The US Court of Appeals for the Federal Circuit has affirmed three Patent Trial and Appeal Board final written decisions holding that Google failed to prove the challenged claims of three virtual-reality viewer patents obvious. The decision in Google LLC v. DDC Technology, LLC, Nos. 2025-1315, 2025-1316 and 2025-1330, was issued as precedential on 5 November 2026 by a panel of Circuit Judges Reyna, Hughes and Stark, with Judge Reyna writing. The appeals came from IPR2023-00707, IPR2023-00709 and IPR2023-00711, directed to US Patent Nos. 9,420,075, 10,528,199 and 11,093,001 — a smartphone-based VR viewer with an input mechanism that reaches through the housing to touch the phone’s screen. Google ran the same combination in all three proceedings: Japanese Patent Publication JP 3176588 (“Kondo”) with US Patent Application Publication No. 2013/0082963 (“Chu”).
What makes the decision worth reading is how the patent owner won. On claim 1 of the ‘075 and ‘199 patents, the Board agreed with Google on the technical merits: the Kondo–Chu combination teaches every limitation, and a skilled artisan would have been motivated to combine the references with a reasonable expectation of success. The claims survived anyway. The Board found that DDC established copying, which it weighed “heavily” and described as strong evidence of nonobviousness, and industry praise, to which it gave “moderate” weight — and concluded that on balance the nonobviousness evidence outweighed Google’s case. The Federal Circuit found no legal error, stating that on this record it could not say as a matter of law that Google’s obviousness evidence was strong enough to outweigh DDC’s.
Objective indicia are not a consolation prize
The practical message for patent owners is that losing the prior-art argument is not the same as losing the IPR. The panel restated, citing Stratoflex v. Aeroquip and WBIP v. Kohler, that secondary considerations must always be weighed when present and that the argument that they can never overcome a strong obviousness showing is contrary to law. It also rejected the framing petitioners lean on: a “prima facie case” is a tool of examination that allocates burdens between examiner and applicant (In re Oetiker, In re Brandt), and the Federal Circuit has emphatically rejected any formal burden-shifting under Graham (OSRAM Sylvania), which “does not apply in the adjudicatory context of an IPR” (In re Magnum Oil Tools).
The rule from Tokai v. Easton — that a strong prima facie case cannot be overcome by a far weaker showing of objective indicia — survives, but the court showed exactly where it stops. It did not apply here because the Board never characterised Google’s evidence as strong, and reached its motivation-to-combine finding only after rejecting two of Google’s three rationales and holding a contested trial on the issue. Contrast Ecolochem, where the obviousness showing was conceded. If you are the patent owner, that is the gap to open: force every motivation rationale to be litigated, concede nothing, and build the indicia record even when the art looks bad. If you are the petitioner, a Tokai argument is only available if the Board actually called your case strong — and Google’s brief characterising its own evidence as “overwhelming” and DDC’s as “marginal” got no traction, because the Board had said the opposite.
What copying evidence has to compare
Google’s lead attack was that the Board compared the products to the claims rather than to each other. The court agreed with the standard and disagreed on the facts. Under Iron Grip Barbell v. USA Sports, not every competing product within a patent’s scope is evidence of copying — otherwise every infringement suit would confirm nonobviousness — so the comparison runs between the patented device and the alleged copy, not between either and the claims. Copying can be shown through internal documents, direct evidence such as disassembling a patented prototype and using photographs of it as a blueprint, or access plus substantial similarity to the patented product. The Board had done the product-to-product comparison, with pictures, between Google’s GCV2 viewer, DDC’s DODOcase VR 1.2 viewer and inventor Patrick Buckley’s prototype.
Equally important for anyone assembling such a record: the Board identified two real differences between the input mechanisms, called them “minor, non-material modifications” affecting attachment and materials but not fundamental features, and still found copying. The Federal Circuit affirmed, pointing to Medtronic v. Teleflex Innovations, 70 F.4th 1331, 1341 (Fed. Cir. 2023), where it accepted substantial similarity despite acknowledged differences. The consequence is that the material deciding an IPR filed in 2023 was generated years earlier in a business-development file — who saw the prototype, when, under what terms, and what the dated demo material showed. Preserve the physical prototype, the dated videos and the correspondence that proves access; that file is the asset, and it cannot be reconstructed after the petition arrives.
