I heard something similar recently. Do you expect pilot hybrid bonding adoption in logic-on-logic for AI accelerators, GPUs, CPUs versus mobile? Which do you expect to come first, and what might lag or come later?

Considering the performance-to-cost ratio, AI accelerators like GPUs or ASICs are eager for high computing power and high data rate transfer. This will be a significant growth engine supporting hybrid bonding. AI accelerators will certainly need hybrid bonding. The question is when companies like AMD, Nvidia, Broadcom, or Qualcomm will adopt it. I expect this will happen around 2028, during the second phase of system scaling. They will need integrated 3D on N2 chips or A14 chips on top of N2. This might happen in 2028, and for Nvidia, perhaps their framing product will require hybrid bonding at that time.