In-depth reporting and analytical commentary on intellectual property disputes and debates. No legal advice.

Xockets now trying to assert patents against Microsoft’s Fungible and Azure Boost DPUs after first focusing on Microsoft’s use of Nvidia chips

Context: Xockets is a company that claims on its patent assertion-centric corporate website to have “invented the DPU” (data processing unit). Three months ago, Xockets announced an antitrust and patent infringement lawsuit against Nvidia (the only defendant mentioned in the headline of the press release), Microsoft and patent aggregator RPX Corporation in the United States District Court for the Western District of Texas (September 5, 2024 press release by Xockets (PDF)). Its first attempt to game the system failed: by filing an antitrust-only complaint followed by an amended complaint with patent infringement claims only a few hours later, Xockets sought to judge-shop by circumventing the Western District’s assignment rules for patent cases, but the case was reassigned from Judge Alan D. Albright to Judge Leon Schydlower.

What’s new: On Wednesday (December 11, 2024), Microsoft filed, with Nvidia’s support, a so-called CSE (customer-suit exception) motion, asking the court to sever and stay the patent infringement claims against it as it makes more sense to resolve the claims against the chipset maker first (and Microsoft agrees to be bound by the outcome of those proceedings). Xockets opposes the motion. Apparently recognizing that the three CSE factors as applied in that district (if not even the traditional stay factors) would be met, Xockets is trying to keep the patent claims against Microsoft in the case by belatedly bringing infringement allegations against Microsoft’s use of non-Nvidia chips: the acquired Fungible DPU (January 9, 2023 corporate blog post by Microsoft) and the in-house Azure Boost DPU (November 19, 2024 corporate blog post by Microsoft). Microsoft asks the court to grant the motion to sever and stay at any rate.

Direct impact: Notwithstanding their patentee-friendly reputation, the Western and the Eastern Districts of Texas do fairly often apply the CSE. Here, Microsoft advocates for applying the CSE regardless of Xockets’s new attack (possibly just in anticipation of the CSE motion) on its “homemade” DPUs. Xockets already amended the complaint once (within hours of the original filing), but nowhere on its 203 pages does it specifically discuss any other accused technologies than Nvidia chips. Another argument is that allegations involving non-Nvidia silicon would not be eligible for joinder.

Wider ramifications: Each plaintiff is the master of its case, but Xockets exposes itself to accusations of gamesmanship. By contrast, another AI/cloud-related patent infringement litigation against Microsoft was brought by supercomputer firm ParTec more than six months ago (June 10, 2024 ip fray article) and is running the ordinary course. Last month, Judge Rodney Gilstrap recused himself (November 13, 2024 ip fray article) and Judge Robert W. Schroeder III took over. A Markman (claim construction) hearing will be held on October 28, 2025, and a jury trial is scheduled to begin on April 20, 2026. ParTec is also suing Nvidia, but in a different jurisdiction: Europe’s Unified Patent Court (UPC) (November 7, 2024 ip fray article). And while ParTec is not formally pursuing antitrust claims at this stage, it accuses Nvidia of a refusal to deal with ParTec in response to its patent assertions against Nvidia customer Microsoft.

Xockets’s antitrust-patent gamble

Xockets is not the first U.S. patentee to have brought a combination of patent infringement and antitrust allegations against the same party (or, as here, two parties: Nvidia and Microsoft; RPX will be addressed further below) in a single complaint.

In theory, one could picture a case in which the antitrust claims are comparably strong as the patent claims, such as a scenario in which a platform gatekeeper has a successful app in its app store, kicks it out arbitrarily and then offers a similar product that infringes the original app maker’s patents. But ip fray has yet to see such a case. In practice, one set of claims will eclipse the other, especially if a patentee believes it can increase its chances of prevailing on its patent infringement assertions by throwing in some antitrust claims. That motivation could be psychological (trying to portray the defendant as particularly lawless, especially with a view to a jury trial), or it could be procedural.

