Semiconductor Patents in the Onshoring Era: Why Domestic Fabs Change Your IP Strategy
By Brandon Furdock, Registered Patent Attorney, Chicago Patent Law LLC
The semiconductor industry is living through two resets at once. Fabrication capacity is moving onshore. And the USPTO has rebuilt how it examines patents and, more sharply, how it decides whether granted patents can be challenged. Most commentary treats these separately. For companies in the semiconductor value chain, the strategy lives where they intersect, and as of this month the intersection has numbers attached to it.
What changed, concretely
On March 11, 2026, Director John Squires issued a memorandum adding three considerations to the decision whether to institute an inter partes review or post-grant review:
- The extent to which the products accused of infringement in parallel litigation are manufactured in the United States or relate to investments in American manufacturing.
- The extent to which competing products made, sold, or licensed by the patent owner are manufactured in the United States.
- Whether the petitioner is a small business that has been sued for infringement of the patent at issue.
It took effect immediately, reaching every pending proceeding in which the patent owner's discretionary brief was not yet due. The definitional detail matters more for this industry than the headline. Manufacturing is not final assembly. Parties may address where components are sourced and whether U.S.-made products are sent abroad for further processing. For method claims, the relevant product is the device used to carry out the method. Small businesses are identified by the SBA's size standards (13 C.F.R. §§ 121.801 to 121.805) and the small-entity fee rules (37 C.F.R. § 1.27(a)).
The memo did not arrive alone. Since the October 17, 2025 memorandum, the Director personally decides every institution question, merits and discretion alike, with routine outcomes issued as summary notices and written decisions reserved for novel issues. Settled expectations, which weighs how long a patent has been in force, is now a standing consideration. And on May 14, 2026, in Magnolia Medical Technologies v. Kurin, IPR2026-00097, the Director designated as precedential the rationale under all of it: AIA review exists as an alternative to district court litigation, not an addition to it. Magnolia also catalogues the considerations the Office now treats as established, and considerations tied to foreign sovereign petitioners and to domestic manufacturing sit in that catalogue. The March memo is no longer a freestanding policy statement; it is part of the precedent.
Which asymmetry, exactly
The quick reading is that the March memo is a win for patent owners. It is not that simple, and acting on the quick reading sends you into discretionary briefing with the wrong evidence.
Two of the three factors point toward institution. The accused-products factor helps a challenger whose own products are American-made. The small-business factor is an open door for defendants using the PTAB defensively. Only the competing-products factor runs the patent owner's way, and only for a patent owner that actually sells something.
So the line does not run between owners and challengers. It runs between companies that manufacture in the United States and companies that do not, and it favors the former on whichever side of the caption they sit. The test is manufacturing activity, not corporate nationality. An American company producing entirely offshore can find these factors working against it; a foreign-headquartered company with substantial U.S. operations can find them working in its favor.
Practitioners will recognize the shape. The International Trade Commission has always conditioned relief on a domestic industry under 19 U.S.C. § 1337(a), a requirement rooted in a trade-protection mission. No other patent forum imposed anything comparable until now.
A caveat about magnitude, with the current numbers
These factors operate at the margin of a system that has become very hard to enter, and the data through the third quarter of fiscal 2026 show how hard.
AIA petitions filed in FY2026 to date number 395, against 1,433 in all of FY2025 and 1,288 in FY2024. June 2026 saw 22 petitions; January 2025 saw 131. Of petitions decided under the current framework, roughly 60% were denied by the Director on discretionary grounds before the merits were reached; the effective institution rate across all decided petitions sits near 38%, though petitions that clear the discretionary gate are instituted about three times in four. Meanwhile ex parte reexamination requests hit an all-time quarterly high of 336 in the second quarter of 2026, and reexamination made up about three quarters of all post-grant filings in the first half of the year, with IPRs barely a fifth.
Read those together. A strong domestic-manufacturing record is not a key that opens the door. It is one input into a judgment that, at present, comes out the same way in most cases. Build the record to preserve optionality and to improve your standing in the narrow band of petitions that still get through. Do not build a litigation strategy that assumes the PTAB will be available to you.
What this means by company profile
If your products are manufactured in the United States, your patents became more defensible and your defensive posture became stronger at the same time. That is an unusual combination, and it belongs in licensing negotiations, assertion decisions, and how you document your footprint. The paper trail connecting patented technology to U.S. manufacturing, meaning which products, which facilities, which investment, is now institution-stage evidence. Assemble it before you need it. Discretionary briefing runs on a short clock, and supply-chain facts are slow to pull from procurement and operations.
If your products are manufactured abroad, the same asymmetry cuts against you, and the response is upstream: deeper freedom-to-operate work at design time, earlier design-arounds, proactive validity work, and a franker assessment of assertion risk in the business plan. Given how narrow the gate has become for everyone, this is less a new disadvantage than an acceleration of a problem you already had.
Fabless is not the same as foreign-manufactured
This is the most common misreading of the memo, and for semiconductors it is the one that matters.
