the onda

There is tons of speculation about what Apple intends to do with PA Semi, the chip company they paid $300 million for last year. Though it varies wildly, one thing that people do agree on is that this means the fruit company is going to push more into CPU design and all of the associated accoutrements (GPUs, custom motherboard chips, compilers, deep OS tweaks, etc.)

In the words of J.J. Abrams's Scotty in Star Trek "that's essitin'" mostly because it puts a company with very deep pockets in a place where fundamental innovation can take place. Back when I was living in California and Steve had recently come back to Apple (and killed the clone business), I remember someone saying that Jobs wanted to be so vertically integrated that trucks full of sand would pull into one end of the building, and full computers would come out the other. With PA Semi they are now practically set to turn sand into chips directly!

In a related note, this past weekend, I saw an Alan Kay talk (at Intel of all places) about how the x86 straightjacket has kept software big, complex, and slow. He seems fascinated with FPGAs, going deep into how it's important for the chip to be able to adapt to the software so that it needs to run more efficiently and with fewer levels of abstraction on top of the hardware. Thanks to an NSF grant, Kay has undertaken a "grand challenge" project of implementing an entire system from the applications to the CPU instructions in 20,000 lines of code which is why he is so obsessed about having control over the individual gates.

The most amazing to me about the prospects of PA Semi inside of Apple products is how this evolution has happened. Steve might have always wanted to pound sand into chips but he started from a bunch of commonly available components glued together with great software (for both the iPhone and the iPod) and only after building a big enough business to justify this kind of vertical integration, has this move taken place. And if it works, it won't matter how many other people can buy capacitive touch displays at Foxconn. With the level of integration that Apple might achieve, they could crack something fundamental enough to stand out in terms of power management, performance, or even programmability in a way that might truly stand out.