InterDigital v Amazon, Part 2: What Changes When You Fix the Inputs
A critic identified a real flaw in my Part 1 analysis. Here is what happened when I corrected it — and what the revised system found that nobody else is talking about.
The Criticism That Earned a Response
After Part 1 published, a well-sourced critic posted publicly that my analysis had committed an “idiocy that can be demonstrated very easily.” He attached a court document showing the Delaware case I had flagged as a potential pressure lever was stayed — and had been stayed for years — as a companion to an ITC investigation. A stayed companion case carries no independent coercive weight. He was right.
I conceded the point publicly and immediately. But I also drew a distinction that matters: the system reasoned over the JUVE Patent chronology as its seed document. That article did not reference the Texas or Virginia cases, or the ITC investigation. That is a source document limitation, not an AI hallucination. The system can only reason over what it is given.
The correct response to a source document limitation is better inputs. So I rebuilt the input layer.
This is Part 2.
What Changed in the Methodology
Part 1 used a single source document — the JUVE Patent chronology — as the seed for the entire six-agent analysis. That article covered the European and UK proceedings in detail but mentioned only one US case, and mentioned it in passing.
For Part 2, I added a pre-processing step: a Python script that calls the CourtListener/PACER API to pull live docket data for all US proceedings, combines that with the original JUVE article, and writes an enriched input file that the swarm then analyzes. The six-agent architecture, the seven roles, the five rounds, the coverage audit, and the final refinement all ran unchanged. Only the inputs improved.
What the CourtListener pull revealed was that the US proceedings picture is not one case. It is four — and they are materially different from each other in status, patent family, and strategic significance.
The US Proceedings Map: What Part 1 Missed
Here is what the corrected fact ledger shows:
District of Delaware — 1:25-cv-01365 — STAYED Filed November 7, 2025. Five video codec patents. This is the case the critic correctly identified — a companion complaint to the ITC investigation, stayed pending ITC resolution. Carries no independent coercive weight. Should not have appeared in the pressure analysis at all.
ITC Investigation — 337-TA-3869 — ACTIVE Filed December 2025. The same five video codec patents as Delaware. This is the primary US enforcement lever — not Delaware. An ITC exclusion order blocks physical import of FireTV, Kindle, and Echo Show devices into the United States. That is commercial disruption at the supply chain level, not a financial penalty. However — and this is the correction that materially changes the timing prediction — a December 2025 ITC filing puts the ALJ initial determination on a standard 15–18 month schedule, meaning approximately March–June 2027. Not Q4 2026.
Eastern District of Virginia — 2:25-cv-00822 — ACTIVE Filed December 18, 2025, originally in the Alexandria Division, transferred to Norfolk. Four different video coding patents: US 8,149,338; 11,252,435; 12,143,606; 12,149,734. This is a separate patent family running on an independent timeline, assigned to District Judge Arenda L. Wright Allen, represented by McKool Smith PC — a specialist patent litigation firm known for aggressive trial posture. This case was entirely absent from Part 1.
Western District of Texas, Midland/Odessa Division — ACTIVE Filed February 2026 by InterDigital and subsidiary DRNC Holdings. Six content delivery patents: US 7,921,259; 10,116,565; 8,745,128; 9,015,416; 8,868,701; 8,583,769. This is not a video codec case. It targets internet traffic management and routing — Amazon’s cloud infrastructure and streaming delivery platform. It was entirely absent from Part 1, and it is the most strategically significant US filing in the entire dispute.
The Edgio Angle: Premeditated Arsenal Assembly
The Texas case contains a detail that the system correctly flagged in the Part 2 fact ledger and that deserves more attention than it has received anywhere else.
InterDigital acquired the content delivery patents through its subsidiary DRNC Holdings — purchased from Edgio Inc., a content delivery network company, in January 2025. Edgio went through bankruptcy. InterDigital, whose core business is video codec and wireless SEP licensing, has no organic presence in internet traffic management and routing technology. These are not patents InterDigital developed. They are patents InterDigital bought.
The acquisition date is January 2025. Amazon filed in the UK in August 2025.
InterDigital was assembling its litigation arsenal — including a completely different patent family specifically targeting Amazon’s cloud infrastructure — seven months before Amazon pulled the trigger on the UK proceedings. That is not reactive enforcement. That is premeditated multi-front preparation.
This reframes the escalation narrative in an important way. The conventional reading is that Amazon initiated by filing in the UK. But InterDigital had already positioned itself to attack Amazon’s most profitable business segment — AWS and Prime Video delivery — before the opening shot was fired. The Texas filing in February 2026 was not an escalation. It was the deployment of a weapon that had been loaded seven months earlier.
From a pro-licensor perspective, this is sophisticated portfolio strategy executed at the highest level. InterDigital identified Amazon’s commercial vulnerabilities across multiple technology layers, acquired the necessary patent coverage where its organic portfolio had gaps, and built a litigation architecture capable of threatening not just Amazon’s device business but its cloud infrastructure. That is what serious IP enforcement looks like.
What the Revised System Found
With the corrected, enriched input layer, the seven-agent swarm produced a materially different analysis in three specific areas.
On the ITC timeline: The system correctly identified that the December 2025 ITC filing produces an ALJ initial determination in approximately March–June 2027 — well past Q4 2026. The Part 1 prediction of a Q4 2026 settlement driven by ITC pressure was therefore built on a timing error. The ITC is a genuine lever, but it operates on a 2027 timeline, not a 2026 one.
