Exafer v. Microsoft: Federal Circuit Rejects Per Se Bar on Unaccused Royalty Bases

The Federal Circuit’s March 6, 2026 decision in Exafer Ltd. v. Microsoft Corp. is an important reminder that Daubert does not create a shortcut to wipe out patent damages theories simply because they use a product-based royalty base. In a precedential opinion, the court reversed a district court’s exclusion of Exafer’s damages expert, holding that a reasonable royalty can, in the right circumstances, be tied to an unaccused product when the accused technology has a causal connection to that product’s value. The panel also vacated the follow-on orders that denied discovery and granted summary judgment for lack of a remedy, sending the case back for further proceedings. (cafc.uscourts.gov)

Overview​

Patent damages fights often turn less on whether a patent was infringed and more on how to measure what that infringement was worth. That is exactly what made Exafer v. Microsoft so consequential: the district court viewed Exafer’s royalty base as defective because it was built around virtual machines that were not themselves accused of infringement, and it excluded the expert under Rule 702. The Federal Circuit rejected that approach as too rigid, emphasizing that the admissibility question depends on the economic relationship between the accused features and the selected base, not on a categorical prohibition against using noninfringing products. (cafc.uscourts.gov)
The facts matter because Microsoft’s Azure platform is not a stand-alone product in the ordinary consumer sense; it is a layered cloud infrastructure offering where efficiency improvements can translate into greater hosting capacity and, in turn, more billable units. Exafer argued that its expert did not try to claim royalties on unrelated activity. Instead, the expert used Microsoft’s VM-hour pricing to quantify the incremental value Microsoft allegedly derived from the accused Azure features. The Federal Circuit accepted that framing, at least for admissibility purposes, because the record showed evidence of a causal chain between the patented functionality and additional VM capacity. (cafc.uscourts.gov)
The opinion lands at a time when patent damages law is already under pressure from both sides. On one hand, patent owners are increasingly trying to prove value in software, cloud, and platform cases where the infringing feature is buried inside a complex stack. On the other hand, defendants continue to argue that damages experts are smuggling in unrelated revenue streams under the guise of an “economic base.” The Exafer decision does not eliminate apportionment requirements, but it does reinforce that apportionment is not the same thing as exclusion. (cafc.uscourts.gov)

Why this case matters now​

The Federal Circuit’s refusal to turn Enplas into a per se rule is the headline. The court said the proper analysis is case-by-case, and it stressed that the hypothetical negotiation asks how the parties would have valued the accused technology at the time infringement began. That matters because many modern patent disputes involve features whose commercial value appears indirectly, through improved throughput, lower CPU usage, higher capacity, or reduced operating costs. (cafc.uscourts.gov)

• The ruling is especially relevant for cloud infrastructure, virtualization, and software platform cases.
• It narrows the risk that a defendant can win Daubert simply by arguing the royalty base includes a nonaccused commercial unit.
• It strengthens the role of internal documents showing how the accused feature affected system performance and revenue.
• It signals that economic reality can matter more than formal product labels when courts evaluate damages methodology.

The procedural domino effect​

Once the district court excluded the damages expert, the case rapidly shifted from a patent merits dispute to a remedies problem. Exafer sought to reopen fact and expert discovery to pursue an alternative damages theory, but that request was denied. Microsoft then obtained summary judgment on the theory that Exafer lacked a remedy after the expert exclusion. The Federal Circuit’s reversal of the exclusion mooted the discovery ruling and undermined the summary judgment, which is why those orders were vacated as well. (cafc.uscourts.gov)
That sequence is a familiar one in patent litigation. A damages ruling can become existential if the plaintiff has no backup theory or the court cuts off a chance to retool the record. The Federal Circuit’s decision therefore has a practical lesson beyond doctrine: when damages experts are excluded, the downstream remedy rulings often become vulnerable if the exclusion itself falls. In that sense, the damages ruling drives the whole litigation posture.

