By: Doug S (foo.delete@this.bar.bar), November 1, 2020 9:45 am
Room: Moderated Discussions
Michael S (already5chosen.delete@this.yahoo.com) on October 31, 2020 8:53 am wrote:
> I am not sure that it makes sense economically for FPGA company to become an early adapter of the
> new process. Very new chips are tiny part of the revenue, so why overpay for their production?
> It seems, typically FPGAs start to ship production parts on the process that is 1-1.5 years old.
> With likes of Apple overpaying for new stuff, may be, by now it should e closer to 2 years.
New chips may be a tiny part of their revenue, but if they can get customers to pay high enough prices for them that they make a nice profit why not sell them? If you assume they will go to that latest process eventually, they are going to have to pay the design/mask costs anyway it is just a matter of when.
Also not sure why you think Apple is "overpaying" for new stuff. Apple is able to use all the new transistors they get with each process - most of them went into the NPU on the A13->A14 jump so they obviously have some reasons for that. In a couple years they will probably have their own modem ready and that will be their use for the additional billions of transistors N3 will provide.
Anyway, supposedly the reason Apple gets first crack is not because they are paying more than others but because they are paying TSMC well in advance. Having a guaranteed customer of Apple's scale providing a lot of the funding and thereby reducing the risk makes it easier for TSMC to stay on the leading edge.
> I am not sure that it makes sense economically for FPGA company to become an early adapter of the
> new process. Very new chips are tiny part of the revenue, so why overpay for their production?
> It seems, typically FPGAs start to ship production parts on the process that is 1-1.5 years old.
> With likes of Apple overpaying for new stuff, may be, by now it should e closer to 2 years.
New chips may be a tiny part of their revenue, but if they can get customers to pay high enough prices for them that they make a nice profit why not sell them? If you assume they will go to that latest process eventually, they are going to have to pay the design/mask costs anyway it is just a matter of when.
Also not sure why you think Apple is "overpaying" for new stuff. Apple is able to use all the new transistors they get with each process - most of them went into the NPU on the A13->A14 jump so they obviously have some reasons for that. In a couple years they will probably have their own modem ready and that will be their use for the additional billions of transistors N3 will provide.
Anyway, supposedly the reason Apple gets first crack is not because they are paying more than others but because they are paying TSMC well in advance. Having a guaranteed customer of Apple's scale providing a lot of the funding and thereby reducing the risk makes it easier for TSMC to stay on the leading edge.