On the Texas cloud infrastructure case: The system elevated this from a footnote to a first-order variable. A broad injunction or adverse ruling in Texas would not just affect Amazon’s device sales — it would threaten Amazon’s ability to serve third-party cloud and streaming customers through AWS. That is a completely different category of commercial disruption. The system’s US/ITC Agent, whose mandate is to evaluate whether US actions create meaningful leverage tied to physical products and supply disruption, correctly identified that the Texas case extends the threat surface from consumer hardware to Amazon’s most profitable business segment.
On the Mannheim breach finding: The system correctly weighted the 27 February 2026 UPC finding that Amazon was actually in breach of the ASI — not merely at risk of breach. Contempt sanctions and the €50M penalty are not hypothetical future risks. They are procedurally ripe now. This materially strengthens InterDigital’s near-term European leverage independent of the May 2026 appeal outcome.
The Revised House View
Prediction: Amazon and InterDigital reach a negotiated global patent license in late 2026 or early Q1 2027, before any binding judicial FRAND determination issues from the UK High Court. The settlement is not driven by any single forum but by the convergence of four simultaneous pressure vectors: UPC contempt sanctions that are procedurally ripe and accruing; the Texas cloud infrastructure case threatening AWS and Prime Video delivery at scale; the UK trial producing a rate anchor both parties can negotiate around; and InterDigital’s existential need to avoid a below-market public RAND determination that would cascade as a de facto ceiling across its entire licensing portfolio. The agreed rate lands above Amazon’s pre-litigation position but below InterDigital’s initial demand, structured as a multi-year global portfolio license with a retrospective lump sum. All active proceedings are withdrawn by consent upon execution.
Alternative Outcome: The UPC Court of Appeal narrows the AILI in May 2026, restoring Amazon’s fragmentation strategy. The ITC timeline slips past Q4 2026 without producing a near-term exclusion order threat. The Texas case is stayed or proceeds slowly. Amazon accepts the UK judicially determined rate under protest, challenges its extraterritorial scope in US courts, and the parties execute a license on judicial terms after a 6–12 month post-judgment enforcement phase. InterDigital suffers a below-demand rate that depresses its broader licensing program. Probability: approximately 30–35%.
Key Driver: The May 2026 UPC Court of Appeal ruling on AILI scope — specifically whether it upholds the AILI broadly and affirms that UK RAND damages proceedings improperly interfere with UPC jurisdiction. If it does, Amazon loses its primary delay forum and faces simultaneous coercive pressure from ripe UPC contempt sanctions, the Texas cloud infrastructure case, and the approaching UK trial rate anchor. Settlement becomes near-certain. If the Court of Appeal narrows the AILI, Amazon’s fragmentation strategy survives and the UK trial proceeds to a public judgment.
Secondary Driver: The trajectory of the Western District of Texas case. If it produces injunctive relief or a damages award on a timeline that precedes the UK trial, Amazon’s exposure extends from consumer hardware to its most profitable business segments and its negotiating posture collapses regardless of the ITC or UPC outcomes.
Main Uncertainty: Whether the Texas content delivery case is stayed, proceeds slowly, or accelerates. This is the variable that most determines whether the settlement pressure is overwhelming or merely significant. It is also the variable that received the least public attention before this analysis.
Confidence: 54% on late 2026 / early Q1 2027 global license. 35% on alternative fragmentation path. 11% on outcomes outside these two scenarios.
What Changed From Part 1 and What Didn’t
The directional prediction survived: a confidential global license remains the most probable outcome. The mechanism and timing got sharper.
What changed: the ITC is no longer the primary forcing mechanism on a 2026 timeline. The Texas cloud infrastructure case is now a first-order variable. The settlement window moved from Q4 2026 to late 2026 / early Q1 2027. The confidence number moved from 52% to 54% — a modest improvement reflecting a more complete fact base rather than a fundamentally different analysis.
What didn’t change: InterDigital’s benchmarking risk remains the most important structural incentive driving both parties toward a confidential rather than public resolution. The May 2026 UPC appeal remains the central hinge point. The February 2026 near-settlement remains the most important data point — parties who were in touching distance once, blocked only by a procedural technicality, will find their way back.
The most important correction is the one the critic forced: the Delaware case is irrelevant. The ITC is real but slow. The Virginia case is active and underreported. The Texas case is the sleeper that nobody is watching.
A Note on the Methodology
Part 2 adds one new capability to the system: live docket data retrieval via the CourtListener/PACER API. Before the swarm runs, a pre-processing script pulls current docket metadata for all identified US proceedings, combines it with the source article, and writes an enriched input file. The swarm then analyzes the enriched input unchanged.
The CourtListener API returned docket metadata for the Delaware and Virginia cases. Docket entries at the filing level require an elevated access tier. The ITC and Texas dockets were included via verified public sources rather than direct API retrieval.
The total cost of the Part 2 run — 37 API calls across seven agents, five rounds, coverage audit, and final refinement — was approximately $2.50 at current Claude Sonnet 4.6 pricing. Runtime was under five minutes.
The full Python implementation — including the CourtListener pre-processor, the swarm architecture, and the enriched input builder — is available at github.com/theharlans/standards-at-risk.
Sources: JUVE Patent, “InterDigital vs Amazon — A chronology of the escalation,” 16 March 2026. CourtListener/PACER docket data retrieved April 2026. Bloomberg Law, Mondaq, and verified public sources for Texas and ITC proceedings. All analysis is the author’s own. This is not legal advice.