Picks‑and‑shovels bets
Some investors are framing companies like ASML and specialty photonics firms as 'picks‑and‑shovels' plays for AI hardware demand rather than pure GPU bets. ( ) That view rests on multi‑year backlogs in equipment and substrate supply, and analysts recently repriced ASML on stronger EUV demand—cited price targets moved higher in the conversation. ( )
Investors hunting for artificial intelligence hardware exposure are moving beyond Nvidia and into the suppliers that build the tools, optics, and packaging behind the chips. (asml.com) ASML, the Dutch company that makes the lithography machines used to print advanced chips, raised its 2026 sales outlook on April 15 to €36 billion to €40 billion, up from €34 billion to €39 billion. It reported first-quarter 2026 sales of €8.8 billion, net income of €2.8 billion, and guided second-quarter sales to €8.4 billion to €9.0 billion. (asml.com) ASML ended 2025 with a backlog of €38.8 billion after booking €13.2 billion of orders in the fourth quarter, including €7.4 billion for extreme ultraviolet, or EUV, systems. Bloomberg reported on April 15 that stronger artificial-intelligence spending helped push ASML to lift its 2026 forecast. (asml.com, bloomberg.com) Lithography is the chip industry’s version of a printing press: chip designers draw the plans, but companies like ASML supply the machines that etch those patterns onto silicon wafers. EUV is the most advanced version of that process, and ASML remains the only commercial supplier of EUV systems. (asml.com, asml.com) The same “picks-and-shovels” logic is spreading to photonics, the business of moving data with light instead of electrical signals. Nature wrote in February that photonic technologies such as co-packaged optics and optical circuit switches are being adopted to cut bandwidth, latency, and power bottlenecks in artificial-intelligence infrastructure. (nature.com) Nvidia underscored that shift on March 2 by investing a combined $4 billion in Coherent and Lumentum, with $2 billion for each company under multiyear strategic agreements tied to optical components and manufacturing capacity. CNBC reported the deals were aimed at securing photonics supply for next-generation artificial-intelligence systems. (cnbc.com, photonics.com) Coherent has spent the past month pitching 1.6-terabit and 3.2-terabit optical transceivers, plus co-packaged optics designs, for artificial-intelligence data centers at the Optical Fiber Communication Conference in Los Angeles. Lumentum said this month it is demonstrating optical products for “scale-out, scale-up and scale-across” artificial-intelligence infrastructure and building a facility for lasers used in large data centers. (coherent.com, coherent.com, lumentum.com) Packaging is part of the same trade. Ajinomoto, better known to many consumers for food seasonings, told investors that high-performance computing substrates use more than 10 times as much Ajinomoto Build-up Film, or ABF, as PC substrates, and linked demand growth to deep learning and natural-language artificial intelligence. (ajinomoto.co.jp) Not everyone is treating these suppliers as a one-way bet. Reuters reported before ASML’s April results that investors still wanted stronger second-quarter bookings to support its growth case, and CNBC reported on April 15 that tighter China restrictions pushed ASML shares lower even after the company beat expectations and raised guidance. (marketscreener.com, cnbc.com) The bet now is that the bottlenecks in artificial-intelligence hardware sit in the factories, optical links, and substrate layers around the processor, not only in the processor itself. ASML’s raised forecast, Nvidia’s photonics deals, and the packaging data point to the same part of the supply chain. (asml.com, cnbc.com, ajinomoto.co.jp)
