The Precedent: Federal Circuit Allows Patent Damages Theory Based on Virtual Machine Usage in Exafer Ltd. v. Microsoft Corp.
Authored by Jason Mueller and Bradley Hough
In this edition of The Precedent, we outline the decision in Exafer Ltd. v. Microsoft Corp.
Overview
In Exafer Ltd. v. Microsoft Corp., the Federal Circuit reversed the district court’s decision to exclude Exafer’s damages expert’s report and vacated summary judgment for Microsoft based on the absence of a remedy. The Federal Circuit held that Exafer’s expert report properly used Azure virtual machine (“VM”) usage as the royalty base where the record showed that the accused networking features enabled Microsoft to sell more VM‑hours.
Issues
- Did the district court err by excluding Exafer’s damages expert’s report under Rule 702 because he used VM usage as the royalty base even though VMs themselves were not accused?
- Was summary judgment proper once the damages report was excluded on the theory that Exafer had no viable remedy?
Holdings
- The district court abused its discretion in excluding Exafer’s damages expert’s report. His methodology was based on a “causal connection” between the economic value of the accused features and VMs and did not improperly seek royalties on unrelated activity.
- Because the exclusion was improper and summary judgment was based on the absence of a remedy, the Federal Circuit vacated the summary judgment ruling and remanded the case.
Background and Reasoning
Exafer owns two patents on optimizing communication paths between virtual network devices in cloud environments. It accused Microsoft’s Azure platform of infringement, focusing on SmartNICs and Azure’s VFP Fastpath. Exafer’s theory was that those features offload network processing, reduce CPU usage and allow Azure to host more VMs on the same physical servers.
Exafer’s technical expert, Dr. Congdon, relied on Microsoft’s internal documents to support his report. One presentation credited VFP Fastpath with a 50% improvement in CPU usage. Another internal estimate attributed roughly 300% CPU improvement to SmartNICs. Dr. Congdon opined that a portion of those gains was due to the patented technology and that the gains translated into increased VM capacity.
Building on that, Exafer’s damages expert, Justin Blok, used VM‑hours as the starting point for a reasonable royalty. Azure charges customers per-VM‑hour. Blok:
- Took Microsoft’s data on total VM‑hours sold;
- Applied Dr. Congdon’s CPU‑improvement percentages to estimate the additional VM‑hours Microsoft could sell because of the accused features; and
- Multiplied those incremental VM‑hours by Azure’s average price and gross margin.
He calculated an incremental benefit of about $534.3 million or $0.0625 per-VM‑hour, attributable to the patented improvements, and used that as the basis for a reasonable royalty in the hypothetical negotiation.
Microsoft moved to exclude Blok’s report under Rule 702, arguing that his model violated Federal Circuit precedent, particularly Enplas Display Device Corp. v. Seoul Semiconductor Co., by effectively taxing unaccused VM usage. The district court agreed and excluded his report, then granted summary judgment for Microsoft as Exafer had an “absence of a remedy.”
On appeal, the Federal Circuit distinguished Enplas. In that case, the patentee had tried to recover royalties on hypothetical, unaccused products with “causal connection” to the patents. Here, by contrast, Exafer’s model rested on a specific causal connection supported by Microsoft’s own documents and other technical testimony.
The patents covered network‑layer improvements. The accused SmartNIC and VFP Fastpath features implemented those improvements, which reduced CPU consumption. Reduced CPU consumption freed cores to host more VMs. Azure’s business model monetizes those VMs through VM‑hours. Under those facts, the Federal Circuit found it methodologically sound to use VM‑hours as the economic basis for measuring the value of the patented technology.
The Federal Circuit emphasized that “[t]he facts associated with the alleged infringing activity must be assessed on a case-by-case basis to determine how the parties would value the accused technology during the hypothetical negotiation. “For example, [for] damages theories for claims directed to methods of manufacture,” that value may be measured using revenue from an unaccused product that the method helps produce. The same logic applied here. Blok was not simply grabbing all VM revenue. He isolated the incremental VM‑hour revenue made possible by the accused features and then assigned a per‑VM‑hour royalty consistent with that incremental value.
Because the methodology of the expert damages report was grounded in record evidence and reflected a standard economic approach to valuing incremental benefits, the Federal Circuit held that excluding the report as “unreliable” was an abuse of discretion under Rule 702. Once that exclusion fell, the summary judgment based solely on the “absence of a remedy” could not stand. The Federal Circuit vacated the judgment and remanded for further proceedings.
Takeaway
This decision distinguishes Enplas and shows that a damages model may properly use a royalty base associated with unaccused items where there is a “causal connection” between the accused features and the unaccused items. Technical and business evidence, tying the patented features to the chosen base, can be used to prove the causal connection.