✦ The Future BetsNarrow moat

A monopolist's future bets look different from a challenger's. AMD must bet to become something; ASML bets to remain the only one — and its four live wagers are best read as new lanes on a toll road it already owns.

The first lane is open: High-NA EUV crossed from promise to product when Intel began shipping the first high-volume chips patterned on ASML's EXE platform¹, with the 14A node — the first designed around High-NA — entering risk production in 2027. The second lane is entirely new territory: the TWINSCAN XT:260, ASML's first machine built for advanced packaging, took the company into the back-end of chipmaking for the first time in its history², just as AI's appetite for stacked and stitched silicon made packaging the industry's scarcest capacity. The third is the strangest capital allocation Veldhoven has ever made: €1.3 billion into the French AI lab Mistral, making ASML its largest shareholder at about 11%³. And the fourth is the long lane — Hyper-NA, the 0.75-aperture platform penciled for pilot systems around 2030⁴, the machine that would re-found the monopoly a third time.

Notice what is absent: risk of ruin. Every bet is funded from a business earning roughly €10 billion a year with a €38.8 billion backlog; none bets the company; and three of the four simply extend the physics franchise ASML already monopolizes. The Mistral stake is the outlier — a lithography company buying an option on the intelligence that designs chips — and it should be judged as strategy, not venture arithmetic.

Grade the lanes separately. High-NA grades on orders and second customers — Intel is proof, TSMC's embrace is the prize. The XT:260 grades on units shipped into the packaging boom. Hyper-NA grades on nothing but its calendar for years yet. And if all three succeed, the effect is the same: the road gets longer, the toll gets larger, and the only company that can build the booth is still Dutch.