While that fact alone does not prove Xockets’s antitrust claims meritless, it is a fact that Xockets hoped to gain something from antitrust claims regardless of their merits or lack thereof: it wanted to put the case before Judge Albright. That is understandable, but filing a thin antitrust complaint followed by an undoubtedly already prepared 203-page combined antitrust and patent infringement complaint a few hours later was an extremely obvious attempt to circumvent the assignment rules for patent cases in the Western District of Texas. If the case had remained on Judge Albright’s docket, but the antitrust claims had been dismissed (which is what the defendants are indeed seeking), Xockets would still have achieved one tactical objective.

What is clear is that the antitrust claims logically build on the patent infringement allegations. There are two counts, both of which relate to buyer’s cartel allegations (though one of them is brought under Section 1 and the other under Section 2 of the Sherman Act) with respect to the terms on which Xockets’s patents might be licensed. If, however, no valid Xockets patent is infringed by Nvidia and Microsoft, then they don’t need a license anyway.

One of the two antitrust counts involves not only Microsoft and Nvidia, but also the third defendant: RPX. There is also a pending motion for preliminary injunction (PI). The court held a hearing in late October. Decisions on the PI motion and on the motion to dismiss the antitrust claims should come down soon. But the question of whether the patent infringement claims against Microsoft should be severed and stayed is a separate one.

Customer-suit exception argument with respect to Nvidia chips

With similar fact patterns as the one presented by Microsoft in this case, defendants are reasonably likely in the Western District of Texas to prevail on a CSE argument. It does not mean that the claims against them are dismissed. They will be resolved against the maker of the accused products, and if anything is still left to be addressed between the patentee and the defendant that is just a customer using those products, there can be further proceedings. Typically, there will be nothing left (as a result of patent exhaustion, meaning that one can’t collect royalties at both levels of a value chain over the same infringement), or so little that the parties can easily settle.

There are cases in which the defendants fail to prevail on a CSE argument, such as when they accept to be bound only be the infringement but not the validity determination in the case against the product maker. That, for example, is clearly not the case here, as Microsoft is unequivocal about being bound by the outcome of Xockets’s patent dispute with Nvidia.

Microsoft’s motion argues that the patent infringement claims against it could be stayed even under the standard that generally applies to stays.

New infringement allegations against non-Nvidia chips

The complaint (technically, the first amended complaint, but for the patent claims that is the original one) does not specifically allege infringements by Microsoft’s homemade DPUs (which are not Nvidia’s business in any way). If there is anything in the complaint that refers to non-Nvidia chips, it is just the expression “at least” in para. 83 (joinder):

“For example, Microsoft infringes the asserted Xockets Patents at least through its use of the accused NVIDIA systems and components that NVIDIA sells to Microsoft, as alleged below.”

But Microsoft’s CSE motion explains that Xockets is now trying to expand the scope of the case:

“… Xockets served patent infringement contentions on November 1, purporting to add infringement claim charts directed only to Microsoft based on unaccused non-NVIDIA products identified as ‘Microsoft Fungible and/or Azure Boost DPUs.’”

Microsoft’s motion tells the court that those infringement allegations “are outside the scope of the pleadings, violate the patent joinder statute, should be disregarded, and further justify severance and stay.”

There are certain procedural possibilities such as Xockets perhaps seeking leave to amend its complaint (again). For the event that this happens and the court allows such an amendment, Microsoft argues that the CSE motion would nevertheless have to be considered first, under the Federal Circuit’s Nintendo precedent.

With respect to the joinder statute, Microsoft’s motion recalls that a patentee cannot simply sue two defendants in the same action only because they allegedly infringe the same patents. It takes more than that, and Nvidia has nothing to do with other DPU’s Microsoft uses.

Documents

Finally, here are the first amended complaint (again, that was the first one to come with any patent infringement assertions at all), Microsoft’s CSE motion and Nvidia’s support of that motion.