Tesseract Systems v. Nvidia: Neural Network Patent Suit Dismissed With Prejudice
Tesseract Systems LLC asserted US10984320B2 — covering highly trainable neural network configurations — against Nvidia Corp. in the Western District of Texas. The case closed with prejudice just 63 days after filing, permanently barring re-litigation of the same claims.
A 63-Day Neural Network Patent Campaign Against Nvidia Ends Permanently
Tesseract Systems LLC filed suit against Nvidia Corp. on 7 October 2025 in the Western District of Texas before Judge Alan D. Albright, asserting US10984320B2. The patent, filed under application number US15/582831, covers highly trainable neural network configurations — technology directly relevant to Nvidia’s core GPU and AI accelerator product lines.
On 9 December 2025, the court granted Tesseract’s own notice of voluntary dismissal with prejudice. Dismissal with prejudice is a permanent resolution: Tesseract cannot refile the same patent claims against Nvidia in any US court. The order was plaintiff-initiated, suggesting the parties either reached a private arrangement or Tesseract elected to exit before substantive litigation costs escalated.
The 63-day lifespan is notably short even by the standards of early-resolution patent cases, indicating no significant motion practice occurred. What drove the withdrawal — licensing agreement, inability to sustain litigation costs, or a pre-suit resolution — remains undisclosed in the public record. The with-prejudice designation, however, means whatever drove the exit, it is final.
Filing to Dismissed with Prejudice in 63 days
Resolved in 63 days — well under the W.D. Texas median for patent cases
Dismissed with prejudice: what the voluntary exit means for both parties
Voluntary dismissal with prejudice is permanent and self-inflicted
A dismissal with prejudice, even when plaintiff-initiated under Fed. R. Civ. P. 41(a), permanently extinguishes the claims. Tesseract cannot reassert US10984320B2 against Nvidia in any US court. Unlike a dismissal without prejudice — which preserves refiling rights — this closure is absolute. Courts grant such motions routinely when the plaintiff requests them, regardless of merits.
Rule 41(a) — permanent barTesseract exits permanently — no public recovery recorded
Tesseract Systems secured no publicly visible judgment, damages award, or injunction. The with-prejudice designation forecloses any future action on the same claims against Nvidia. This is consistent with either a confidential settlement — where withdrawal with prejudice is the consideration — or a unilateral decision to abandon. The public record does not confirm which scenario applies.
No public recoveryNvidia walks away free of this patent claim permanently
Nvidia faces no liability, no injunction, and no ongoing exposure from Tesseract’s US10984320B2 claims. The with-prejudice dismissal provides Nvidia with a permanent shield against re-litigation of these specific neural network configuration claims by Tesseract. No fee-shifting order was recorded, meaning each party likely bears its own costs — a common outcome in early voluntary dismissals.
Full immunity from these claimsNeural network IP remains a live risk — this exit resolves only one vector
The rapid closure suggests Nvidia’s AI and GPU product lines faced minimal disruption. However, US10984320B2 remains a granted patent that could be asserted against other defendants. Companies developing highly trainable neural network architectures — particularly those competing with or supplying to Nvidia — should note the patent’s existence and scope regardless of this case’s outcome.
Patent remains enforceable vs. othersFull party and counsel information
| Role | Name | Type | Detail |
|---|---|---|---|
| Plaintiff | Tesseract Systems LLC | Company | Patent assertion entity — holder of US10984320B2 (neural network configurations)Search in Eureka ↗ |
| Defendant | Nvidia, Corp. | Company | Nvidia Corp. — global GPU and AI accelerator manufacturerSearch in Eureka ↗ |
| Plaintiff counsel | Isaac Rabicoff | Attorney | Counsel for Tesseract Systems LLCSearch in Eureka ↗ |
| Plaintiff law firm | Rabicoff Law LLC | Law Firm | Representing Tesseract Systems LLCSearch in Eureka ↗ |
| Defendant counsel | Christopher S. Ponder | Attorney | Counsel for Nvidia, Corp.Search in Eureka ↗ |
| Defendant counsel | Harper S. Batts | Attorney | Counsel for Nvidia, Corp.Search in Eureka ↗ |
| Defendant counsel | Jeffrey Liang | Attorney | Counsel for Nvidia, Corp.Search in Eureka ↗ |
| Defendant law firm | Sheppard Mullin Richter & Hampton LLP | Law Firm | Representing Nvidia, Corp.Search in Eureka ↗ |
| Presiding judge | Judge Alan D Albright | Judge | Texas Western District CourtSearch in Eureka ↗ |
Official order — verbatim text
The order adopts plaintiff’s own framing — ‘notice of voluntary dismissal with prejudice’ — confirming this was Tesseract’s election, not a court-imposed outcome. The with-prejudice designation is the legally operative phrase: it activates res judicata as to these claims against Nvidia. Because no merits ruling was issued, US10984320B2’s validity and scope were never adjudicated, leaving the patent fully enforceable against third parties.
