… I would rephrase. Transceiver makers and hyperscalers are the customers and foundries are hired to produce the wafers for these customers . NVIDIA tells TSMC what it wants, TSMC agrees on specifications, volumes and price deliverables. Yes, in time of shortages or availability issues the customer may compromise and take what’s available. It seems that the current supply situation helps to extend the life of incumbent solutions given that demand exceeds supply.
Yves indicated that either the transceiver or hyperscaler (I would have to listen again) has their own foundry preferences, which I would think would not be one of the two they developed/developing the PDKs with. It doesn't mean they start completely over, but it suggests to me that the new foundries have to go through a learning process even with a PDK that might not even work in their plants the first time around.
Market Data 
Markets 
AI