Praise from the other side’s own engineers counted
The industry-praise finding rested on emails from three Google engineers replying to Buckley, then a prospective business partner, after he shared videos of his prototype: two called the button genius or ingenious, and a third said it looked really neat. The Board discounted the third as politeness, credited the other two at moderate weight, and expressly offset them for the on-going business relationship, noting the engineers had interpersonal and business reasons to respond positively. Google’s attempt to dismiss the emails as self-serving under In re Cree and Geo M. Martin failed on a distinction worth remembering: those cases involved parties praising their own work, whereas here the engineers were praising a third party’s prototype. Note also that the Board tied the praise to the claimed input mechanism specifically, not to the product at large — untethered enthusiasm about a product is far weaker currency.
There is an appellate lesson buried in the same section. The weight assigned to objective indicia is a factual determination reviewed for substantial evidence, and the court recorded that Google did not challenge the weight the Board gave copying and praise — only the ultimate balancing. It therefore declined to disturb those assignments. A petitioner who intends to attack an indicia-driven loss has to attack the weight findings themselves, in the brief and at argument, not just the legal conclusion that follows from them.
Two procedural traps that cost Google claims 27 and 28
On claim 27 of the ‘001 patent, Google argued the Board had sprung an implicit construction of “made from a flexible material” on it in the final written decision, in violation of the Administrative Procedure Act. The court distinguished SAS Institute v. ComplementSoft, where the problem was a construction the parties could not have anticipated. Here the institution decision gave no construction, recorded Google’s position, made clear the Board did not accept it, and welcomed further argument and evidence at trial — and Google never came back on the claim. A construction a petitioner was invited to address is not an unexpected one, and the Board cannot be faulted under the APA for the petitioner’s failure to develop the position. The court also corrected a mischaracterisation: the Board never found Chu’s shaft inflexible, only that Google had not shown it was flexible.
Claim 28 failed on a mapping error compounded by the same silence. Google relied on Chu’s biasing member 40 as the “memory material” and as part of the “linkage element”, so the limitation requiring the memory material to be coupled to the linkage element read as that component being coupled to itself. The Board flagged the inconsistency at institution; Google did not address it; and its proposed construction of “coupled to”, raised for the first time on appeal, was held forfeited under Netflix v. DivX, 84 F.4th 1371, 1378 (Fed. Cir. 2023). The operational rule for petition drafting is unglamorous and expensive to ignore: never map a single prior-art component to two claim elements that the claim requires to be joined to each other, and treat every invitation in an institution decision as a deadline, because an unanswered invitation converts into both an APA defence for the Board and a forfeiture on appeal.
Claim-construction notes for drafters
Two constructions were affirmed and both reward careful specification practice. “Disposed on” was held not to require a direct connection but to exclude connections that are “far removed” — the claim paired each surface with its own portion, and the specification used “disposed on” interchangeably with “located on”, which defeated Google’s broader reading. And “lever” in claim 20 was construed as “a stiff body pivoted on a fulcrum”, the Board having credited the Oxford Dictionary of Mechanical Engineering over the general-purpose dictionaries both sides filed, supported by the fulcrum action shown in Figures 7A–7E. Paired structural recitations narrow a term whether you want them to or not, and a technical dictionary matched to the art will usually beat a general one at the Board.
For portfolio owners, the sequencing that mattered here is the point to act on. Objective indicia only exist if someone preserved the evidence long before the challenge; prior-art strength is assessed on a record that is fixed by the petition and the institution decision. If you are facing or planning an inter partes review, the two workstreams are an exhaustive prior-art and litigation search on the merits and a disciplined invalidity and validity assessment that prices in what the secondary-considerations record will actually support. Google v. DDC Technology is a reminder that the second workstream decided three proceedings in which the first one went the challenger’s way.