The Apple Broadcom deal showed up first as an 8-K filing on July 7, 2026, and the number inside was $30 billion. Apple’s press release landed the next day, the normal choreography for a material customer agreement: the SEC filing goes first, the marketing follows. The commitment runs through 2031, it’s the biggest single item under Apple’s American Manufacturing Program so far, and it does several jobs at once for two companies with very different reasons to want it.
The filing describes new multi-year agreements for "custom ASIC silicon products for use in multiple generations of Apple products." That phrasing is calibrated. Precise enough to mean something, loose enough that you can read a lot into it, which is exactly what everyone did.
None of it is a new building. The physical anchor is Broadcom’s Fort Collins facility in Colorado, which has been making FBAR filters for Apple for years. Broadcom is putting $1.5 billion of capex into expanding and modernizing it, and Apple’s committed volume is the thing that makes that spend safe. Hock Tan and Tim Cook both showed up for the announcement. Cook thanked the Trump administration by name, the White House called it "another major win for America," and Apple, a company that normally says as little politically as it can get away with, went along with all of it.
The production numbers are real: more than 15 billion U.S.-made chips over the term, "hundreds" of American jobs. Apple folds this into the $600 billion, four-year domestic investment pledge it announced in 2025. It has never broken that figure into new money versus spend it was going to make anyway, and that gap is worth keeping in mind.
FBAR filters are the established piece of those 15 billion chips. They’re radio-frequency components that use thin piezoelectric films to resonate at exact frequencies and strip interference out of the cellular radio, small, brutally hard to manufacture, and non-negotiable for signal quality in every iPhone. Broadcom has been one of Apple’s primary FBAR suppliers for years, up against Murata, TDK, and Qorvo, all of whom lean heavily on Asian fabs.
The custom ASIC line is the more interesting part. An application-specific integrated circuit is a chip built for one defined workload instead of general-purpose computing. Apple already designs its own A-series and M-series application processors, fabbed at TSMC, but ASICs tuned for connectivity and RF signal processing are a different design world. "Multiple generations of Apple products" tells you these aren’t one-off parts; there’s a sustained roadmap here reaching to at least 2031. Wireless connectivity rounds out the scope, the cellular, Wi-Fi, and Bluetooth integration Broadcom has supplied to Apple for a long time, now deepened and pulled onto U.S. soil.
Tariff exposure is the most immediate force behind the timing. The Trump administration has kept an aggressive posture on electronics components out of China and, to varying degrees, other Asian hubs. RF filters have historically come out of Japan and elsewhere in Asia, so moving them to Fort Collins sidesteps tariff risk on a part that ships in the billions of units a year. Run the arithmetic at Apple’s volumes and even a modest per-unit tariff on RF filters piles up into hundreds of millions of dollars annually.
That commercial logic doesn’t run separately from Cook’s exit. He built Apple’s supply chain into a weapon over his tenure, and locking a domestic silicon partnership of this size with the White House applauding is the kind of thing a departing CEO wants on his ledger. His public thanks to the administration reads as deliberate, a deal engineered to serve operations and politics in the same paragraph.
The AI angle sits underneath both. Custom ASICs are the same category powering the AI accelerator market: Google’s TPUs, Amazon’s Trainium, Meta’s MTIA are all ASICs. Apple’s Neural Engine already handles on-device inference inside the A-series, and as those workloads get heavier and spread across more of the device, the appetite for purpose-built silicon only grows. A supply relationship locked through 2031 lets Apple iterate on inference hardware across product generations without reopening supplier terms mid-cycle.
There’s a catch that cuts against the "Apple owns its silicon now" narrative. The independence story is genuine as far as it goes: the 2020 Apple Silicon transition brought Mac CPUs in-house, the 2019 purchase of Intel’s smartphone modem business is now shipping as the C1 modem in the iPhone 16e, and chip design has been steadily migrating inward across the stack. But Apple isn’t designing these chips. It’s commissioning custom ASICs from Broadcom, which keeps the IP and the fab process while Apple gets parts built to spec. Apple owns the specification and the roadmap commitment; Broadcom owns the IP and the process. Apple gets supply certainty and design continuity without pouring capital into its own RF manufacturing, a domain where Broadcom’s decades of process expertise would be slow and expensive to copy.
The deal also pulls Apple away from Murata, TDK, and Qorvo in the filter segment, all three of whom feed Android OEMs too. A dedicated U.S. supply line with Broadcom hands Apple a differentiation rivals can’t stand up on short notice, because FBAR isn’t a commodity process. It needs specialized deposition equipment and process control that take years to build.
The $30 billion, spread over roughly five years, is about $6 billion annually against Apple revenue north of $380 billion, so this doesn’t strain the budget. It does concentrate a meaningful slice of component spend on one supplier in one place. Apple is trading multi-sourcing flexibility for pricing certainty and supply assurance, and it decided the certainty was worth more.
For Broadcom the math is different, and better. Apple has historically been something like 20% of Broadcom’s total revenue, and a contract through 2031 turns a relationship-dependent stream into something close to a booked backlog. The $1.5 billion Fort Collins capex stops being a bet. Broadcom can plan equipment, hiring, and process work against a known demand curve instead of a pile of design-win probabilities, a serious advantage in a business where fab capacity gets committed 18 to 36 months ahead of production. AVGO rose on the news for the obvious reason: revenue certainty at scale trades at a premium. It also reinforces Broadcom’s standing as a custom silicon shop across markets, alongside its aggressive push into AI networking silicon and custom AI ASICs for Google and Meta.
Broadcom’s exposure is the same trap that already sprang shut on the modem side. Apple has a habit of internalizing chip design once it has enough expertise and volume in hand. It bought Qualcomm modems for years while quietly building its own, then acquired Intel’s modem unit, and now ships the C1. If Apple eventually decides to design its own filters or develop an in-house FBAR process, maybe through an acquisition, the 2031 contract buys Broadcom a runway, not a moat. When Hock Tan leaned on "decades of success" in his statement, part of that was a nod to switching costs as much as to friendship.
The Fort Collins framing deserves the hardest scrutiny for the same reason. The site has been making FBAR filters for years and already staffs process engineers and technicians who know piezoelectric thin-film deposition. The $1.5 billion expansion scales and modernizes a running operation instead of building from dirt, so the ramp should beat a greenfield timeline, though neither company has said when the new capacity comes online. The political pitch leans on American jobs and domestic resilience, but the plain fact is Fort Collins was already doing this work. The investment is real and the expansion is real; calling it a fundamental reshoring of chip production oversells the structural shift. What’s actually new is the committed volume, the ASIC scope, and the contract length, not the address.
Set against the $600 billion pledge, this $30 billion is 5% of the total and the largest single item disclosed to date. The pledge itself still carries the open question of whether it’s net new money or capital Apple would have spent in the U.S. regardless, and Apple hasn’t answered it. The Broadcom deal at least comes with a dollar figure, a named counterparty, a specific facility, and an SEC filing behind it, which puts it on much firmer ground than the aggregate number.
Read it as one thing and you’ll get it wrong. It’s a supply chain decision, a tariff hedge, a political set piece, and a multi-generation silicon roadmap all at once. Watch the ASIC scope hardest, because that’s the line where Apple has quietly walked away from a supplier before, and Broadcom just bet $1.5 billion that it won’t do it again.