There was an interesting analysis from the awesome team at Chipworks (with a good analysis from Ars as well) out last week showing that Apple’s new A9 processor for its iOS devices is being manufactured by two different ‘foundries’. A foundry is a company that owns a semiconductor manufacturing plant, aka a fab, and manufacturer chips that someone else designed. Apple now designs all its own chip and has contracted with both Samsung and the Taiwan Semiconductor Manufacturing Company (TSMC) to actually manufacture the chips.
This dual-sourcing strategy is one of those engineering decisions that Apple makes which both dazzle and terrify me. First, designing a chip is incredibly expensive. Most companies use ‘merchant’ chips, these are general purpose chips which anyone can buy. Most of the time, the merchant vendor wins because they can spread the costs of designing that chip across many customers. Apple’s A-Series and other internal chips, must cost several hundred million dollars a year to maintain, possibly as high as $1 billion. Very few companies can manage that.
Of course, Apple can afford it. And if you read some of the recent commentary on this, (for example, Gruber at his best), you can see the benefits of taking this approach. Companies that can afford this, can build chips custom suited to what they want, which means they are faster and more power efficient. By some metrics, Apple is now 3 years ahead of the competition who rely on merchant silicon. Three years in silicon terms is two generations, a huge competitive moat.
The other thing that fascinates me is the fact that Apple has chosen to dual source this production. Having two suppliers translates into immense leverage when it comes to pricing negotiation. This is very common among all electronics companies, but Apple is taking it to a whole new level. Apple is known for being incredibly un-sentimental in the way it treats its suppliers. There have been many Wall Street Jounral, New York Times and Harvard Business School profiles about how this gives Apple immense flexibility in expanding, or contracting, production of its products. Another big competitive advantage.
But if you are one of those suppliers, sitting on the sharp end of that flexibility, life can be hard. I know whole villages in China that have been essentially de-populated overnight when the factory that anchors the town lost Apple business and had to shut down.
The rational side of me understands that this is just part of the reality of modern capitalism. What staggers me about last week’s news is that Apple is now putting semiconductor foundries in that position. It costs several billion dollars to build a fab. Imagine spending that kind of money to ensure having enough capacity for Apple, only to lose that business a year from now when the next version o the iPhone comes out.
I know that TSMC, in particular, has angered another very large customer because they devoted so much effort to winning the Apple business. They are going to be in a very tight spot when they go back to Apple to negotiate for A10 production. They have already lost a lot of business from that other customer, and at some point will have to negotiate with Apple for the next chip. If they lose Apple, they will have a an empty $5 billion building. Apple knows that, and will use this fact when it comes time to talk about price.
I do not want to paint Apple as some ruthless group of cutthroats. This is just business. Apple may want to reconsider its approach to suppliers, as they may someday need to call in favors, but that day is not today.
This is just the reality of the market. Apple did not invent these practices. When Nokia had 50+% share of handsets, they operated in a similar fashion.
However, I do think that there is a stark reality which everyone in the consumer electronics supply chain faces. One customer commands a huge share of the industry. You either do business with that customer, or fall behind competitors who do business with them.
Most hardware executives I speak with today are very aware of this fact. It is a big factor driving consolidation inside the semiconductor industry. It is causing a lot of transition among senior executives in hardware looking to move to software and Internet companies.
The really troubling part is that I see no sign of it changing any time soon. This reality is part of the reason the companies have been marketing the “Internet of Things” (IoT) products so aggressively, it is the only business out there big enough to re-balance things. But as I have cautioned (and here and here) IoT is going to take a long time to emerge as a business, and many parts of could be just as imbalanced.