1. The number that explains the case

In 2013, a smartphone processor had four, six or eight CPU cores. It sold for somewhere between $20 and $60, and it did one job: run a phone. The architecture license Qualcomm signed with Arm that year reflected that reality. For chips with at least five CPU cores, the royalty was capped at $1.88 per unit.

In 2026, a data center processor can carry 288 CPU cores. It sells for thousands of dollars and runs workloads that did not exist when the contract was drafted — model training, inference serving, in-memory analytics, virtualized desktops. The cap, if it still applies, is $1.88.

That is the whole dispute in two sentences. Arm's lawyer made the point directly during cross-examination: under Qualcomm's reading, a chipmaker pays for five cores and receives the other 283 for free. Qualcomm's answer is that the cap is a contract term, freely negotiated, and that Arm cannot rewrite it simply because the product became more valuable.

Both statements are true at the same time. That is why this case has lasted four years, outlived one jury, and now occupies a federal courtroom in Wilmington.

2. What happened in Delaware

On October 5, Qualcomm Technologies' breach-of-contract case against Arm Holdings opened before Judge Maryellen Noreika in the United States District Court for the District of Delaware. The trial was scheduled for five days. The complaint was filed on April 18, 2024, and is docketed as a diversity action for breach of contract — a state-law claim brought in federal court because the parties are domiciled in different jurisdictions.

The posture is unusual because the same two companies litigated a related dispute only two years earlier. In 2022, Arm sued Qualcomm, two of its subsidiaries and Nuvia, the design startup Qualcomm had acquired, alleging breach of license agreements and trademark infringement. Arm also sought an order requiring destruction of certain Nuvia designs. That case reached a jury in December 2024.

The jury found that Qualcomm had not breached the relevant agreements, and that chips developed with Nuvia technology were covered by Qualcomm's own license. It could not reach a verdict on whether Nuvia itself had breached its agreement, and that portion ended in a mistrial. The court subsequently dismissed the remaining architecture-license claims, denied Arm's motion for a new trial, and Qualcomm described the result as a complete final judgment in its favor.

This time Qualcomm is the plaintiff. The subject, however, is unchanged: what Qualcomm owes for designs it acquired, and which generation of Arm's architecture sets the price.

3. Two lawsuits, one argument

The first case was about permission. Arm argued that Nuvia's license did not travel with the acquisition, and that Qualcomm needed a new agreement to sell chips built on Nuvia cores. Qualcomm argued that its own architecture license already covered the work. The jury accepted Qualcomm's position on the central question, and the court's later rulings closed most of what remained.

The second case is about price. Having established that it may sell Nuvia-derived designs, Qualcomm now argues that Arm's conduct during the licensing transition — withholding testing tools, threatening termination, and publicizing that threat — breached the agreement and caused measurable harm. Arm responds that it breached nothing, that no harm occurred, and that the lawsuit is leverage in a commercial negotiation.

The shift matters for anyone who buys semiconductors. A permission dispute threatens a product's existence; a pricing dispute threatens its margin. Existence disputes end in verdicts. Pricing disputes tend to end in contracts — and contracts set the terms that flow down to module makers, OEMs, distributors and eventually to the bill of materials.

Understanding which kind of dispute this is tells you what to watch. If you are waiting for a dramatic injunction, you are watching the wrong signals. The signals that matter are the clause ruling, the bench trial findings, and any sign that the two sides are talking about a tier structure.

4. What Qualcomm is asking for

Qualcomm's requested remedy is the most consequential part of the case, and the least discussed outside the courtroom.

It is not asking only for money. It is asking the court to allow it to stop paying royalties to Arm for up to five years — an amount that could reach into the billions of dollars — and to compel Arm to grant a license to V10, the tenth generation of its compute architecture.

Those two requests solve different problems. The payment pause addresses the past: if Arm breached the contract, Qualcomm argues the remedy should reflect the value of what was withheld, and a royalty holiday is a way to deliver that value without a cash transfer. The V10 license addresses the future: Qualcomm needs access to the next architecture to keep its roadmap intact, and it does not want that access to depend on a negotiation it believes Arm has conducted in bad faith.

There is a strategic logic underneath. A damages award is a one-time event that ends the dispute. A payment pause changes the bargaining position for every negotiation that follows, because Qualcomm's roadmap no longer depends on a near-term deal. Arm's lawyers have characterized that logic as pressure rather than compensation, and that characterization is now central to the defense.