Background​

Exafer sued Microsoft over the Azure platform, focusing on Azure SmartNICs and Virtual Filtering Platform Fastpath technology. The asserted patents were directed to network-traffic optimization and forwarding methods, and Exafer alleged that Microsoft’s accused features improved CPU efficiency inside Azure servers. According to the Federal Circuit, Exafer’s technical expert tied those efficiency gains to Microsoft’s ability to free up CPU cores and host additional virtual machines. (cafc.uscourts.gov)
Microsoft attacked the damages proof through a Daubert motion. Its argument was straightforward: Exafer’s expert allegedly used unaccused virtual machines as the royalty base, so the methodology supposedly swept in activity that did not itself constitute infringement. The district court accepted that framing, reasoning that Enplas foreclosed use of noninfringing activities in the royalty base. That is the step the Federal Circuit found too broad. (cafc.uscourts.gov)
Enplas itself is important context because it has long been cited in damages disputes as a cautionary tale against overbroad royalty bases. But the Federal Circuit reminded readers that Enplas was tied to a different factual setting: there, the unaccused products had no causal connection to the accused products. That distinction is crucial. The Exafer panel did not say unaccused products are always fair game; it said they can be if they are linked to the accused feature in a way that reflects the value the parties would have negotiated over. (cafc.uscourts.gov)

The Enplas problem​

The district court treated Enplas as if it imposed a bright-line prohibition. The Federal Circuit said that was wrong because the earlier case involved royalty-base elements that truly had no causal connection to the infringement. In Exafer, by contrast, the record allegedly showed that the accused Azure features increased VM capacity, and that incremental capacity was exactly what the damages expert tried to monetize. (cafc.uscourts.gov)

• Enplas is a warning against fictional or untethered royalty bases.
• Exafer is about tying the base to a real operational benefit.
• The difference is not semantic; it is the difference between speculation and economic linkage.
• That linkage can make a royalty base involving an unaccused product admissible.

The role of internal Microsoft documents​

One of the more interesting aspects of the opinion is that Microsoft’s own documents helped Exafer. The court said internal presentations and estimates showed the accused features improving CPU usage and enabling more VM hosting. That evidence mattered because it allowed the damages expert to connect the accused technology to a measurable business benefit rather than simply using VM revenue as a proxy for convenience. (cafc.uscourts.gov)
This is a recurring pattern in cloud and software cases. Internal engineering and product documents often reveal how a defendant values efficiency gains, scalability, or capacity improvements long before litigation begins. When those documents speak in terms of additional units sold, improved margins, or lower resource consumption, they can become the backbone of a reasonable royalty model. The paper trail can be more powerful than the marketing pitch.

The Federal Circuit’s Reasoning​

The panel’s core holding was that the district court abused its discretion by reading Rule 702 and Enplas too aggressively. The court emphasized that the damages expert did not simply slap a royalty rate onto an unrelated product stream. Instead, the expert used the VM-hour base to capture the incremental benefit Microsoft allegedly gained from the accused features, which the court viewed as tethered to the patented invention. (cafc.uscourts.gov)
The opinion is careful in tone. It does not say Exafer will ultimately win a large damages award; it says the expert’s theory was sufficiently reliable to get to the factfinder. That distinction matters because Daubert is supposed to screen out junk science, not resolve close economic disputes that depend on competing factual narratives. In the Federal Circuit’s view, the district court crossed from gatekeeping into merits resolution. (cafc.uscourts.gov)
The panel also leaned on familiar Federal Circuit damages principles, including the highly case-specific nature of reasonable royalty analysis. The court pointed to precedent explaining that courts must ask how the parties would value the accused technology in the hypothetical negotiation, and it noted that some method claims commonly use the unaccused product made by the claimed method as the royalty base. That analogy helped the court show why a nonaccused product base is not inherently improper. (cafc.uscourts.gov)

No categorical ban on unaccused products​

This is perhaps the most important doctrinal clarification in the opinion. The Federal Circuit expressly rejected Microsoft’s argument that Enplas created a categorical rule against using sales of unaccused products to avoid expanding the patent monopoly. Instead, the court held that the admissibility inquiry is contextual. If the base reflects the value of the accused feature as actually used in the market, the methodology can survive Rule 702. (cafc.uscourts.gov)
That matters because modern products are increasingly modular. A single software feature can affect storage, processing, latency, routing, and billing all at once. A rigid insistence on limiting damages only to the specific accused line item can undercount the actual value extracted from infringement. At the same time, the court did not relax apportionment; it simply refused to confuse apportionment with exclusion. That is a subtle but vital line.

