The parties and the court
The case is VB Assets, LLC v. Roku, Inc., No. 1:26-cv-01336 (D. Del.), filed October 16, 2026. The complaint describes VB Assets, a Delaware limited liability company based in Bellevue, Washington, as the successor to VoiceBox Technologies Corporation and VoiceBox Technologies, Inc., which it says were founded in 2001 to bring natural language understanding to a wide range of computer applications. Roku, Inc. is a Delaware corporation headquartered in San Jose, California. Venue is pleaded under 28 U.S.C. 1400(b) on Roku's Delaware incorporation. VB Assets is represented by Farnan LLP and Greenbaum Law LLC and demands a jury.
What the patents cover
The complaint lists thirteen VoiceBox patents in three families and asserts seven of them, quoting claim 1 of each. U.S. Patents 8,073,681 (issued December 6, 2011), 10,510,341 (December 17, 2019), and 10,755,699 (August 25, 2020) share the title System and Method for a Cooperative Conversational Voice User Interface. Claim 1 of the '681 patent covers a method that receives an utterance containing words with different meanings in different contexts, accumulates short-term shared knowledge about the current conversation and long-term shared knowledge about past conversations with the user, has a conversational speech engine identify a context from both and establish the intended meaning to disambiguate the user's intent, and generates a response that the engine grammatically or syntactically adapts to that meaning. Claim 1 of the '341 patent covers accumulating short-term knowledge from utterances received during a predetermined time period, expiring older items of short-term knowledge, keeping long-term knowledge that includes at least one expired item, and using both to determine a context and an interpretation for a new utterance and generate a response. Claim 1 of the '699 patent covers accumulating short-term knowledge from utterances in a single conversation and long-term knowledge from earlier utterances, identifying the manner in which an utterance was spoken based on both, and generating a response based on the interpretation and that manner of speaking.
U.S. Patent 7,818,176 (October 19, 2010) is titled System and Method for Selecting and Presenting Advertisements Based on Natural Language Processing of Voice-Based Input. Its claim 1 covers recognizing words in an utterance at a speech recognition engine by mapping a stream of phonemes to syllables represented in an acoustic grammar and generating a preliminary interpretation, interpreting the words at a conversational language processor that establishes a context for the utterance, and selecting and presenting an advertisement in that context. U.S. Patents 9,269,097 (February 23, 2016) and 11,080,758 (August 3, 2021) share a title on delivering targeted advertisements and natural language processing based on advertisements. Claim 1 of the '097 patent covers presenting an advertisement for a product or service, receiving the user's natural language utterance, and interpreting it based on the advertisement, including deciding, when the utterance contains a pronoun, whether the pronoun refers to the product or service or to its provider. Claim 1 of the '758 patent covers determining a context for an utterance from words output by a speech recognition engine, selecting and delivering a purchase opportunity based on that context, tracking an interaction pattern that includes the device completing a transaction, building or updating a user-specific profile from that pattern, and using the profile to interpret a later utterance and select a later purchase opportunity.
U.S. Patent 11,087,385 (August 10, 2021), titled Voice Commerce, has a claim 1 that covers receiving a single natural language utterance, searching databases of products or services based on the recognized words, selecting a product or service to buy without further user input, receiving a second input confirming the purchase, and completing the transaction without further input.
Accused products and theories
The complaint accuses the voice-based functions of the Roku Voice Remote, Roku Voice Remote Pro, and Roku Mobile App, along with other hardware and software that let users interact with the Roku voice features shown in its claim charts. It calls the accused voice, natural language understanding, and AI functions the "AI Functionality" and alleges, on information and belief, that the speech recognition engines, language processors, and inference code behind them exist as discrete modules that Roku develops and updates separately from unaccused features. Each count alleges that Roku itself performs every step of the claimed methods and that, to the extent end users perform any step, Roku directs or controls that performance through its software design, terms of service, and instructions. The complaint also pleads induced and contributory infringement. For each patent it points to an exemplary claim chart, Exhibits H through N, and says VoiceBox expects to identify additional asserted claims under the case schedule.
