Google has settled claims from a startup accusing it of stealing technology to boost its AI products. (AP Photo/Peter Morgan)
The tentative resolution came half an hour before the counsel for the parties were scheduled to appear before Judge Saylor and the jury for closing arguments.
"We are pleased to have resolved this matter," Google spokesman José Castañeda said in a statement Wednesday. "As we showed in court, Singular's patent doesn't apply to our tensor processing units, which were independently designed and built by Google engineers using Google technology over many years."
Representatives for Singular were not immediately available for comment. No terms of the deal were detailed in the filing.
Singular founder Joseph Bates claimed he had repeatedly pitched Google his computing architecture ideas as a way to increase the capacity of "artificial intelligence" computers by having the chips perform less precise calculations. Google then adopted the technology, Singular claimed, and used it in a quarter-million computer chips.
Google denied lifting the idea from Bates, argued his patents should never have been granted, and said the technology underlying the computer chips at issue was fundamentally different from Singular's technology and patents.
In the end, Singular's technology wasn't right for Google, the company's attorney Robert Van Nest of Keker Van Nest & Peters LLP said at opening arguments.
"His patented claim is a device using approximate math, which means getting incorrect answers on a regular basis," Van Nest said, also calling Bates a "disappointed investor" who unsuccessfully pitched his ideas to dozens of other tech firms. "That approach was simply not right for Google or anyone else."
Singular attorney Kerry Timbers of Sunstein LLP told the jury that Bates' presentations had been widely distributed around Google's computer engineers, so much that one employee had claimed the company was "quite corrupted with Joe's ideas because we already heard them."
The jury was promised at openings to see how Google saved about $1.6 billion by using Bates' technology, a sum reached by assuming the tech giant would have had to buy 1 million off-the-shelf computer chips to achieve the same output. If the jury agreed, the damages sum would have been one of the largest patent verdicts ever.
In its response, Google claimed it spent $143 million over five years with 25 engineers to develop the chips that power its machine-learning applications.
After Singular filed the infringement action, Google challenged the validity of the patents through the inter parties review process at the Patent Trial and Appeal Board in November 2020, claiming Bates' purported invention was obvious.
The PTAB in May 2022 found only some of the claims Bates made in the patent to be obvious and unpatentable, and Google has appealed that decision to the Federal Circuit.
The patents-in-suit are U.S. Patent Nos. 8,407,273 and 9,218,156.
Singular is represented by Matthew D. Vella, Kevin Gannon, Daniel McGonagle, Brian Seeve, Robert R. Gilman, Aaron S. Jacobs, Jeffrey Jackson Pyle and Adam R. Doherty of Prince Lobel Tye LLP and Kerry L. Timbers of Sunstein LLP.
Google is represented by Gregory F. Corbett, Nathan R. Speed, Elizabeth A. DiMarco and Anant K. Saraswat of Wolf Greenfield & Sacks PC, Robert Van Nest, Michelle Ybarra, Andrew Bruns, Vishesh Narayen, Christopher S. Sun, Anna Porto, Deeva Shah and Stephanie J. Goldberg of Keker Van Nest & Peters LLP, Michael S. Kwun and Asim Bhansali of Kwun Bhansali Lazarus LLP and Matthias A. Kamber of Paul Hastings LLP.
The case is Singular Computing LLC v. Google LLC, case number 1:19-cv-12551, in the U.S. District Court for the District of Massachusetts.
--Editing by Brian Baresch.
Correction: A previous version of this article listed an attorney for Singular who has withdrawn from the case. The error has been corrected.
For a reprint of this article, please contact reprints@law360.com.