The causal-connection test in practice​

The panel’s logic effectively asks whether the unaccused product is being used as a stand-in for something that the accused feature makes more valuable. In Exafer, the answer was yes, because the disputed features allegedly allowed Microsoft to host more VMs per hour. The royalty base thus functioned as a measure of incremental capacity, not as a claim to unrelated revenue. (cafc.uscourts.gov)

• A royalty base may be acceptable when it tracks incremental output attributable to the accused feature.
• The court will look for technical proof and business proof working together.
• Internal documents can help connect performance gains to monetizable units.
• A base becomes risky when it is merely a revenue grab disconnected from the infringement story.

Damages Theory and the Hypothetical Negotiation​

The opinion is especially instructive on how damages experts can frame value in software and cloud cases. Exafer’s expert apparently used VM-hour pricing not because VMs were accused products, but because they represented the commercial consequence of the accused features’ efficiency gains. That is a sophisticated but defensible theory: if a patented optimization allows a provider to host more billable units, the royalty may be measured by the economic lift from those additional units. (cafc.uscourts.gov)
That approach aligns with the fundamental purpose of a reasonable royalty, which is to estimate what the parties would have agreed to at the time of hypothetical negotiation. The Federal Circuit cited LaserDynamics and other authorities for the proposition that the value inquiry seeks to capture the patented technology’s worth in the marketplace, not to mechanically isolate a single product label. In other words, the model has to mirror how sophisticated parties would actually bargain. (cafc.uscourts.gov)
This is where the case could influence future expert reports. Patent plaintiffs in platform cases will likely lean harder into incremental-capacity and throughput-based models, especially where internal evidence shows that the accused feature improves utilization. Defendants, in turn, will argue that the base still needs stricter apportionment and that capacity gains must be carefully bounded. The battle will be over how directly the accused feature drives the revenue stream. (cafc.uscourts.gov)

The economics of VM-hours​

A VM-hour is not just a billing metric; it is a way of translating infrastructure efficiency into revenue. If a feature reduces CPU load enough to fit more virtual machines on the same hardware, the financial value is not abstract. It is embodied in the additional units the provider can sell, and that can be an appropriate starting point for damages analysis. (cafc.uscourts.gov)

• VM-hour pricing can reflect capacity monetization.
• CPU savings can create incremental sellable inventory.
• The royalty base should map to that incremental value, not to unrelated commerce.
• The stronger the technical evidence, the more persuasive the royalty model becomes.

Hypothetical negotiation under modern cloud economics​

In older hardware cases, the royalty base often centered on chips, devices, or modules. Cloud cases are different because the product being sold is often a service layer over infrastructure, not a discrete boxed item. That makes a usage-based metric, like VM-hours, feel more natural than a unit-count tied to a physical device. The market has changed, and damages law has to keep up.
The Federal Circuit did not announce a new doctrine for cloud computing, but its reasoning suggests that courts will be more receptive to damages bases that mirror service economics. For plaintiffs, that is a door to meaningful recovery. For defendants, it is a warning that operational metrics buried in engineering slides may become damages evidence if they reveal the value of the accused feature.

Daubert, Rule 702, and Appellate Deference​

The district court’s mistake was not merely factual; it was legal. The Federal Circuit framed the issue as an erroneous view of the law because the court treated Enplas like a per se exclusion rule. Since evidentiary rulings are reviewed for abuse of discretion, the question on appeal was whether the lower court applied the right legal standard and assessed the evidence correctly. The panel concluded it did not. (cafc.uscourts.gov)
That is important because Daubert motions in patent cases often become strategic weapons. Parties use them not just to police shaky methodology, but to eliminate the other side’s only damages path. The Exafer ruling suggests that courts should be cautious before using Rule 702 to resolve factual disputes over how a technology creates value. If the expert’s theory is grounded in record evidence and economically coherent, exclusion may be too harsh. (cafc.uscourts.gov)
The decision also underscores the difference between reliability and weight. A defendant may have compelling arguments that a damages model overstates the value of the patented feature, but those arguments usually go to cross-examination and jury assessment, not automatic exclusion. The Federal Circuit’s point is not that every incremental-value model is sound, but that a model is not unreliable merely because it uses a nonaccused product as the base. (cafc.uscourts.gov)

Gatekeeping versus fact-finding​

The opinion is a reminder that trial judges must avoid substituting their own damages theory for the expert’s if the expert has a supportable methodology. That line is easy to state and hard to apply, especially in complex technology cases. But the appellate court’s message is clear: when the expert has connected technical performance gains to a revenue metric through admissible evidence, the court should be wary of excluding the testimony wholesale. (cafc.uscourts.gov)

• Daubert is about reliability, not perfection.
• A dispute over assumptions often belongs in cross-examination.
• Courts should distinguish between a flawed model and a legally impossible one.
• The exclusion of the only damages expert can effectively decide the case, so courts should be especially careful.