The accused-products factor asks where the products are made, not who owns the fab. A fabless designer taping out at a domestic foundry has a real manufacturing story to tell, which is the entire premise of the CHIPS-era buildout. Combined with the Office's statement that manufacturing extends to component production, a fabless company may be considerably better positioned than its business model suggests.
That surfaces the hard case. Advanced packaging, assembly, and test remain overwhelmingly offshore. A die fabricated domestically and packaged in Asia is an ordinary fact pattern, and it sits exactly on the memo's language about domestic products sent abroad for further processing. No decision has yet said how much that costs. If your products follow that path, treat it as an open question and prepare evidence on both sides of it rather than conceding it.
The prosecution side: hardware is having a moment
Examination favors the sector. Device structures, process improvements, and packaging innovations rarely draw a § 101 fight, which consumed software prosecution for a decade. And the technologies driving current filings sit where fab investment and patent value compound: advanced packaging and chiplets, high-bandwidth memory, AI accelerators, thermal and power delivery for dense compute.
Timing deserves a careful read. On April 10, 2026, the USPTO announced that first office actions within the fiscal year were exceeding new filings for the first time in nearly a decade, with unexamined inventory at 776,995, down from the January 2025 peak of 837,928 and projected to keep falling through the fiscal year. Backlog and pendency are different measures, and improvement in the first shows up in the second only with a lag; total pendency has not yet turned. Check the Office's current dashboard before building a filing plan around either.
Two prosecution disciplines pay outsized dividends here.
Claim across the value chain. A semiconductor invention rarely lives at one layer: the device structure, the process that makes it, the equipment configuration, the resulting module, the system that incorporates it. Each is a different claim, a different infringer, and a different venue, including the ITC, whose importance grows as PTAB relief narrows. Portfolios claimed at a single layer leave most of their leverage unclaimed. And the memo's method-claim rule, which looks to the device used to perform the method, means claim architecture now shapes which manufacturing facts are even relevant.
Draft for detectability. Process innovations are hard to police because you cannot see inside a competitor's fab. The craft response is to claim the fingerprints a process leaves in the finished device, measurable by teardown. Where no fingerprint exists, the patent-versus-trade-secret decision deserves real weight. A recipe with no detectable trace is often better kept than published.
For the AI-hardware companies
The most valuable filings in the current wave sit at the hardware-software boundary: compiler and scheduling techniques for accelerators, memory-hierarchy management for training workloads, sparsity and quantization implemented in silicon, interconnect protocols for distributed training.
These are the exception to the eligibility comfort above. They earn their eligibility from architectural anchoring, not from sitting in a hardware technology center. The discipline is tying the claim to the silicon, meaning the memory hierarchy, the datapath, the physical constraint the technique exists to solve, rather than reciting the algorithm and hoping the hardware context carries it.
They are also the most awkward organizationally, because they fall between the hardware team's invention process and the software team's, and go unfiled as a result. If invention harvesting does not deliberately trawl that boundary, competitors' portfolios are being built out of your gap.
How durable any of this is
The framework rests on memoranda and designated decisions, not rules. In Apple Inc. v. Squires, decided February 13, 2026, the Federal Circuit held the Director's institution guidance exempt from notice-and-comment rulemaking because it is a general statement of policy that does not bind the Director. That insulates the framework from procedural challenge. It also means a future Director can rescind or rewrite it without process, exactly as the prior administration's 2022 guidance was discarded in early 2025. Because 35 U.S.C. § 314(d) makes institution decisions final and nonappealable, mandamus petitions against summary denials have gained no traction, and the question of what review remains has reached the Supreme Court's docket more than once without being taken up.
Rulemaking that would codify much of this has been proposed but not finalized. The October 2025 proposal would require a broad stipulation as a condition of institution and would bar IPR where claims were upheld in almost any other forum, including ex parte reexamination. A final-rule package was at the Office of Information and Regulatory Affairs as of July 22, 2026, and remains there.
That last point matters, because reexamination is the workaround everyone now recommends, and the numbers above show everyone taking it. It remains free of the discretionary gate. But the Office is visibly attending to the migration: on July 22, 2026, it proposed requiring third-party requesters to identify every real party in interest, which would end anonymous filing and give the Office a mechanism for estoppel. Treat reexamination as available now rather than as a durable structural answer, and confirm its status the week you rely on it.
The takeaway
Semiconductor IP strategy used to be a closed calculation: file broadly, cross-license, keep the peace. The onshoring era adds a variable. Where things are made now shapes what patents are worth, and the current USPTO has made that explicit and, for the moment, made it evidence.
The companies adjusting fastest treat their manufacturing footprint, their claim architecture, and their post-grant posture as one question. That was always the right instinct. Policy has just made it visible.
Brandon Furdock is a registered patent attorney and computer engineer. Chicago Patent Law LLC, 33 N Dearborn St #1000, Chicago, IL 60602 · (312) 800-1843 · chicagopatent.law. This article is general commentary, not legal advice.