US10984320B2 — Highly Trainable Neural Network Configurations
US10984320B2, filed under application number US15/582831, protects configurations for highly trainable neural networks — architectures designed to optimise learning efficiency and adaptability across training workloads. The patent sits squarely in the deep learning infrastructure domain, covering structural and methodological aspects of neural network design that are foundational to modern AI model training pipelines.
For the AI hardware sector, this patent’s claim scope is commercially significant. Nvidia’s GPU platforms underpin the majority of large-scale neural network training globally. Any patent asserting rights over neural network configuration methods creates potential exposure across Nvidia’s data centre, automotive, and edge AI product lines — and by extension, across the broader ecosystem of software frameworks and cloud providers building on Nvidia silicon.
Should your team run an FTO against US10984320B2?
Any organisation commercialising neural network training systems, AI accelerator hardware, or deep learning software frameworks should treat US10984320B2 as a live FTO consideration. Tesseract’s withdrawal with prejudice against Nvidia does not limit its right to assert the same patent against other parties. Cloud providers, AI chip startups, and enterprise ML platform vendors are all potentially within scope.
PatSnap Eureka’s FTO Search Agent can map your product’s neural network architecture claims against the full scope of US10984320B2, identify prosecution history disclaimers that may narrow claim breadth, and surface related continuation applications filed under US15/582831. This analysis is essential before launching or scaling any highly trainable neural network product in the US market.
Run a freedom-to-operate analysis on US10984320B2 to assess your product’s exposure
Run FTO in Eureka →Similar Neural Network Patent Cases in W.D. Texas
Comparable patent assertion cases targeting AI hardware and neural network technology in the Western District of Texas, including related PAE enforcement actions.
What this case signals for the AI hardware and neural network IP landscape
Short-lived assertions against AI chipmakers are increasing — understanding the pattern matters for portfolio strategy.
W.D. Texas remains a magnet for PAE actions against AI hardware firms
Judge Albright’s docket continues to attract patent assertion entities targeting large AI infrastructure companies. Even cases that terminate in under 90 days generate docket exposure and legal costs for defendants like Nvidia. In-house teams should maintain active monitoring of neural network patent filings in W.D. Texas.
With-prejudice dismissals in 63 days often signal pre-suit or early confidential resolution
When a plaintiff voluntarily exits with prejudice this quickly — before any substantive motion practice — it typically signals a confidential settlement or licensing arrangement. R&D and IP teams should treat rapid PAE withdrawals against Nvidia as a potential indicator of royalty-bearing outcomes, even absent public disclosure.
Tesseract v Nvidia — key questions answered
The case was dismissed with prejudice on 9 December 2025, 63 days after filing. Tesseract Systems voluntarily requested the dismissal, and Judge Alan D. Albright granted it. The with-prejudice designation permanently bars Tesseract from reasserting US10984320B2 claims against Nvidia in any US court.
Tesseract asserted US10984320B2, filed under application number US15/582831. The patent covers highly trainable neural network configurations — structural and methodological innovations in neural network architecture design relevant to AI model training systems and hardware acceleration platforms.
A voluntary dismissal with prejudice, typically filed under Fed. R. Civ. P. 41(a), is a plaintiff-initiated permanent exit from litigation. It extinguishes the asserted claims and activates res judicata — the defendant cannot be sued again on the same patent claims by the same plaintiff. It differs from a dismissal without prejudice, which preserves refiling rights.
Yes. The dismissal with prejudice applies only to claims against Nvidia Corp. US10984320B2 remains a granted, enforceable patent. Tesseract retains full rights to assert it against any other party — including AI hardware firms, cloud providers, and deep learning software companies — in future litigation.
The 63-day duration suggests no substantive motion practice occurred before the voluntary dismissal. This timeline is consistent with either a confidential pre-trial settlement, a licensing arrangement, or a plaintiff decision to exit before incurring significant litigation costs. The public record does not disclose the underlying reason, and no damages or fee orders were entered.
PatSnap Eureka searches patents and litigation data to answer instantly.