Thoughts About Apple’s Secret iPhone 7 Chip

An intriguing chip has been discovered in the teardown of the iPhone 7. We know that it’s a field-programmable gate array (FPGA), but we don’t know what it’s intended to do. Speculation abounds, and John Martellaro thinks it my be for encryption.

An FPGA from Altera.

The FPGA was noted in the bill of materials tear down by Chipworks back in September. Its part number is listed as “Lattice Semiconductor FPGA ICE5LP4K.”

An FPGA is an integrated circuit that works like an Application Specific Integrated Circuit (ASIC) but can be reprogrammed in the “field.” That suggests that Apple wants to be able to change specific, low-level, programming logic at the hardware level if necessary.

This Forbes article notes that the Galaxy S5 in 2015 had one, but it was dropped from the S6.

That article also notes that Microsoft “… has been investing heavily in making its own custom FPGAs to boost the artificial intelligence capabilities in its data centers.”

A Matter of Scale and Encryption

However, just because Microsoft is using FPGAs in its AI research data center, it doesn’t follow that an FPGA in an iPhone will be used for AI/Siri. Although it could, my own hunch is that a single FPGA in an iPhone isn’t going to have the scale required to implement the massive heuristics required for AI. If Apple can’t implement what it wants in Siri now, a small FPGA isn’t going to push Siri to the next level.

My own guess, and it’s just that, is that an FPGA would be a good place to implement encryption code that cannot be accessed or modified by malware from iOS and which can be updated in the future as needed. The state of the art is continuing to advance with encryption code, and many iPhone 7s can likely be expected to still be operating in 2019-20.

Given the resourcefulness and intent of both legal and illegal attempts to break into an iPhone, for a variety of reasons, it would seem sensible for Apple to grant itself some flexibility in this arena.

Again, this is a hunch, but it sounds a lot more reasonable to me than artificial intelligence, which requires grand scope. I could be wrong. I probably am.

Clever researchers will likely be on the case, and we may know more soon.

Share this post