When news broke earlier this week that Apple had finally inked a deal with TSMC to produce processors with the company beginning in 2014, the first round of chatter was on what this might mean for Samsung. Apple, after all, has been by far Samsung's single largest customer. With that topic out of the way, talk has turned to Intel. Once again, the tired rumor that Intel walked away from the manufacturing equivalent of Cibola is being floated across the Internet.
At first glance, it makes sense. Intel currently enjoys a hefty lead over every other country in terms of its processor technology. The company has admitted that its fab utilization is as low as 50%, which may be a record. With PC demand slumping and Intel's mobile business still scaling upwards, Santa Clara has every reason to seek a solution that would allow it to bring fab utilization up and start shipping more parts. Apple, meanwhile, can afford to pay a premium for Intel's technology. It's a match made in heaven!
It's The Little Things
Our trip to ARM last week actually highlighted some of the reasons why a foundry relationship between Intel and Apple would be difficult. Apple's A4 and A5 cores were custom implementations of a standard ARM Cortex-A8 / A9 processor, but the Apple A6 inside the iPhone 5 is based on Apple's own "Swift" architecture. Apple has its own processor design team now, and that means it would want a foundry license that gave it a great deal of flexibility and input into implementation and design.
Intel's D1X Fab
, meanwhile, simply isn't used to sharing control of those variables. The ARM license is another headache -- while Apple could certainly negotiate a license with ARM
to fab their chips, none of ARM's designs have been ported for construction in Intel's factories. Intel, meanwhile, would take a serious PR hit for agreeing to fab ARM for Apple while simultaneously plugging its own x86 solutions.
Finally, there's the issue of time-to-ramp. Apple
has been talking to TSMC about foundry work since at least 2011 but only signed an agreement recently. That speaks to the specifity of what Apple wants, the volume guarantees it required, and a careful negotiation process over what each party would control as far as semiconductor designs, foundry technologies, and IP production. Even assuming Intel could do the work more quickly than its Taiwanese competitor, it could still take 12-24 months for Cupertino and Santa Clara to hammer out an agreement.
After being burned (in its own opinion) by Samsung, Apple is going to insist on tough non-compete clauses. Intel, meanwhile, has a 22nm Atom architecture that it's itching to pitch at the smartphone market. Fabbing ARM chips for Apple could cripple that processors' reception -- who's going to believe it's a top-notch design if you're simultaneously building ARM processors for the most-visible smartphone vendor in the United States. It's hard to see Apple being thrilled with Intel if Intel ends up building chips for Samsung, Nokia, or HTC down the line -- particularly if those devices outperform Apple's own products.
This is the kind of deal that looks great at first glance but falls apart when both companies bring their long-term desires to the table. Intel would love to win the iPhone and iPad for Bay Trail or future Atom chips, but is going to be less enthusiastic about building ARM. Apple would love to take advantage of cutting-edge 22nm and 14nm technology, but would want a great deal of input into the manufacturing process and Intel's ability to compete with its own products.
Maybe one day -- but not today.