The Federal Circuit’s precedential decision in Netlist, Inc. v. Micron Technology, Inc., No. 24-1707 (Fed. Cir. Sept. 2, 2026), affirmed the Patent Trial and Appeal Board’s determination that all challenged claims of a computer-memory patent were obvious, while leaving unresolved an important question about how far the Board may rely on prior art that was not part of the petitioner’s asserted ground.

The Board had cited Netlist’s own U.S. Patent No. 7,532,537 for a teaching concerning the timing effects of a data buffer, even though Samsung had not included that patent in its obviousness ground. Netlist argued that the Board had crossed the line between using evidence to illuminate the knowledge of a skilled artisan and creating an invalidity theory outside the petition. The Federal Circuit declined to decide whether the Board crossed that line because a separate finding based on the petitioned prior art independently supported the Board’s conclusion.

That disposition is the most important aspect of Netlist. The decision preserves the principle that the petitioner defines the grounds of an inter partes review, but it also demonstrates why identifying questionable Board reasoning may not be enough on appeal. If the Board has an independently sufficient, petition-grounded basis for the same finding, the Federal Circuit may treat any potential reliance on an unasserted reference as harmless.

The Patent and the IPR

Netlist owns U.S. Patent No. 10,949,339, which concerns computer memory systems using dual in-line memory modules, or DIMMs. The claimed architecture places byte-wise buffers between a memory controller and DDR DRAM devices. Data paths through the buffers are ordinarily disabled, electrically isolating the DRAM devices, and are selectively enabled when data must be transferred. Representative claim 1 requires, among other things, that a buffer data path be enabled for a specified period “in accordance with a latency parameter” and that tristate buffers drive write data toward the DRAM devices during that period.

Samsung petitioned for inter partes review of claims 1 through 35, asserting that the claims were obvious over the combination of U.S. Patent Application Publication No. 2006/0277355 to Ellsberry and U.S. Patent No. 7,024,518 to Halbert. Micron filed a petition challenging the same claims over the same prior art and was joined to Samsung’s proceeding. Samsung later settled with Netlist and withdrew from the appeal, leaving Micron to defend the Board’s determination that every challenged claim was unpatentable.

The technical disputes concerned whether Ellsberry disclosed enabling a buffer data path according to a latency parameter and whether the prior art disclosed enabling and disabling the claimed data path. Those questions mattered not only to obviousness, but also to Netlist’s more significant procedural argument concerning the Board’s reliance on an additional patent outside the petition.

Petition Boundaries Matter, but So Does Harmless Error

Inter partes review is not an open-ended inquiry in which the Board may assemble its own invalidity case. In SAS Institute Inc. v. Iancu, the Supreme Court described IPR as a party-directed adversarial proceeding in which the petition defines the challenge.

The Federal Circuit developed the distinction further in Koninklijke Philips N.V. v. Google LLC. The Board may consider a skilled artisan’s general knowledge when evaluating an obviousness theory properly presented in a petition, but it may not create a new obviousness combination using a reference that the petitioner did not assert as part of the ground. More recently, International Business Machines Corp. v. Zillow Group, Inc. reiterated that the Board errs when it holds a claim unpatentable on a ground absent from the petition.

Netlist presented a potentially difficult application of that boundary. The ‘537 patent was not part of Samsung’s asserted obviousness ground, yet the Board cited the patent’s teaching that an additional clock cycle could account for propagation delay through a data buffer. The Board then stated that, “[p]rimed with this teaching,” a skilled artisan would have implemented Ellsberry by budgeting an additional clock cycle for the buffer.

Netlist argued that the ‘537 patent was therefore doing more than illustrating general knowledge. In its view, the Board was using the patent’s substantive teaching to fill a gap in the petitioned obviousness theory. Had the Federal Circuit accepted that characterization and found the teaching necessary to the Board’s result, SAS, Philips, and IBM v. Zillow would have presented a serious obstacle to affirmance.

The Federal Circuit instead avoided deciding the issue. The Board had separately found that Ellsberry itself disclosed the relevant timing adjustment, and Samsung’s expert explained why a skilled artisan would understand Ellsberry’s notation as accounting for propagation delay through the data buffer. Because that independent finding was supported by substantial evidence and remained within Samsung’s asserted ground, the Federal Circuit held that any potential error involving the ‘537 patent was harmless.

The distinction is important. Netlist does not authorize the PTAB to introduce additional references to repair deficiencies in a petition, nor does it clarify precisely when an outside reference moves from permissible evidence of background knowledge to an impermissible component of a new obviousness ground. Instead, the decision adds an appellate dimension to the petition-boundary doctrine: even if part of the Board’s reasoning arguably exceeded the petition, reversal may depend on whether that reasoning was actually necessary to the result.