Courts are cautious with remedies of this kind. Specific performance — an order compelling a party to do something rather than pay — is available in contract cases, but judges hesitate to supervise ongoing commercial relationships. A five-year royalty holiday is easier to administer than an injunction, but it still requires the court to accept that the contract's remedy provisions support it.

Compelling a V10 license is the harder of the two requests. Courts can order a party to execute an agreement, but they are reluctant to set its terms, because doing so turns a judge into a contract negotiator. Qualcomm's request is therefore likely to be framed as an order to license on terms the court finds consistent with the existing agreement, rather than an order to license at a price the court invents. That distinction matters: it gives the judge a narrow path to grant access without setting a royalty rate, and it gives Arm a narrow path to comply while preserving its pricing arguments for a later negotiation.

5. Arm's defense: no harm

Arm's opening argument, delivered by its lawyer Gregg LoCascio, took the shortest available path: deny that any damage occurred.

According to courtroom reporting, Arm described Qualcomm's case as an attempt to use litigation as commercial leverage, and argued that Qualcomm had failed to show actual harm. The line Arm pressed was blunt: Qualcomm has not been harmed at all.

The Meta deal is the test of that argument. Qualcomm says a chip deal with Meta lost $170 million in value after Arm notified Qualcomm of a license breach and the notification reached Bloomberg. Arm says the deal was speculative, and that Meta's real reason for changing terms was a shift in its own priorities — from VR headsets toward AI glasses.

Arm also counterattacked on the leak claim, arguing that Qualcomm had itself disclosed non-public details about an antitrust investigation into Arm, and therefore could not complain about the disclosure of a termination letter.

This is a conventional structure in contract litigation: attack causation before damages. If the jury accepts that Meta renegotiated for independent business reasons, the $170 million becomes an opinion rather than a loss. If the jury accepts that Qualcomm leaked first, the leak claim loses its force even if the underlying facts are proven.

6. The $170 million question

Qualcomm's lawyer Karen Dunn told the jury that Qualcomm was close to closing a chip deal with Meta when Arm notified the company that it was in breach of its architecture license. Qualcomm alleges the notification was leaked to Bloomberg. On Qualcomm's account, Meta grew concerned that Qualcomm might lose its Arm license, and the value of the deal fell by $170 million.

Arm's version differs in the part that decides the case. Arm says Meta adjusted the terms because its business focus shifted from VR headsets to AI glasses, and it wanted to recut the commercial terms for Qualcomm's chips. Under that account, the license dispute was background noise rather than cause.

The difference between those stories is the difference between a compensable loss and a negotiation that would have happened anyway. It is also why the trial has spent so much time on documents and sequencing: who knew what, when, and in what order.

For a contract case, the standard is causation plus foreseeability. Qualcomm must connect the disclosure to the price change, and must show the loss was the kind of harm the contract contemplated. Arm's defense is designed to break that chain at both ends: first by arguing the disclosure did not cause the change, then by arguing the change was not the sort of loss the agreement protected against.

7. 1800%: how to read a frightening number

The most striking figure in the trial is not a damages number. It is a percentage.

Qualcomm's lawyers repeatedly displayed charts claiming that, between the ninth and tenth generations of Arm's compute architecture, Arm sought to raise the royalties Qualcomm pays by 1800%. Qualcomm CEO Cristiano Amon testified during the bench phase, and the royalty trajectory was central to that testimony.

The figure is best read as a negotiating position rather than a rate card. Arm has been explicit that it intends to capture more value from newer architecture generations; the v9-to-v10 transition is where that intention becomes visible. Whether 1800% is the right characterization depends entirely on the baseline. A large percentage increase on a small per-chip royalty can look extraordinary while remaining a modest fraction of a chip's selling price.

That ambiguity is what makes the number effective in a courtroom. It is large enough to sound like exploitation and elastic enough that Arm can dispute the arithmetic without disputing the direction.

8. The $1.88 cap: how Arm pricing actually works

Arm sells two different things, and the distinction explains the dispute.

An architecture license lets a customer take Arm's instruction set and design its own cores. Qualcomm holds one. A core license lets a customer take Arm's finished core designs and integrate them. Most of the mobile market uses a mix of both, and the largest customers negotiate both.

Royalties are usually calculated per chip — either as a percentage of the chip's selling price or as a fixed amount per unit, subject to caps, floors and tiering negotiated at signing. Those terms are long-lived. A 2013 agreement can govern products shipped in 2026 unless the parties renegotiate, and renegotiation is precisely what has failed.

