Semiconductor Verification After the Arm–Qualcomm Verdict: Why Two Dollars of Damages Rewrites the Compliance Playbook

For a hardware team planning tape-out, compliance sign-off is the last gate before a processor can be marketed as what it is. That gate is normally invisible in reporting on licensing disputes, which is why the $2 damages figure has been read as the story. For the Electronicseng Semiconductors audience — silicon architects, verification leads, component buyers and supply-chain owners — the usable story sits somewhere else: a jury held that the party writing the tests can, contractually, be forced to hand over the tooling those tests depend on.

What follows unpacks that finding as an engineering and sourcing problem rather than a courtroom scoreboard: which artifacts are involved, how a gap in them shows up on a verification schedule, what the royalty clause in Qualcomm’s license actually implies, how the alternative architecture scores on the same axis, and what a custom-core team should insist on in writing before it commits.

Read the finding, not the damages line

On Oct. 13 a federal jury in Delaware returned the second verdict in two years in the long-running dispute between Arm and Qualcomm. Attention locked onto the award: $2, one for each claim Qualcomm won. Both sides could fairly claim something — Arm prevailed on four of the six claims, and Qualcomm’s result included an interference finding involving Meta. The operative determination for anyone who designs cores is different: the jury found that Arm breached Qualcomm’s Architecture License Agreement by withholding two deliverables, known in the trade as ACKs and OOBs.

Those acronyms are exactly why coverage drifted toward royalties. They name nothing abstract, though: they identify the machinery a custom core must pass through before anyone is entitled to call it an Arm processor.

Two ways to buy an architecture — and only one of them needs a referee

Arm licenses its technology in two ways, and the compliance question only bites on the second one.

Dimension	Technology License Agreement (TLA)	Architecture License Agreement (ALA)
What the licensee receives	Arm-designed cores taken “as is” and placed inside the customer’s chip	Permission to design its own cores executing the Arm instruction set
Who owns the microarchitecture	Arm	The licensee
Where behavioral risk sits	Largely inherited from Arm’s own design	Must be demonstrated by the licensee, including obscure corner cases, or Arm-targeted software eventually fails
Dependency on compliance artifacts	Limited	Total: tests, test-selection data and test patches gate sign-off
Who takes this path	Many customers	A smaller group, including Apple, Qualcomm, and Nvidia

Read as a sourcing decision, the ALA trades freedom for evidence. You get a microarchitecture of your own; in exchange you must prove it behaves as the architecture specifies, and the proof is only accepted if it is produced with the licensor’s own reference instrumentation.

What the compliance kit actually controls

A licensee demonstrates conformance with the Architecture Compliance Kit, or ACK. Qualcomm’s filing says its engineers run roughly 48,000 ACK tests. Tirias Research, an analyst firm that reported from the courtroom daily and relied on Qualcomm’s pre-trial statement of proof, explained the division of labor: out-of-band (OOB) packages identify which tests apply to a particular processor design, while ACK patches repair flaws in the tests themselves. Qualcomm alleged that Arm withheld these materials from late 2022 into 2025.

Artifact	Function inside the verification flow	Symptom if it is missing
ACK suite	The tests a custom core must satisfy to be called an Arm processor	Sign-off cannot be closed; compatibility claims cannot be evidenced
OOB packages	Select which tests apply to a given design	The team runs the wrong test set, or silently skips required cases
ACK patches	Correct defects in the tests themselves	Failing tests point at silicon that is not at fault
Engineering support	Interpretation of ambiguous failures	Arm argued this layer was not a mandatory deliverable; the jury rejected that reading on Qualcomm’s facts

Schedule physics: why a test gap is not a paperwork gap

Anyone who has run a large verification team can see the mechanism. When a test fails, the first question in triage is whether the fault lives in the design or in the test. Across programs built on tens of thousands of tests, custom microarchitectures and many person-years of engineering, a defective test with no patch available can waste weeks of engineering time chasing a problem that does not exist in the silicon.

Aggregated, that becomes a strategic exposure rather than an inefficiency. As Tirias Research observed, licensing decisions made today shape products that reach customers several years later. Arm’s counter-position was that the underlying ACK remained available, that quarterly releases continued, and that the additional OOBs, patches and engineering support were not mandatory deliverables under the contract. The jury sided with Qualcomm on the contract question, but the underlying issue is still fair to argue — Tirias framed it sharply by asking when withholding support effectively becomes withholding part of the product.