Notice and willfulness allegations
Willfulness is pleaded on all seven patents and rests on a business relationship the complaint says ran from 2014 to 2017. It alleges that Roku sent VoiceBox a mutual nondisclosure agreement on June 13, 2014; that VoiceBox's written pricing proposal of October 10, 2014 described its offering as its "own speech technology (ASR) combined with our patented natural language understanding (NLU) technology" and that an updated proposal of December 17, 2014 repeated that representation; that Roku engineers held a technical deep dive on VoiceBox's architecture on January 5, 2015; and that after talks resumed in late 2016, Roku issued a request for information in May 2017, told VoiceBox "Voicebox is on the shortlist," signed a Voicebox Cloud Platform Evaluation Agreement in June 2017, held an executive meeting on June 27, 2017, and asked for final pricing on September 4, 2017. The complaint alleges that Roku did not take a license and instead developed the accused products. It also alleges, on information and belief, that Roku learned more about the patents when two former VoiceBox computational linguists joined Roku in 2018 and 2021. For the five patents that issued after the 2014 proposals, it alleges, on information and belief, that Roku knew of or was willfully blind to each patent no later than its issue date, based on the earlier dealings.
The record against Amazon
Three of the seven patents have been tried. In VB Assets, LLC v. Amazon.com Services LLC, No. 1:19-cv-01410-MN (D. Del.), a jury found on November 8, 2023 that Amazon willfully infringed claim 13 of the '681 patent, claim 40 of the '176 patent, claim 23 of the '097 patent, and claim 25 of U.S. Patent 9,626,703 through Alexa, found none of those claims invalid, and awarded $46.7 million in reasonable royalty damages. On September 30, 2024, Judge Maryellen Noreika granted Amazon judgment as a matter of law of no infringement of claim 25 of the '703 patent, which reduced the award to $40,007,600, denied Amazon's motion on the other claims, and declined to enhance damages. A December 12, 2024 opinion set ongoing royalties of $0.45 per net new U.S. Alexa user for the '681 patent and $0.25 per net new U.S. Alexa shopping user for the '176 and '097 patents. The May 16, 2025 final judgment holds the three claims willfully infringed and not invalid for lack of written description, for obviousness, or under 35 U.S.C. 101, and adds $7,137,894 in prejudgment interest. The Federal Circuit dismissed the cross-appeals, Nos. 2025-1854 and 2025-1855, on the parties' agreement on April 29, 2026, with each side bearing its own costs.
The September 2024 opinion also notes that the Patent Trial and Appeal Board held the claims of another VoiceBox patent, U.S. Patent 9,015,049, unpatentable in IPR2020-01346, and that U.S. Patent 8,886,536 was dropped before trial. Neither of those patents, nor the '703 patent, is asserted against Roku. The '341, '699, '758, and '385 patents were not part of the Amazon trial.
Relief requested
VoiceBox asks for a judgment of direct and indirect infringement, damages under 35 U.S.C. 284, a willfulness finding with damages increased up to three times, an exceptional-case fee award under 35 U.S.C. 285, and pre-judgment and post-judgment interest. The prayer does not ask for an injunction.
What it means
The suit carries the VoiceBox portfolio from smart speakers to streaming-device remotes and a phone app. The Amazon case ended in a running royalty measured per Alexa user; the Roku complaint instead frames the accused voice functions as separable software modules, an allegation that bears on how any damages base is apportioned. The willfulness theory rests on years of pre-suit commercial dealings. The notices the complaint quotes refer to VoiceBox's "patented natural language understanding (NLU) technology," and for five of the seven patents they predate issuance.
What to watch next
Watch for Roku's answer or a Rule 12 motion, including any challenge to eligibility under 35 U.S.C. 101 or to the willfulness allegations, and for any inter partes review petitions on the asserted patents. Watch also how the parties address the 2025 Amazon judgment on the '681, '176, and '097 patents. The case can end at any time, including by settlement or voluntary dismissal.