As hybrid bonding matures, a wave of consolidation is bound to follow, and Besi will eventually have to share the pie.
Not long ago, investors seemed to believe Besi was the ASML of the back-end. Now that the number of hybrid-bonding players is growing, CEO Richard Blickman has little choice but to let them down gently. After years of inflated expectations, reality is creeping back into the Q&A sessions with analysts. At the most recent capital markets day, Blickman openly admitted that a full twenty bonders are now in development worldwide. In fact, he brought it up himself.
Most CEOs only ever tell half the truth – Blickman is no exception. They’re obligated to keep their investors and shareholders properly informed, so they’ll only volunteer information about competition once it starts to affect the numbers and the hot potato can no longer be avoided. In conversations with analysts over the past few years, Blickman was happy to let the comparison between EUV and hybrid bonding stand. He never confirmed it, but he never contradicted anyone either, and he was content to let the “ASML of the back-end” myth live on.
Now that Besi’s chief himself is bringing up the competition, you can count on what comes next: News will soon break about a second hybrid-bonding supplier at TSMC, by far Besi’s most important customer today. That second player will show up sooner or later, because the Taiwanese will certainly not want to depend on the Dutch – one monopolist from the Low Countries is more than enough. Not just because of the technological dependency, but also for a stronger negotiating position. Bonders are expensive and you need a lot of them, since they still aren’t very productive. That’s exactly where the race will be decided: delivering flawless productivity.
Blickman himself says twenty hybrid bonding platforms are in development worldwide. Back in 2024, Trendforce already pointed out that a handful of equipment makers in Taiwan’s ecosystem were working on hybrid bonding. More than anything, that shows established players can clear this technological hurdle just fine.
It mainly makes clear that Besi’s market position looks quite different from ASML’s, the company that dominates chip lithography. EUV technology is so complex that its biggest rivals – Canon and Nikon – dropped out at an early stage. Hybrid bonding is tough, but it’s achievable for equipment makers with enough R&D muscle and the willingness to invest.
It increasingly looks like many of them have come to see hybrid bonding as a critical part of their strategy. They want this 3D technology in their portfolio no matter what. Just look at the interest from companies mainly active in the front-end, like ASML, Canon and TEL.
We’re in the middle of a hybrid-bonding evolution that, in hindsight, we’ll probably view as a revolution in the semiconductor industry. In any case, the folks in Veldhoven clearly understand they can’t afford to miss this development. They’ve dived into hybrid bonding themselves, if only because lithography scanners need to compensate for disruptive effects like warpage (chips warping after bonding).
In the old, bloody, back-end market, Besi faces its stiffest competition. Serious players like ASMPT, Süss and Shibaura are already shipping hybrid bonding systems there. On top of that, four companies in Korea alone are competing for the favor of SK Hynix and Samsung in particular. By far the largest market sits with this pair, since these giants need tools to stack their high-bandwidth memories.
Richard Blickman built his success mainly by shifting nearly all production to Asia and ruthlessly cutting costs. In hybrid bonding, he has to play a different game: spend heavily to hold on to the lead. Eric Beyne, the Imec back-end veteran who recently retired, told me not long ago that you can also invest too much. Either way, I’m curious to see how Blickman plays this game. If it succeeds, the potential payoff is huge – but it doesn’t come free. I expect Besi to stay on top for years to come, but as hybrid bonding technology inevitably matures, a wave of consolidation is bound to follow, and Besi will eventually have to share the pie.
Right now, Besi undoubtedly has the strongest starting position. It’s the only supplier of production-worthy, qualified bonders for high-volume manufacturing at TSMC. And the Taiwanese foundry’s fabs happen to be where all the major AI platforms get made. Whoever supplies TSMC – currently just Besi – effectively serves nearly the entire AI market with bonders. The foundry giant is undoubtedly working to line up a second supplier, so Taiwan is currently the most important battlefield for Besi.
Then there’s the most promising hybrid-bonding market: high-bandwidth memory. Besi expects HBM makers to adopt hybrid bonding starting with the HBM4e generation, the first units of which are now rolling out of factories. It’s already in at Micron in the US, but almost nothing is known about its relationships with the two biggest players, Samsung and SK Hynix.
What’s notable is that Besi’s management team was able to talk, at the analyst day, about Samsung’s results applying hybrid bonding in its HBM4e memory. The Koreans produced HBM stacks of twelve 10nm DRAMs on a 4nm logic base. A stack with 16 DRAMs is in development. While Samsung’s own press release remained silent about hybrid bonding, Besi revealed that its customer had managed to demonstrate that the technology significantly improved both power consumption and communication speed in a 12-stack HBM, compared to an HBM stacked using solder balls. It’s possible that this was a signal from Duiven that it also has a foot in the door at the Korean company. Either way, the pioneer will run into fierce competition in the largest HBM market, too.
Hybrid bonding, in any case, is no longer the blue ocean Besi once dreamed of – the water will keep turning redder and redder in the years ahead. Welcome back to back-end reality.
Top image credit: Applied Materials
