It is crystal clear that MS and the other "hyperscalers" want to design their own custom AI chips. This is for two reasons. 1) They are sick of getting ripped off by NVIDIA and sick of being at being at the mercy of essentially just one company. 2) Their own custom designed AI chips will be better than any general purpose chip designed by NVIDIA, Intel or AMD.
Intel has a tremendous opportunity here with MS. MS is the world's most valuable company. They have the means, they have the money, and they have the drive to stay dominant here....They are dominant now with their partnership with Open AI. Once MS has it's own chips, they will be a major force in AI. MS see that Intel's 18A process, with GAA transistors, and backside power delivery is a better process than TSMC's at least for the next few years. TSMC is still using FinFET transistor and frontside power delivery, and by their own admission will continue with this until 2026. TSMC is now surpassed. 18A is real and here now. Intel is fabricating devices for customers with 18A now on test chips. 18A is way way beyond R&D at this point and is just a few months away from beginning full fledged high volume manufacturing. Customers who have had test chips for months are extremely happy.
Google, Amazon, Meta, and Tesla also are designing their own AI chips. There is a high probability that at least one more of these will come knocking on Intel's door. Then there's Apple. They too could use Intel fabs.
Comments