ASML raised its 2026 sales forecast on July 15 for the second time this year, and the stock rose roughly 3% on the news -- a notable move for a company whose order book is one of the most reliable leading indicators anywhere in the AI supply chain. Unlike a chipmaker whose quarterly numbers reflect demand that's already happened, ASML's extreme UV (EUV) lithography machines are ordered years in advance of a fab actually producing a finished chip, which makes a guidance raise this large, twice in one year, a real-time signal that AI capex intentions are still accelerating rather than plateauing.
The context matters. ASML holds a genuine monopoly on the EUV technology required to manufacture the most advanced logic chips at the process nodes Nvidia, AMD and the hyperscalers' own custom silicon programs all depend on -- there is no second supplier at this level of the stack. As recently as 2023 and 2024, ASML's guidance was clouded by uncertainty over China export restrictions and a broader post-pandemic semiconductor inventory correction; the fact that its 2026 forecast keeps climbing, driven explicitly by AI-linked orders rather than a cyclical recovery, marks a clear inflection from that more cautious period.
The competitive and customer landscape here is straightforward but worth naming explicitly: TSMC, Samsung and Intel are ASML's core customers, and all three are themselves racing to add leading-edge capacity for Nvidia's GPUs, AMD's Instinct accelerators, and the custom ASIC programs Google, Amazon, Microsoft and Meta are each running in-house. Every dollar of incremental ASML order flow is effectively a forward-looking vote from those foundries that AI-chip demand from their own customers justifies committing to more capacity years before it ships.
“What makes the timing especially interesting is the divergence sitting right next to it.”
What makes the timing especially interesting is the divergence sitting right next to it. SK Hynix shares jumped roughly 8% in Asia trading the same day, extending a rally that's been building since the company's own record-setting Nasdaq debut earlier this year, driven by high-bandwidth memory demand for Nvidia's data-center GPUs. Nvidia itself, meanwhile, has been sliding for weeks on reports that hyperscalers are accelerating a shift toward custom ASIC chips to reduce reliance on merchant GPU vendors. Put those three data points side by side and the picture is clear: the equipment layer (ASML) and the memory layer (SK Hynix) of the AI chip stack are strengthening at the same moment the merchant GPU layer (Nvidia) is being repriced downward.
That's a meaningfully different read than "AI capex is slowing" -- it isn't. It's a read that the market is getting more precise about which part of the value chain actually captures the durable margin as the buildout matures. Equipment and memory suppliers sell into every chip architecture a fab produces, whether it's a merchant GPU or a hyperscaler's in-house ASIC; a company like Nvidia is more exposed to which specific architecture wins.
For VCs and allocators with exposure anywhere in the semiconductor supply chain, ASML and SK Hynix's same-day strength is a reminder that picking the right layer of the AI infrastructure stack matters more right now than simply being long "AI hardware" as an undifferentiated basket. Equipment and memory suppliers are structurally insulated from the specific GPU-versus-ASIC architecture debate in a way merchant chip vendors are not.
The bear case: ASML's business remains genuinely cyclical, and a forecast raise built on current order momentum can reverse quickly if hyperscaler capex growth decelerates even modestly, or if China export restrictions tighten further and remove a meaningful customer segment from ASML's addressable market. What to watch next: whether TSMC's own upcoming capex guidance corroborates ASML's optimism, and whether SK Hynix's rally holds through its next earnings report or proves to be a single strong trading day inside a more volatile memory cycle.