That structure produces a predictable conflict. Arm wants each architecture generation to reset the price. Customers want the old terms to keep applying, because those terms were priced when their leverage was different. Every generation becomes a renegotiation, and every renegotiation is an opportunity for the relationship to break.

Qualcomm's position is that Arm tried to use the V10 transition to reset pricing by an order of magnitude. Arm's position is that the existing agreement does not describe what Qualcomm now ships, and that the royalty should reflect the value of the current architecture rather than the value of a 2013 phone.

9. Why per-chip pricing broke

Strip away the personalities and the case is about a pricing unit that no longer matches the product.

A per-chip royalty made sense when a chip was a chip: one die, a handful of cores, a defined function, a stable price band. Today a single package can contain dozens of cores, heterogeneous accelerators, chiplets and stacked memory. The value delivered scales with all of it. The royalty, under a legacy cap, does not.

Columns three and four are illustrative arithmetic, not Arm's rate card. They show why a fixed cap becomes economically strange as core counts and selling prices rise.

The table makes Arm's argument without requiring the jury to accept any particular rate. If the cap was a proxy for value in 2013, it stopped being a proxy once the product changed shape. Arm's cross-examination was designed to make that gap visible: not "Qualcomm is cheating", but "the contract no longer describes the product".

Qualcomm's counter is equally structural. A cap is a cap. If Arm wanted core-count tiering, it could have negotiated it; instead it signed a term that limits its upside and now wants a court to undo a bargain it made when it needed Qualcomm more than Qualcomm needed it.

10. The Nuvia inheritance

The Nuvia acquisition is the origin story neither side can escape.

Qualcomm bought the startup in 2021 for roughly $1.4 billion, acquiring a team of former Apple silicon engineers and a set of custom core designs built around Arm's instruction set. The strategic logic was straightforward: Qualcomm wanted cores it controlled, rather than cores it licensed, so it could compete in PCs and eventually data center without depending on Arm's roadmap.

Arm's original complaint argued that Nuvia's license did not permit transferring those designs to Qualcomm, and that a new license was required. Qualcomm argued its own architecture license covered the work. The December 2024 jury agreed with Qualcomm on the central claim.

But winning that case settled permission, not price. Qualcomm may use the designs; what it owes for them, and under which generation of the architecture, remained open. The current trial is the invoice arriving after the permission was granted.

That sequence explains why Qualcomm seeks a payment pause rather than a refund. It is not arguing it overpaid in the past. It is arguing that Arm's conduct during the transition — withholding tools, threatening termination, publicizing the threat — should change the terms going forward.

11. What a five-year royalty holiday would actually do

A royalty holiday sounds like a windfall. Operationally, it is a reallocation of risk.

For Qualcomm, five years without royalties would lower the effective cost of its compute roadmap and give it room to price aggressively in PCs and data center — two markets where it is challenging entrenched competitors and where every dollar of bill-of-materials cost matters. It would also remove the immediate pressure to conclude a V10 deal on Arm's terms.

For Arm, the effect would be more subtle than a revenue hole. Arm's business model depends on the expectation that each generation earns more than the last. A court-sanctioned pause for its largest customer would undermine that expectation for every other licensee negotiating a renewal. The precedent, not the money, is the larger risk.

For the market, the effect would depend on how the pause is structured. A blanket holiday changes competitive dynamics. A holiday offset against a future rate restores the status quo later. A holiday limited to specific product lines would create exactly the kind of segmentation that both companies say they want to avoid.

This is why the clause ruling matters more than the verdict. If the clause survives, the remedy is live and the settlement calculus changes. If it is struck, the case becomes a conventional damages dispute and the five-year figure disappears from the table.

12. The procedural wildcards

Three unresolved issues could shape the outcome before the jury finishes.

The clause. Judge Noreika is considering whether to strike the contract provision Qualcomm relies on to support the payment pause. If she does, the remedy Qualcomm can pursue may shrink substantially. This is the single most important ruling in the case, and it may arrive before the verdict.

The bench trial. Alongside the jury trial, Judge Noreika is presiding over a bench trial in the same consolidated action, focused on whether Arm negotiated the V10 license in good faith. The same facts are being tested twice, with different standards and different decision-makers. A judge's findings on good faith can survive a jury's view of damages, and vice versa.

