Ready, Player, M1
A long time ago I taught introductory computer architecture at Stanford. I started before you could get a bachelor’s degree in CS at Stanford. It was that long ago. When I started teaching we were using single board Motorola 6809 computers but with the arrival of the Mac we moved all of the lab work to the Motorola 68K. Some luminaries like George John, founder and CEO of Rocket Fuel, remember having to write a multi-tasking kernel as their final project.
My other job was at Tandem Computers, one of the first Kleiner-Perkins startups, which was disrupting traditional IBM mainframes with minicomputer technology. Along with companies like Digital Equipment we used proprietary computer architecture to provide more performance and reliability at a lower cost than the traditional mainframes.
It didn’t take long to realize a second wave was starting, driven by microprocessors used in PCs and workstations. I even wrote a paper at Tandem about how our instruction set could be emulated by the Motorola 68K microprocessor. Years later the company got rid of it’s proprietary computer architecture and moved to the MIPS microprocessors. And some years after that the transition was complete as the company moved to Intel.
For the past twenty years in enterprise computing we’ve been happy to think Intel was going to supply the CPUs. Just look inside AWS, Azure and Google’s cloud data centers and it’s pretty much all Intel.
But maybe that’s going to change. Yesterday, I attended a Stanford seminar, where the VP of platform architecture at Apple was the guest speaker. A similar old timer he talked about minis taking over from mainframes, and then PC/workstation technology replacing mini computers. So having seen two of these transitions he presented the third wave: M1.
Apple has developed a SOC they call M1, which was born from the iPhone business, which many now see as the modern PC/workstation. M1 today is being made available in the MacBooks and he couldn’t stop talking about how much faster the M1 MacBook was than the Intel MacBook it replaces. I would hate to be the Intel rep on the Apple account.
So what if Apple takes M1 and it’s derivatives and brings it to the rest of the world? Why couldn’t Apple become the computer architecture of the center cloud services of the future? Why couldn’t they start with their own cloud services (iTunes, Apple Music, Apple TV, Podcast, etc.)? And what stands in the way of exposing that compute & storage as a “public cloud service”. Maybe call it Apple Web Services — or on second thought, maybe not.
If you were Intel you’d be thinking my Xeon data center business isn’t safe. If you’re NVIDIA you have to be hoping the ARM merger happens quickly. But even then, you’d be stuck with the problem of who is going to purchase your version of “M1”. Qualcomm invested 1000 engineer team for more than 5 years before it abandoned it’s server cpu architecture in 2018. Is that because there was no customer of scale? So should Samsung just buy NVIDIA & ARM and create a competitor before Apple decides to go to after the rest of world?
The third wave has begun. Let the games begin.