Alternative Evidentiary Paths Can Control the Appeal

The harmless-error holding was possible because the Board had developed more than one evidentiary basis for its technical conclusions. That aspect of Netlist may be as important in practice as the unresolved petition-scope issue itself.

On the latency limitation, Ellsberry disclosed that initialization commands, including the Posted CAS_n parameter, were sent to and stored in its memory bank switch. Ellsberry also described its memory modules as compatible with Joint Electron Device Engineering Council, or JEDEC, standards. Those standards associated read and write latency with Posted CAS latency, also known as additive latency, and Samsung’s expert explained how a skilled artisan would understand those disclosures together.

The Federal Circuit therefore did not require Ellsberry to use the precise language of Netlist’s claim. The relevant factual inquiry was what Ellsberry communicated to a person of ordinary skill, understood in the technical setting in which the reference operated. Ellsberry’s disclosures, the JEDEC materials, and the expert explanation collectively provided substantial evidence supporting the Board’s finding.

The Board also relied on an Ellsberry notation concerning subtraction of one clock cycle. Samsung’s expert testified that a skilled artisan would understand the notation as reflecting an adjustment for propagation delay introduced by the buffer. That finding proved particularly important because it supplied, within the petitioned prior art, the same proposition for which the Board had also cited Netlist’s ‘537 patent.

A similar pattern appeared in the dispute over whether the prior art disclosed enabling and disabling the claimed data path. Netlist argued that Ellsberry merely enabled or disabled memory banks located outside the switch ASIC rather than connecting and disconnecting paths within the buffer. But Ellsberry’s Figure 2 depicted Ports A and B as components of the switch ASIC, and the reference expressly stated that the memory bank switch included those ports. Figure 4 also depicted data paths passing through bidirectional drivers, which Samsung’s expert explained would enable and disable those paths.

The Board had an additional basis for its finding in Halbert. Halbert depicted bidirectional buffers and data registers using symbols recognized as tristate buffers, and the Board explained why a skilled artisan would have combined those teachings with Ellsberry. Netlist did not adequately defeat that alternative rationale on appeal.

These holdings are better understood together than as separate claim-construction or prior-art disputes. The common lesson is that alternative findings can determine whether an appellate challenge matters. A patent owner may identify a defect in one route through the Board’s analysis and still lose if another route independently reaches the same result.

For petitioners, the corresponding lesson is not simply to make a petition longer or more redundant. It is to ensure that important limitations are supported by well-developed evidentiary paths within the grounds actually asserted. A robust petition-grounded record can matter twice: first when the Board evaluates obviousness, and again when the Federal Circuit asks whether an asserted procedural or evidentiary error affected the outcome.

Technical Context Can Explain What Prior Art Teaches

Netlist also illustrates an important feature of appellate review in technically complex obviousness disputes. Although the Federal Circuit reviews the Board’s ultimate obviousness determination de novo, it reviews the underlying factual findings, including what the prior art teaches, for substantial evidence. Prior-art disclosure is evaluated from the perspective of a person of ordinary skill, which means a reference need not necessarily state every proposition in the terminology later used by a patent claim.

Here, the Federal Circuit accepted the Board’s reliance on Ellsberry’s textual disclosures, technical figures, JEDEC standards, and expert testimony to determine how a skilled artisan would understand the reference. The JEDEC materials and expert testimony were used to explain the significance of Ellsberry’s disclosures within the standardized technological environment in which the memory system operated, rather than as an additional asserted obviousness ground. It explained the significance of what Ellsberry disclosed within the standardized technological environment in which the memory system operated.

That distinction matters in light of the petition-boundary issue elsewhere in the same case. Technical context can assist in determining what an asserted reference teaches, but there remains a legal difference between explaining a petitioned reference and introducing a different substantive prior-art basis for unpatentability. Netlist does not define that boundary more precisely, but the contrast between the court’s acceptance of JEDEC and expert evidence and its refusal to decide the status of the ‘537 patent shows why the characterization of supplementary evidence can matter.

The substantial-evidence standard further limits appellate reconsideration of these technical findings. The Federal Circuit was not deciding whether Netlist’s competing interpretation of Ellsberry was conceivable. It was deciding whether a reasonable factfinder could reach the Board’s interpretation from the evidentiary record. Where patent text, figures, standardized technical materials, and expert testimony point toward the same understanding, that standard gives the Board considerable latitude.

The APA Requires a Discernible Path Through the Board’s Reasoning

Netlist separately argued that the Board violated the Administrative Procedure Act by inadequately explaining its analysis of several dependent claims. That issue produced a more conventional, but still useful, appellate lesson.