The motion. On February 13, 2026, Arm filed a letter asking the court to dismiss Qualcomm's claim that Arm breached the implied covenant of good faith and fair dealing, arguing the claim imposes obligations beyond what the parties agreed and is time-barred. If that claim falls, part of Qualcomm's narrative loses its legal hook even if the facts remain.

Beyond the trial court, an appeal is likely regardless of outcome. A diversity contract case from Delaware goes to the Third Circuit, and the standards for reviewing a jury verdict are deferential — which means the trial record, not the appellate court, is where this case will be decided.

The jury's task is narrower than the public conversation suggests. It is not being asked whether Arm's pricing strategy is fair, or whether per-chip royalties are obsolete. It is being asked whether a specific contract was breached, whether that breach caused a specific loss, and — if the clause survives — whether the requested remedy is available. Jurors will hear about 288-core processors because the number is relevant to interpreting the agreement, not because they are being asked to redesign Arm's business model. That framing is why the case can produce a verdict that satisfies neither side's public narrative.

The burden sits with Qualcomm. In a civil case the standard is a preponderance of the evidence, which is lower than the criminal standard but still requires the jury to conclude that Qualcomm's version is more likely true than not. Arm does not have to prove its counter-narrative; it only has to make Qualcomm's version less convincing. That asymmetry explains why Arm's defense is built around doubt — about causation, about the Meta deal, about whether any harm occurred at all — rather than around an affirmative story about fair pricing.

13. Settlement math

Most licensing disputes of this size end in a contract, not a judgment. The reason is arithmetic rather than sentiment.

Arm needs revenue predictability. A verdict that validates a 2013 cap across generational transitions would damage its pricing power with every large licensee. Even a narrow win would leave the core question unresolved.

Qualcomm needs roadmap certainty. A verdict that gives it money but not a V10 license leaves its next generation exposed to a negotiation it says has been conducted in bad faith. Cash does not solve an access problem.

A tiered royalty is the obvious landing zone: price by core count, by market segment, or by selling-price band. It preserves Arm's data center upside while keeping mobile economics intact. The obstacle is not imagination. It is that any tier structure requires both sides to concede that the old contract is obsolete — which is the one thing neither wants to say in a courtroom.

There is also a timing dimension. A settlement before verdict preserves optionality for both sides; a settlement after verdict is priced off the verdict. That asymmetry is why settlement talks often intensify right before a jury goes out.

14. Four scenarios

The scenario that gets the least attention is the one most likely to occur: a partial plaintiff win that produces damages but not a pricing reset, followed by a negotiated V10 deal with a face-saving structure. That outcome would be reported as a Qualcomm victory and understood by the industry as an Arm victory, because the royalty model would survive.

15. Who pays

Qualcomm's customers. Android phone makers, PC OEMs and cloud operators buy chips whose licensing cost is embedded in the price. A royalty reset of the size Qualcomm describes would eventually appear in prices, roadmap timing, or both.

Other Arm licensees. Every large licensee with a legacy agreement is watching. A Qualcomm win on the payment pause would create a template for challenging old caps. An Arm win would validate generational repricing and make renewals more expensive across the industry.

Arm's investors. Arm's valuation depends on royalty growth per chip, not only unit growth. If courts cap that growth for its largest customers, the growth story changes shape.

The concentration problem. Arm's disclosure that Qualcomm represented 9% of revenue in the fiscal year ended March 31, 2026 is more than a fact about one customer. It is a description of how much of Arm's growth depends on a single company's product decisions. If Qualcomm shifts volume toward its own cores, or delays a platform, Arm feels it directly. That dependency explains why Arm has been willing to litigate so aggressively over pricing: the alternative — accepting a cap that does not scale — compounds every year as the customer's volumes grow.

RISC-V and alternative architectures. Every licensing dispute raises the relative appeal of an open instruction set. That appeal is strongest where switching costs are low — microcontrollers, edge AI, custom accelerators — and weakest where software ecosystems are deep.

Distributors and component buyers. Memory, power management, connectivity and analog parts are designed around specific SoC platforms. When a platform's licensing future is uncertain, qualification timelines lengthen and buffer inventory rises.

16. What buyers should do

This is a legal dispute, not a supply disruption. Nothing has stopped shipping, and no court has ordered a licensing change. But the case is a useful reminder of how platform-level risk propagates.

Map exposure by platform, not by part number. If a design depends on a Qualcomm SoC, the relevant risk is the SoC roadmap, not the price of the PMIC. List the designs where a platform change would force re-qualification, and estimate the cost of that re-qualification.

