Ever since the iPhone launched, it’s used a Qualcomm modem. Even after Apple began fielding its own
custom CPU cores, the Cupertino company relied on Qualcomm for its cellular baseband.
Intel reportedly wants to change that, and is pushing hard to build a modem solution that would meet
Apple’s design targets.
Intel, as we’ve previously covered, is one of the few players left in the 4G modem business at all.
While there were a number of companies in the 3G modem business, many of them have dropped out of the
race or consolidated their efforts at 4G. Qualcomm controls the overwhelming majority of the market.
Intel, meanwhile, has eked out a handful of wins, but not much more than that. Several years ago,
Intel decided to keep its modem efforts focused on TSMC and the 28nm node rather than porting their own
technology to 14nm and 4G.
This raises the rather interesting question of which solution a hypothetical iPhone 7 might use.
Intel’s XMM 7360 is a 28nm chip built at TSMC, but it’s possible that Intel would develop a 14nm
solution for Apple if it felt doing so would win the company’s business. Moving cellular baseband to
lower process nodes is a very different animal than performing the same work on traditional semiconductors.
Analog and RF process nodes require their own implementations and do not gain the same power and
performance benefits from smaller nodes as other components within an SoC.
VentureBeat is reporting that Intel has over a thousand people working to develop a modem that would
fit Apple’s specifications and claims that a deal “will happen if Intel continues to hit its project
milestones.” We’d take this with a larger grain of salt. There’s no doubt that Intel would love to be
Apple’s modem supplier of choice, or even to serve as a dual supplier in some markets — either
situation would give the mobile division a much-needed revenue and visibility boost.
In theory, winning the modem could open the door to fabbing integrated SoCs for Apple, though we suspect
such a deal would need to clear mammoth structural hurdles to be attractive to either party.
The problem with predicting an Intel / Apple marriage is that the two companies often have
contradictory goals. Apple doesn’t share its brand space; Intel typically demands top-billing.
Apple would likely want to develop its own CPU architectures, Intel wants to sell Apple on an x86 design.
Apple appears to be moving towards second-sourcing its products, since it split the A9 between
TSMC and Samsung, while Intel would be extremely reticent to allow second-sourcing — that’s the mistake
(from Intel’s point-of-view) that created AMD’s x86 license.
All of these problems could be overcome with sufficient amounts of cash, but collectively they’re
significant enough that we’ve never been all that gung-ho on the “Intel builds chips for Apple” concept.
Selling a modem into the iPhone would be a huge win for Santa Clara, but locking down the entire design
is orders of magnitude less likely.