Intel’s biggest interest and profit center is data center and AI if they can get back into that game.
From a margin and revenue growth perspective sure. But I suspect that CCG will always be the main operating income driver for the company just due to the shear volume differential.
So you want me to believe IFS and IDM 2.0 won’t have a conflict or potential conflict if Intel is presented to manufacture advanced TO for Nvidia
I'm just saying what I see. And what I see is Jensen Huang saying he liked what he was seeing. Does NVIDIA's prior positive statements on IFS mean they must use intel? No. But if NVIDIA thought that IP theft was a concern that made IFS a nonstarter, then why would they have their engineers waste their time designing a testchip and then testing said testchips in their labs? Said testchip might also have "steal-able" IP
or other company who’s new chip is directly displacing likely a potential Intel chip? Or with limited capacity IFS much choose between making wafer starts for Intel Product or Nvidia or other be it wafer or package?
Part of your question was cloud vendors making their own silicon. They themselves do not compete with intel, they are not selling their chips to anybody, and as a result they aren't a merchant chip vendor. Intel competes with them but not the other way around. As an example AWS will use the best CPU for any given AWS instance. If the chip with the best TCO/TVO is a Xeon so be it, and if it is Graviton they will use that. So for them their is no competitive risk of their IP leaking to DCAI. Inversely if DCAI wanted their IP, they could steal a mask set... OR they could just licence the IP AWS licenced from ARM et al. to build the chip since the cloud guys have no native IP development capability.
Lastly I would like to hear an actual chip designer weigh in on this, but I am skeptical of how much information a chip designer can gain from looking at someone else's mask set. There is nothing you couldn't learn from the mask set that you couldn't learn from cutting open the realized product. And without the context of it working and what the original design's intent was I don't know if you can really draw a whole ton from seeing the mask set pre launch. Another area that I think would hinder reverse engineering would be the evolutionary nature of chip designs. Using NVIDIA as our example: what if intel graphics had with ADA's and they wanted to get the same RT performance as NVIDIA? Their whole GPU arch is build differently with different implementations of different render features. I wouldn't doubt if I was wrong on this, but I assume you can't just slap NVIDIA's RT cores onto an intel GPU, because their GPU,drivers, and pipeline was built around doing it their way. Redoing all of that wholesale sounds way harder than just developing a better arch. Funnily enough that was the same conclusion AMD came to when they first started designing their own CPUs
Of course people put in contracts but we already know how poor companies are in forecasting demand two years out and it takes more than that these days to put leading edge capacity in place.
You have just answered your own question. Contracts are legally binding agreements. If intel signed a deal saying hey we will give you x wafers at y time; then their only way out of said agreement is to renegotiate the deal with their customer, or choose not to deliver to the terms of the contract. If they go with the later option, well okay, have fun losing a lawsuit where the customers get their money back. Oh and also IFS would never get another customer again... So in other words if they want to survive they have to follow their contract. Intel said they have over 10B in deal value and by my count they have announced like 13 customers for IFS. Okay imagine the legal fees for 13 cases that they will almost certainly lose and will then they need to comp any of the $10B they got for product that they haven't delivered plus any damages sustained by the customer. At current rates that is nearing a whole quarter of revenue down the drain, and for what? To sell like $2B worth of units. If intel did something so dumb they deserve to be out of business.
I don't remember Intel or TSMC saying building in AZ and OH was contingent on CHIPs Act funding but it is an excuse not to build now? I do understand the overbuilding concern. As I have said, if all of the fab builds that were announced since the pandemic were actually built we would have a capacity glut and I predicted there would be empty fab shells. So here we are.
Maybe Intel will be using TSMC N3 for longer than imagined? Good thing for chiplets! The whole chip does not need to be bleeding edge silicon and not even from the same fab. Progress!
TSMC said something to the tune of "We are only building in AZ with the understanding that the USG will subsidies our cost delta to TW.". Intel for its part said something to the tune of "Two fabs are guaranteed but the speed of the build out depends on subsidies as well as if/when they will build the other modules.".