With all the hyper scale providers building or launching their own inference chips, this seems untimely. If intel wants a return on that $2B, I think it will come down to how fast they can execute mass-production of these chips. Unless I have missed a distinct advantage here.