The royalty clause sitting behind a two-dollar number

Tirias Research describes a far larger exposure in Section 8.3 of Qualcomm’s architecture license: if Arm breaches certain obligations and fails to cure within 30 days of written notice, royalty rates on qualifying products can fall to zero for five years. Qualcomm says it gave notice of these breaches in November 2022, repeated it in December, and received a denial from Arm.

The dispute’s largest possible remedy — one Arm itself values in billions of dollars — rests precisely on the test-deliverable breach. It does not depend on the Meta interference finding, on access to Arm’s v10 architecture, or on any pricing dispute. It could have financial consequence only if a court finds both that the clause covers Qualcomm’s products and that it is enforceable, and both questions are open: Arm maintains the provision is unenforceable, while Judge Maryellen Noreika must decide whether it applies.

According to Bloomberg, she heard closing arguments in the parallel bench trial on Oct. 20. Bloomberg also noted her reluctance to involve the court in setting licensing terms and her description of Qualcomm’s request that the court order Arm to negotiate v10 terms in good faith as “very, very drastic.” That request is separate from the royalty clause, since it asks whether Arm negotiated v10 terms in good faith, and a ruling on one issue does not necessarily settle the other. Both are now before the judge, with a written ruling to follow. Neither outcome is predictable — but a verification issue is already sitting at the center of the money question.

Referee and competitor inside one company

A structural reason explains why control of these artifacts matters commercially. The Arm model concentrates the roles in one company: that firm writes the architecture specification, writes the compliance tests, chooses which tests apply to which design, corrects the tests when they are wrong, and acts as the authority on whether a core is compliant. Arm’s own 2022 complaint shows the last role, describing Qualcomm seeking Arm’s verification that a new Qualcomm core complied with the architecture so it could be incorporated into a product; on timing, that core was most likely the Nuvia-derived design that later became Oryon.

Other ecosystems split those roles. For PCI Express and USB compliance, the process runs through industry bodies, PCI-SIG and USB-IF, where competing vendors use a common test process that no single competitor owns. Arm’s arrangement persisted for decades because Arm was called the industry’s “Switzerland,” a word that came up in court, and because a neutral supplier holding the referee’s whistle raises few concerns.

That neutrality is shifting by Arm’s own account. During trial testimony, Rene Haas was asked whether roughly two-thirds of Arm’s revenue would come from chips by 2030; he said yes, adding that Arm is “adding to our business a chip business.” He was equally direct about who should set the price of Arm’s technology in licensing negotiations: “It’s our product. Of course, we should decide.” He was speaking about pricing, not compliance, but the same logic of ownership applies to both.

Competing with customers is often complicated, yet not automatically misconduct. Still, once the architecture owner is also a chip vendor, the compliance toolchain becomes a control point over how quickly a competitor’s custom core can reach the market. That is likely why such a narrow contract claim ended up carrying so much weight, and why the FTC may find it relevant too. The FTC is not a party to this case, but it reportedly opened an antitrust investigation into whether Arm is illegally trying to dominate parts of the market by refusing or degrading licenses.

By the numbers: figures on the record

• roughly 48,000 ACK tests — the volume Qualcomm’s filing says its engineers run
• $2 in damages — one for each claim Qualcomm won
• four of the six claims resolved in Arm’s favor
• late 2022 into 2025 — the window over which Qualcomm alleged the deliverables were withheld
• Oct. 13 jury verdict; Oct. 20 closing arguments in the parallel bench trial, as reported by Bloomberg
• roughly two-thirds of Arm revenue projected to come from chips by 2030, agreed on the stand by Rene Haas

RISC-V: neutral governance, unfinished trust layer

RISC-V is the standard answer for teams that want to stop depending on a single architecture owner. It removes the need for a proprietary architecture license from that owner, although individual RISC-V cores and implementations can still carry their own intellectual property and licensing constraints. The licensing picture is more mature than the compliance picture.

RISC-V International’s own documentation describes its Architectural Compatibility Test suite as a minimal filter, and states that passing it does not mean a design complies with the architecture. To keep their compatibility listing, cores must re-pass the tests yearly whenever the suite changes. A Certification Steering Committee has been formed to build a formal certification program, but that work is still in progress. Meanwhile the RVA23 profile is becoming the practical target: Ubuntu now requires it, while Linux is only beginning to offer a simple way to report whether a system is RVA23 compliant.

