Patent filings describe a bonder on a track

ASML has announced nothing, and SemiAnalysis says so at the outset. What the firm argues is that the company's patent filings, a news report and its own roadmap together point to a hybrid bonding tool in development, and that it will start with die-to-wafer rather than wafer-to-wafer. Die-to-wafer is the harder of the two to build, which is precisely where the analysts think ASML could gain an edge.

The design detail they draw out of the filings is the transport. The tool runs on a track: each die rides its own holder past the process stations, and the stations close in on the dies at each step rather than the other way round. The independent shuttles on that track, in their reading, lift station utilisation, save duplicate tooling and enable heterogeneous integration.

Their case rests on the two factors that set bonding yield, alignment accuracy and particle control, and on ASML already solving both in lithography. They are equally clear about what is unsolved: feed-forward metrology, the die-to-wafer contact itself, modifying the non-bonding tools so they can process dies on the track, and keeping the whole loop clean.

The competitive reading is the sharpest part, and it is an assessment rather than a fact. SemiAnalysis expects advanced logic, where die-to-wafer already runs in volume, to adopt such a tool first, and calls that bad news for Besi, whose hybrid bonding business is mostly logic and concentrated at TSMC and Intel. ASML has worked with both for more than two decades and announced a joint photomask initiative in September. The analysts also see Besi's high-bandwidth memory prospects weakening, with Korean challengers catching up and HBM adoption, in their expectation, slipping to 2029 or beyond.