14 DAY TRIAL // Many of the big contract manufacturers like TSMC and UMC started charging their fabless customers “by wafer” instead of charging them “per working chip.” This has increased the costs for the fabless companies like AMD and Nvidia, but has assured the FAB owners that they won’t operate at a loss.
There are also some additional advantages in charging per wafer – if the chip designer really optimizes its IC architecture, it can obtain more working chips per wafer without having to pay more.
Moreover, if they have a powerful and efficient architecture with a smaller die surface, they will also be able to fit more chips per wafer.
Many fabless chip designers are rooting for fast 450mm wafer adoption, as they will be able to fit more chips per wafer and the overall manufacturing costs will go down despite the wafers will be more expensive themselves.
Several chip makers have reportedly formed an alliance called Global 450 Consortium and they are putting up $4.8 billion for the development of the necessary technology.
Members of the Global 450 Consortium are Intel, IBM, Globalfoundries, Samsung and TSMC and market analysts like Gartner’s Bob Johnson say that 450mm wafer production will only start in 2017 or even 2018, with wide spread availability in 2020.
Chip designers know that a 450mm wafer will house more than double the chips a 300mm wafer is able to do right now. As we said before, the cost per wafer will not be much higher and it certainly won’t be doubled.
Thus, an IC designer like AMD or Nvidia will be able to pay a bit more for a wafer, but it will get almost double the number of chips.