Qualify a second source before you need one. For commodity functions around the SoC — power, timing, interface — dual sourcing is cheap insurance. For the SoC itself, the realistic alternative is usually a different platform generation rather than a different vendor.

Watch the V10 timeline, not the verdict. The commercially meaningful date is when Arm's next architecture becomes mandatory for new designs. A verdict without a license deal does not change that date; a settlement might.

Do not speculate on inventory. Licensing disputes rarely cause short-term shortages. Building buffer stock against a courtroom outcome ties up capital for a risk that usually resolves in contracts rather than injunctions.

Read the filings. Arm's quarterly reports and Qualcomm's disclosures are published on a schedule and are more reliable than headlines. The useful signals are language about revenue concentration, license renewals and contingencies.

17. The unresolved question

Whatever the jury decides, the industry still needs a licensing unit that scales with heterogeneous computing.

Candidates include royalty per core, royalty as a percentage of selling price with a floor, royalty per performance tier, or a flat platform fee plus a per-unit component. Each has a flaw. Per-core pricing penalizes the architectural direction the industry is moving toward. Price-based pricing is volatile and vulnerable to transfer pricing. Tiered flat fees require someone to define the tiers, and the definition becomes the next dispute.

The deeper issue is that Arm's value proposition has changed. Arm once sold an instruction set and a core design. It now sells an ecosystem, a software stack and a roadmap. Pricing that ecosystem per chip, with a cap written for a 2013 phone, is a mismatch that no verdict can fix.

That is why the most important sentence in this case may not be in the verdict at all. It may be in whatever contract the two companies eventually sign — because that contract will become the template everyone else negotiates against.

18. Timeline

19. FAQ

Is Qualcomm going to stop paying Arm immediately?

No. Any pause requires a ruling in Qualcomm's favor on the relevant contract clause. Until then, the license terms continue to apply.

Has the trial finished?

The jury trial was scheduled for five days beginning October 5. As of this writing no verdict has been announced, and the bench trial on good-faith negotiation is proceeding in parallel.

Does this affect chip supply right now?

No interruption to production or shipments has been reported. The dispute concerns licensing terms and money, not an injunction against selling chips.

What is V10?

V10 is the tenth generation of Arm's compute architecture, following v9. Qualcomm wants a guaranteed license to it so its roadmap does not depend on the outcome of commercial negotiations.

Why does $1.88 matter so much?

It is the per-chip royalty cap in the 2013 agreement for chips with at least five CPU cores. Arm argues a cap designed for small phone chips cannot sensibly apply to 288-core data center processors.

What did the 2024 jury decide?

That Qualcomm did not breach the relevant agreements and that chips built with Nuvia technology were covered by Qualcomm's license. It could not agree on whether Nuvia itself breached, and that portion was mistried.

Could this end in a settlement?

Yes, and that is a common outcome. A tiered royalty — priced by core count or market segment — is the most likely landing zone, because it preserves Arm's data center upside while keeping mobile economics intact.

What should distributors watch?

Any change in Arm's revenue guidance tied to Qualcomm, and any language in Qualcomm's filings about license terms or contingencies. Both are published quarterly.

20. Glossary

• ALA (Architecture License Agreement) — a license allowing a company to design its own CPU cores using Arm's instruction set.
• TLA (Technology License Agreement) — a license to use Arm's finished core designs and related IP.
• ISA (Instruction Set Architecture) — the interface between software and hardware; defines the instructions a processor executes.
• Royalty cap — a contractual ceiling on the per-chip payment, often a fixed dollar amount.
• Bench trial — a trial decided by a judge rather than a jury.
• Implied covenant of good faith and fair dealing — a doctrine requiring honest performance of a contract even where the text is silent.
• Diversity jurisdiction — federal jurisdiction based on the parties' domiciles rather than a federal question.
• Specific performance — a remedy ordering a party to perform an obligation rather than pay damages.
• Mistrial — a trial ending without a verdict, often because the jury cannot agree.

21. Sources

• Arm Holdings investor relations and SEC filings: investors.arm.com
• Qualcomm newsroom and investor disclosures: qualcomm.com/news
• U.S. District Court for the District of Delaware: ded.uscourts.gov
• SEC EDGAR company filings: sec.gov/edgar

22. Related reading

For context on how licensing fights and platform shifts move through the component supply chain, see our analysis of the Vicor Section 337 investigation and our review of the four signals reshaping the 2026 chip supply chain.