Under Alacritech, Inc. v. Intel Corp., the Board must identify the reasons and record basis for its conclusions sufficiently to permit meaningful appellate review. The APA does not, however, require the Board to repeat the same analysis in every section of a final written decision when its reasoning can reasonably be discerned from the decision as a whole.

For dependent claims 7, 16, and 21, the Board had found that Ellsberry’s Posted CAS_n latency parameter was sent to and stored in the switch ASIC during initialization and had separately explained how subsequent operations occurred according to those programmed latency parameters. The claims required module control signals to control timing “in accordance with” the latency parameter, not that the signals themselves contain the parameter. Reading the Board’s findings together, the Federal Circuit concluded that its analytical path was sufficiently clear and supported by substantial evidence.

Netlist also asserted, in only two sentences of its opening brief, that the Board had failed to address numerous other limitations. Citing Monsanto Co. v. Scruggs, the court held that the argument was insufficiently developed for appellate review.

The two APA holdings impose complementary obligations. The Board must provide enough reasoning to permit review, while an appellant claiming that the reasoning is deficient must identify the specific claims, limitations, and analytical gaps at issue. Neither side can rely on generality.

Practical Implications

The principal lesson from Netlist v. Micron is not that the Federal Circuit relaxed the rules governing the scope of an IPR. It did not. SAS, Philips, and IBM v. Zillow continue to constrain the Board to the invalidity case initiated by the petitioner.

What Netlist demonstrates is that the consequences of a possible petition-scope violation depend on the rest of the record. The Board’s reliance on an unasserted patent raised a legitimate doctrinal issue, but resolving that issue would not have changed the result because Ellsberry independently supported the relevant finding. The disputed reasoning therefore became unnecessary to the judgment.

That has practical consequences for both sides of an IPR. Petitioners benefit from building complete evidentiary support within the asserted grounds rather than depending on the Board to fill gaps later. Patent owners challenging Board reasoning should examine whether another factual finding or asserted combination independently supports the same limitation before investing heavily in a petition-scope argument. On appeal, identifying error and establishing prejudice are separate tasks.

The decision also provides a broader lesson for patent disputes involving standardized technologies. A prior-art reference is not necessarily limited to isolated sentences read without their technical surroundings. Figures, incorporated technological conventions, relevant standards, and expert testimony can inform what the reference would communicate to a skilled artisan. But Netlist simultaneously cautions that contextual evidence cannot automatically be converted into a new substantive prior-art combination that the petitioner never presented.

Viewed that way, the opinion is less a routine affirmance of an obviousness determination than a decision about the resilience of PTAB decisions supported by multiple, petition-grounded evidentiary paths. Petition boundaries remain enforceable, but an appellate challenge must reach the reasoning that actually sustains the Board’s result.

Key Takeaways

• SAS, Philips, and IBM v. Zillow continue to prohibit the PTAB from holding claims unpatentable on an invalidity ground absent from the petition.
• The Federal Circuit did not decide whether the Board improperly relied on Netlist’s ‘537 patent because Ellsberry independently supported the relevant timing finding, making any potential error harmless.
• The significance of a petition-scope objection therefore depends partly on prejudice. A patent owner must consider whether another petition-grounded finding independently supports the Board’s conclusion.
• Patent text, figures, technical standards, and expert testimony may collectively establish how a skilled artisan would understand an asserted prior-art reference.
• Alternative evidentiary support within a properly asserted obviousness ground can become decisive on appeal when one part of the Board’s reasoning is challenged.
• The APA requires the Board’s analytical path to be reasonably discernible, while appellants must specifically identify and develop alleged deficiencies in that reasoning.

Related Analysis

• IBM v. Zillow: When “Discloses” Still Means Discloses in IPR Practice examines the precedent expressly relied upon in Netlist and the distinction between evaluating what asserted prior art actually teaches and substituting an invalidity theory absent from the petition.
• Qualcomm v. Apple: Express Reliance on Applicant Admitted Prior Art Bars IPR Ground addresses the related distinction between information that may illuminate the knowledge of a skilled artisan and material that impermissibly becomes part of the substantive basis for an IPR ground.
• Nielsen v. TVision: Federal Circuit Links Claim Breadth to Analogous Art and Obviousness provides related analysis of APA constraints on PTAB proceedings and the distinction between developing an existing theory and changing the theory placed in dispute.
• Intellectual Pixels v. Sony: Federal Circuit Clarifies What Remains Open After a PTAB Remand examines related questions involving substantial-evidence review and procedural limits on PTAB decision-making.

By Charles Gideon Korrell