks have upgraded their full-year semiconductor capital expenditure estimates.

The question for investors is no longer whether AI chip demand is real. It is whether the supply chain can scale fast enough to meet it. ASML's order book suggests new EUV machines will not arrive at customer fabs until late 2027 at the earliest. TSMC's Arizona expansion faces construction delays and labor shortages. And the tightening export controls on China mean that a significant fraction of global chip demand is being rerouted through a shrinking number of allied-nation fabs.

The bottleneck is no longer software. It is not algorithms. It is not even energy, though that is tightening. The bottleneck is lithography machines and clean-room square footage. ASML and TSMC are the toll booths through which every AI ambition in the world must pass, and their Q1 results show that the line is getting longer, not shorter.