Philip Elmer-DeWitt calls attention to public remarks from Apple's COO Jeff Williams regarding their bet on TSMC.? I guess there's "betting the company" and then there's betting the company.?
And Apple decided?it?s our first engagement?Apple decided to have 100% of the new iPhone and new? iPad chips [A11 application processors] sourced at TSMC.
So Apple was betting their workhorse product on TSMC.? Completely.? And this was a $9B bet by TSMC:
And TSMC invested $9 billion and had 6,000 people working ?round the clock to bring up a [unintelligible] fab in a record 11 months and in the end the execution was flawless. And we?ve gone on to ship over half a billion chips together in that short window.
The scale of these investments is incredible.? And in Asymco's blog "Silicon Valley" we get another view into the kind of bets Apple makes on its own business.? Horace Dediu calls attention to Johny Srouji, who runs the silicon development at Apple:
Under Johny?s leadership, Apple has shipped 1.7 billion processors in more than 20 models and 11 generations. Currently Apple ships more microprocessors than Intel.
The Apple A11 Bionic processor has 4.3 billion transistors, six cores and an Apple custom GPU using a 10nm FinFET technology. Its performance appears to be almost double that of competitors and in some benchmarks exceeds the performance of current laptop PCs.
I had modest expectations for Apple's ability to design its own chips when they kicked things off with the A-series chips.? I just assumed that most of the "surprise" in the outstanding performance of those chips was due to (relatively) cheap wins in the design of these chips by forgoing elements that weren't needed and focusing on those that were.? And perhaps I was right for the early generations - that the wins were relatively easy and as a result of optimizing the inclusion or exclusion of major generic components. But at this point, with the A11, it is clear that the advantage is significant and due to internal design capability, not just the internal consistency of building their own software for their own hardware.
There's no doubt that Apple has the scale to make this investment worth it.? But it was probably not so clear in the year when Apple started down this path.? As Horace writes:
This across-the-board approach to silicon is not easy or fast or cheap. This multi-year, multi-billion dollar commitment is rooted in the Jobsian observation that the existing supplier network is not good enough for what you?re driving at. Tiny EarPods, Smart Watches, Augmented Reality, Adaptive Acoustics require wrapping your arms around all parts of the problem. The integration and control it demands are in contrast to the modular approach of assembling off-the-shelf components into a good-enough configuration.
I'm sure Apple can still hedge its bets - by sticking with an old processor for another year for example - but so far it has kept a steady drumbeat of deliveries of new chip designs and bringing online new partnerships like TSMC.? Someday someone will do business case studies on how Apple managed to pull off projects of this scale.