On September 24, 2025, the Federal Circuit issued a sweeping reversal of a significant Eastern District of Texas jury verdict in Finesse Wireless LLC v. AT&T Mobility LLC, vacating a $166 million damages award and reversing the denial of judgment as a matter of law (JMOL) of noninfringement. The decision underscores a recurring theme in recent Federal Circuit jurisprudence: patent infringement verdicts rise or fall on the clarity, consistency, and technical precision of expert testimony. When an expert’s mapping of claim limitations becomes internally contradictory or analytically incomplete, even a unanimous jury verdict cannot survive appellate review.
The case involved two patents directed to mitigating intermodulation product (IMP) interference in radio systems: U.S. Patent Nos. 7,346,134 and 9,548,775. Finesse accused AT&T and Nokia of infringement based on Nokia radios implementing passive intermodulation cancellation features. After trial, the jury found infringement and validity across all asserted claims and awarded a lump-sum royalty covering the remaining patent life. The district court denied post-trial motions for JMOL and a new trial. On appeal, the Federal Circuit reversed across the board.
The ’134 Patent: Sampling Signals That Do Not Yet Exist
The court’s analysis of the ’134 patent focused on a fundamental claim requirement: the accused receiver must sample, at the front end, a passband containing both “signals of interest” and “interference generating signals.” Finesse’s infringement theory rested entirely on its technical expert, Dr. Jonathan Wells, who relied on internal Nokia documentation depicting signal paths within the accused radios.
At trial, Dr. Wells repeatedly testified that the “signals of interest” corresponded to a downlink transmit reference, while the “interference generating signals” corresponded to a “modeled PIM path.” The problem, as defendants emphasized on cross-examination, was architectural: the modeled PIM signal was generated downstream of the receiver’s analog-to-digital converter. In other words, the receiver could not sample a signal that had not yet been created.
Confronted with this issue, Dr. Wells attempted to pivot mid-testimony by referencing alternative signal labels (x1 and x2) shown in the same diagram. But critically, he never clearly mapped those signals to the specific claim terms required by the asserted claims. Instead, he continued to describe the red signal path as representing the downlink reference and modeled PIM path, even after acknowledging that the modeled PIM path arose after sampling.
The Federal Circuit found this testimony fatally inconsistent. Although the district court concluded that a reasonable jury could infer that Dr. Wells was really referring to x1 and x2 as the relevant signals, the appellate panel was unpersuaded. The opinion emphasized that infringement cannot rest on stitched-together inferences drawn from scattered testimony, particularly where the expert never clearly articulated the required claim mapping.
Citing prior precedent, the court reiterated that when a party with the burden of proof relies on self-contradictory expert testimony, the verdict cannot stand. The court characterized Dr. Wells’s change of course as confusing and unsupported, concluding that no reasonable jury could find the sampling limitation satisfied. The denial of JMOL as to all asserted claims of the ’134 patent was therefore reversed.
The ’775 Patent: Seven Multiplications Means Seven Multiplications
The ’775 patent presented a different, but equally unforgiving, problem. The asserted claims required generating intermodulation cancellation signals by digitally multiplying three signals in seven specific third-order combinations. The infringement case again turned on Dr. Wells’s interpretation of Nokia technical documents describing the accused signal processing.
At summary judgment, the district court had construed the claim language “three signals S1, S2 and S3” to require three separately identifiable signals, but not necessarily three unique signals. On appeal, defendants did not challenge that construction. Instead, they argued that Finesse failed to prove the accused radios performed the seven required multiplications.
The Federal Circuit agreed. At trial, Dr. Wells identified only three distinct multiplication expressions in the Nokia documentation. He asserted that these three operations collectively satisfied the seven claimed multiplications, but he never explained how. Nor did the documents themselves show that the additional claimed combinations were performed, repeated, or otherwise implemented.
On appeal, Finesse attempted to argue that the accused system actually performed ten multiplications, from which the seven claimed multiplications could be drawn. But the court rejected this effort outright. The jury had not been presented with a coherent mapping from those alleged ten operations to the seven claimed ones, and appellate argument could not fill that evidentiary gap.
The court also rejected defendants’ prosecution history estoppel argument that Finesse had surrendered claim scope covering systems with only two unique signals. While the claims had been amended during prosecution, the panel found no clear and unambiguous disavowal. That issue, however, did not save the verdict. Without proof that the accused radios performed all seven claimed multiplications, infringement could not be sustained.
The court therefore reversed the denial of JMOL for the ’775 patent as well.
Damages Fall With Liability
Because the Federal Circuit reversed the infringement verdicts for all asserted claims of both patents, the damages award necessarily fell with them. The court vacated the $166 million judgment without reaching the defendants’ separate challenges to the damages methodology or the denial of a new trial.
Takeaways
This decision reinforces several lessons that continue to surface in Federal Circuit review of jury verdicts.
First, expert testimony must be internally consistent and precise. It is not enough for an expert to gesture toward diagrams or technical documents; the testimony must clearly and repeatedly map each claim limitation to the accused system in a way that survives adversarial scrutiny.
Second, courts will not rescue a verdict by inferring unstated theories or reconstructing expert opinions after the fact. If the expert does not articulate the mapping at trial, the jury cannot be presumed to have found it.
Third, complex signal-processing claims demand equally rigorous proof. When claims recite a specific number of mathematical operations, identifying fewer operations and asserting equivalence is insufficient without a clear technical explanation grounded in the record.
Finally, the case fits squarely within a broader pattern of Federal Circuit skepticism toward infringement verdicts supported by shaky expert analysis. Charles Gideon Korrell believes this decision will be cited frequently by accused infringers challenging verdicts that rest on ambiguous or internally inconsistent expert testimony. Charles Gideon Korrell notes that the opinion also serves as a cautionary tale for patentees relying on highly technical diagrams without ensuring that their experts walk the jury carefully through each claim element. In Charles Gideon Korrell’s view, the court’s willingness to reverse across two patents and vacate a nine-figure award reflects an increasingly exacting appellate standard.
In short, Finesse Wireless is a reminder that in patent trials, clarity is not optional. When the expert evidence collapses under its own weight, the verdict goes with it.