The A7 in the iPhone 5s is the first 64-bit processor in any phone, a Samsung 28nm chip that caught the whole industry flat-footed, Qualcomm included. Samsung built it, and that is quietly the more interesting fact, because this is almost certainly the last flagship Apple chip Samsung will make. The foundry that has produced every Apple processor since the A4 is about to lose the account that built its foundry business, and the 64-bit triumph coming off its 28nm line is a goodbye it cannot say out loud.
28nm is the workhorse node of this moment, the generation Samsung reached after the 32nm gate-first bet, and it is where mobile chips get genuinely powerful without setting the battery on fire. The A7 is a monster for something that fits in a pocket: a dual-core 64-bit CPU Apple calls Cyclone, the successor to last year’s Swift, a quad-core PowerVR Rogue graphics block, a separate motion coprocessor, a Secure Enclave walling off the data from the new fingerprint sensor, more than a billion transistors on roughly a hundred square millimeters. None of that ships without a mature, high-yield 28nm process underneath it, and Samsung’s is mature. On manufacturing alone, the A7 is a showcase.
The dismissive read came fast and aged badly. Qualcomm’s marketing chief called 64-bit a gimmick, zero benefit to anyone, something you only need past four gigabytes of memory when the phone has one. Within days Qualcomm called its own executive’s comments inaccurate and moved him out of the job, because the gimmick line missed the point entirely. The 64-bit jump was never really about addressing more memory; it was the move to ARM’s modern v8 instruction set, twice the registers, faster cryptography, a cleaner core that did more work at the same clock. Apple shipped that while the Android camp was still counting cores, Samsung trumpeting eight of them in an Exynos that was busy fighting its own bugs. No rival had a 64-bit core anywhere on its near-term roadmap, and now every one of them has to go build one.
Everyone in the industry can see what is coming. Apple has spent years uneasy that its silicon is built by the company it is suing and being sued by, and the rumors that TSMC will take the next Apple processor have hardened into something the whole supply chain treats as settled. Apple wants out from under a supplier it is at war with, TSMC has spent the last few nodes proving it can handle that kind of volume, and the A8 is widely expected to be built in Taiwan rather than Korea. Samsung is about to find out what its foundry looks like without the customer it was shaped around.
Losing Apple is not a rounding error, it is the anchor tenant walking out mid-lease. Samsung is TSMC’s only credible number two at the leading edge, and a large slice of that standing is Apple’s wafers; strip them out and Samsung slides back toward the pack while TSMC holds roughly half the foundry market. Apple’s volume is what justified Samsung’s leading-edge logic spending, what kept the front of the process roadmap busy. Without it Samsung has to go find new takers for its most advanced nodes, the Qualcomms and Nvidias and AMDs of the world, companies that have spent years defaulting to TSMC and have no obvious reason to move. The 28nm A7 is a high note played on the way out the door.
The smart read is that Samsung already knows this and is repositioning around it. If Apple is leaving, the answer is to leapfrog: skip the incremental knife-fight on mature nodes where TSMC is dug in, and bet everything on being first to the next real transistor change, the one that finally moves the industry off flat planar devices. Samsung has the memory profits to fund a swing that big and the institutional stomach for a gamble. Beating TSMC to a genuinely new process is the only move that wins back what Apple’s exit is about to cost.
Maybe Apple stays. There is a version where the manufacturing entanglement is simply too deep to cut and Apple keeps dual-sourcing to keep everyone honest on price. I doubt it, the strategic case for leaving is overwhelming and the bad blood is real, but stranger things have held together in this business on switching-cost inertia alone. Either way, Samsung’s foundry is standing at a hinge. The 64-bit chip it just built is the best work it has ever done for Apple, and it may be the last thing it does for Apple at all. The company that taught the world a phone could be 64-bit is about to learn what a foundry is worth when it has nothing left to build.