Axis	Arm	RISC-V
Architecture governance	One company, which is also entering chip sales	Community-governed, no proprietary architecture license needed
Compliance program	Mature and rigorous; single authority declares conformance	ACT suite described as a minimal filter; formal certification still under construction
Test maintenance	Quarterly releases plus patches and OOB selection packages	Re-pass yearly whenever the suite changes, to keep the listing
Software signal today	Broad ecosystem built around the architecture	RVA23 as the practical target: required by Ubuntu, only beginning to be reported by Linux

A designer selecting an architecture therefore faces a tradeoff rather than a clean escape: Arm offers mature, rigorous compliance under one company’s control, while RISC-V offers neutral, community-governed compliance that is still being built. Qualcomm’s acquisition of Ventana, and the recently reported addition of the Akeana team, reads as a bet that this gap will close, or at least as insurance against further disputes with Arm. The decisive question for RISC-V in high-performance computing is less about the instruction set than about who builds trusted compliance and certification at scale, and who can match the broad software ecosystem Arm has built around its architecture.

Five misconceptions this verdict exposes

1. “A $2 award means the case is trivial.” The amount is symbolic; the exposure lives in Section 8.3, which Arm itself values in billions of dollars if it applies and is enforceable.
2. “A compliance kit is documentation.” It is gating instrumentation: without the applicable test-selection data and patches, a custom core cannot be signed off, regardless of how good the RTL is.
3. “Continued quarterly releases prove nothing was withheld.” That was Arm’s argument; the jury rejected it on Qualcomm’s contract claim, while the broader question of when support becomes the product is still debated.
4. “RISC-V means full independence.” It removes the architecture license, not other IP and licensing constraints carried by individual cores and implementations.
5. “A ruling on the royalty clause settles the v10 argument.” They are separate questions before Judge Maryellen Noreika; one does not necessarily resolve the other.

Checklist: what a custom-core team should put in writing

• Inventory the compliance artifacts your flow actually consumes — suite, applicable-test selection data, patches, interpretation support — and name the owner of each.
• Require the licensor to list those artifacts as deliverables in the agreement, not merely reference “the kit” as an abstract entitlement.
• Fix release cadence, notice language and cure mechanics in the contract, so a delivery dispute has a defined path instead of becoming litigation.
• Budget test-defect triage explicitly; track and report the weeks of engineering time lost to suspected bad tests as a program risk line.
• Maintain an independent oracle — directed tests or formal models of critical corner cases — so a single failing reference test is not fatal to the schedule.
• Treat architecture-generation access as a tooling question, not only a royalty question: newer generations carry their own compliance kits.
• If you are evaluating a second architecture, test against the current practical profile and verify the freshness of any compatibility listing before qualifying a supplier.

FAQ for engineering and procurement readers

Was the damages number the point of the verdict?

No. The $2 award reflects one for each claim Qualcomm won. The consequential determination was the contract breach over two deliverables, because that breach is the hook on which the Section 8.3 royalty remedy hangs.

In practical terms, what does an OOB package do?

Per the description in the record, it identifies which tests apply to a particular processor design, while ACK patches fix flaws in the tests themselves. Without them, a team can neither scope its regression correctly nor trust its failures.

Does Arm’s “not a mandatory deliverable” defense still matter?

It failed before the jury on the contract question, but the design question it raises — how much of a toolchain is the product — is unresolved, and Tirias Research framed it as the issue worth watching.

Is certification the same thing as compliance?

Not today in RISC-V: passing the compatibility test suite is explicitly described as a minimal filter, not proof of compliance, and a formal certification program under the Certification Steering Committee is still being built.

Should a team hedge now?

For any roadmap that depends on a custom core, the prudent step is the same one the jury just priced: make the compliance toolchain an explicit, contracted supply item, and evaluate a second architecture path on its certification maturity rather than its instruction set.

Three things to watch next — and what to do about them

1. Judge Maryellen Noreika’s ruling on Section 8.3 and on whether Arm negotiated v10 in good faith.
2. v10 itself, because access to a new architecture generation includes access to its compliance kit, so the eventual terms will cover tests as well as royalties.
3. RISC-V certification, because a credible and independent program would change the calculus for every company weighing a custom core.

Only the first is a court decision, and its timing is not yet known; the other two play out over a longer period. Practical guidance in the meantime: verification leads should confirm that every artifact their flow consumes is named in the license; procurement and supply-chain owners should treat compliance tooling as a long-lead input with a documented fallback; and analysts should read the royalty clause as a verification-linked exposure, not a pricing dispute.

The Two Dollars headline will fade quickly, but the finding underneath it should stay in the planning documents of anyone shipping a custom core: whoever writes the test suite controls the path from design to a shipping product, and this trial put an unusually concrete value on that control.