Why summary judgment fell with the expert ruling​

Once the exclusion was reversed, Microsoft’s absence-of-remedy summary judgment could not stand on the same footing. The district court had treated the exclusion as fatal to Exafer’s damages case, but that premise disappeared when the appellate court reinstated the expert testimony. The Federal Circuit therefore vacated both the order denying discovery and the summary judgment ruling. (cafc.uscourts.gov)
That procedural result is a useful reminder that remedy rulings in patent cases often depend on the survival of expert testimony. If the damages evidence comes back into the case, the plaintiff may regain leverage not only on money but also on settlement and trial posture. The appellate correction reopened the whole litigation.

Competitive Implications for Microsoft and the Cloud Market​

For Microsoft, the immediate consequence is practical: Exafer gets another chance to prove damages based on a theory tied to Azure’s economics. But the broader market impact is bigger. Cloud providers, hyperscalers, and enterprise software vendors all rely on complex infrastructure metrics that can convert efficiency improvements into revenue. This ruling tells them those metrics can be fair game if they are sufficiently linked to the accused technology. (cafc.uscourts.gov)
That is bad news for defendants who have hoped to compartmentalize patent damages into narrow, formally accused components. The more a business model monetizes performance, scaling, and utilization, the more likely a court may accept an incremental-capacity royalty base. That does not mean plaintiffs win by default. It means defendants must build a stronger record showing that the proposed base captures unrelated value or double counts nonpatented features. (cafc.uscourts.gov)
The competition angle is equally important. Cloud providers compete on reliability, throughput, and efficiency as much as on headline price. If patent damages can be tied to the business value of those improvements, then even relatively narrow technical features can carry substantial monetary exposure. That is a signal the market will not ignore.

Enterprise versus consumer effects​

The enterprise impact is likely greater than the consumer impact because enterprise products are usually sold on performance economics. Buyers care about uptime, cost per workload, density, and resource optimization, which gives damages experts more room to argue that efficiency gains have concrete monetary value. Consumer products can also generate similar theories, but the linkage is often less direct and more speculative. (cafc.uscourts.gov)

• Enterprise vendors should expect more damages models built around throughput, capacity, and utilization.
• Consumer product cases may still favor entire market value or apportionment disputes, but the evidence may be thinner.
• Internal telemetry and engineering benchmarks may become more important in discovery.
• Pricing documents can become damages evidence even when they are not tied to a single accused product.

A warning to hyperscalers​

Hyperscalers often think in terms of fleet-wide efficiency. That is exactly what makes the Exafer theory potent. If a feature lets a provider add more workload capacity to the same hardware, the business value can be enormous even if the feature is invisible to end users. The Federal Circuit’s opinion suggests that invisible infrastructure value is still value, and value can support royalties. (cafc.uscourts.gov)
That has strategic consequences. Companies may need to be more careful about how internal presentations describe gains from software and hardware optimizations. A slide showing a 50% or 300% CPU improvement is not just an engineering boast anymore; it can become a damages exhibit. What engineers celebrate, litigators may later quote.

How This Fits the Federal Circuit’s Broader Damages Jurisprudence​

Exafer sits comfortably within a long Federal Circuit tradition that resists bright-line damages formulas while insisting on economic grounding. The court repeatedly emphasized that damages must be case-specific and that reasonable royalty analysis depends on the facts of the alleged infringing activity. That is consistent with earlier opinions cautioning against overbroad entire-market or royalty-base theories, but it also leaves room for sophisticated models tied to real business effects. (cafc.uscourts.gov)
This balance has been the hallmark of Federal Circuit damages law for years. The court does not want patentees to expand claims to unrelated commerce, yet it also knows that modern technology monetization often happens indirectly. The line between a proper surrogate for value and an improper enlargement of the monopoly is often thin, and Exafer reinforces that courts must police that line with attention to evidence rather than slogans. (cafc.uscourts.gov)
The panel’s method also fits the broader trend of appellate skepticism toward rigid Daubert exclusions in patent cases where damages models are highly technical. Judges are increasingly being asked to evaluate whether expert assumptions are wrong or merely contestable. Exafer suggests that, when in doubt, the better course may be to let the jury hear the theory and test it through adversarial process. That is not a soft standard; it is a realistic one.

Relationship to product-based royalties​

The opinion does not make every product-based royalty base admissible. Instead, it restores nuance. If the unaccused product is the direct commercial embodiment of the accused feature’s benefit, the base may be acceptable. If the base is merely a convenient revenue pool with no economic tie to the patented invention, exclusion remains possible. (cafc.uscourts.gov)

• The key question is whether the base reflects value of the invention or merely value in the marketplace.
• Courts will likely scrutinize the chain from feature to efficiency to revenue.
• The stronger the chain, the harder it is to exclude the testimony under Rule 702.
• The weaker the chain, the more likely the theory collapses under Daubert.

Strengths and Opportunities​

The opinion gives patent litigants a clearer roadmap for damages proof in complex technology cases. It also preserves flexibility for plaintiffs whose infringement theories are economically tied to service-layer performance rather than to discrete sold units. For defendants, the decision provides a useful reminder that the right response is often a better factual record, not a categorical objection.

• Clarifies that unaccused products are not automatically barred from the royalty base.
• Affirms that case-specific economic analysis remains central to patent damages.
• Supports damages models based on incremental capacity and operational efficiency.
• Reinforces the importance of internal documents showing technical value.
• Helps plaintiffs in cloud, SaaS, networking, and virtualization disputes.
• Preserves Daubert’s gatekeeping role without letting it become a merits shortcut.
• Signals that apportionment arguments should be sharpened with evidence, not labels.

Risks and Concerns​

The same flexibility that helps legitimate damages claims also creates risk. If courts read Exafer too broadly, plaintiffs may be emboldened to wrap royalties around large revenue pools and call them “incremental.” That could invite more expensive litigation, more motion practice, and more inconsistent rulings over where legitimate economic linkage ends and overreach begins.

• Potential overextension of royalty bases into broad revenue streams.
• Higher discovery burdens as parties fight over internal performance metrics.
• More Daubert motions focused on economic framing rather than methodology alone.
• Risk of double counting when multiple features contribute to the same capacity gain.
• Uncertainty over how much causal connection is enough.
• Greater pressure on district judges to separate reliability from weight.
• Possibility that parties may weaponize efficiency documents without proper context.

Looking Ahead​

The most immediate question is how district courts will apply Exafer outside the specific Azure facts. If a plaintiff can show that an accused feature increases throughput, reduces cost, or expands salable capacity, trial courts may be more willing to admit a royalty model built on the resulting commercial metric. But if the link is attenuated, the opinion will not save the testimony. The real test will be whether the damages story is anchored in the defendant’s own business logic.
A second question is whether this decision shapes negotiation behavior before trial. Sophisticated defendants may become more reluctant to create internal records that quantify efficiency gains in revenue terms, while plaintiffs may become more aggressive in discovery requests aimed at those records. That could make early settlement more likely in some cases, but it could also harden positions when each side believes the other’s documents tell the story. The litigation economy may shift before the courtroom does.

• Watch for district courts to cite Exafer when admitting incremental-value damages models.
• Expect defendants to respond with tighter apportionment and double-counting challenges.
• Look for more disputes over whether internal metrics are truly tied to the accused feature.
• Monitor whether plaintiffs in SaaS and cloud cases begin framing bases around capacity monetization more explicitly.
• Pay attention to whether later panels treat Exafer as a broad damages principle or a fact-specific cloud case.

Exafer v. Microsoft does not rewrite patent damages law, but it does sharpen an important point: the right royalty base is not always the accused product itself. If the evidence shows that the accused technology creates value by enabling more sales, more capacity, or more efficient operations, a court may allow the plaintiff to measure that value through the unaccused product the feature makes more profitable. That is a subtle but powerful message for the patent bar, and it will likely echo through cloud and software disputes for years.

---

Source: The National Law Review Not So Fast, Daubert: Expert Report OK After All