¹ Lawrence Der & Behzad Razavi, A 2-GHz CMOS Image-Reject Receiver with LMS Calibration, 38(2) IEEE J. Solid-State Circuits 167 (2003).

² Behzad Razavi, Rf Microelectronics (1998).

³ Valla et al., A 72-mW CMOS 802.11a Direct Con- version Front-End With 3.5-dB NF and 200-KhZ 1/f Noise Corner, 40(4) IEEE J. Solid-State Circuits 970 (2005). Case: 20-1664 Document: 103 Page: 5 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 5 28, and 31 after disagreeing that these claims were obvious in view of Der, Razavi, and Burgener. 4 Id. at *28, *30. After Intel appealed, Qualcomm moved to dismiss for lack of standing. We denied that motion and directed the parties to address standing in their briefs. 5 We have juris- diction over final Board decisions under 28 U.S.C. § 1295(a)(4)(A). As discussed below, Intel has standing to invoke that jurisdiction. DISCUSSION I Before proceeding to the merits, we address the thresh- old question of our jurisdiction. We are limited to deciding “cases” and “controversies.” U.S. CONST. art. III, § 2. This “fundamental limitation” is reflected in the Article III standing requirement. Summers v. Earth Island Inst., 555 U.S. 488, 493 (2009). Although “not necessarily a re- quirement to appear before an administrative agency,” it “kicks in when a party seeks review in a federal court.” Ap- ple Inc. v. Qualcomm Inc., 17 F.4th 1131, 1135–36 (Fed. Cir. 2021) (cleaned up). That party must demon- strate: (1) an “injury in fact” (2) “fairly traceable” to the de- fendant’s challenged conduct and (3) “likely to be redressed by a favorable judicial decision.” Spokeo, Inc. v. Robins, 578 U.S. 330, 338 (2016). Relevant here, an injury in fact must be “concrete and particularized and actual or imminent, not conjectural or hypothetical.” Id. at 339 (cleaned up). That’s generally so when an IPR petitioner “has engaged in, is engaging in, or will likely engage in activity that would give rise to a pos- sible infringement suit.” Grit Energy Sols., LLC v. Oren 4 U.S. Patent No. 6,804,502.

⁵ We also denied Apple’s motions to intervene and to participate as amicus curiae. Case: 20-1664 Document: 103 Page: 6 Filed: 12/28/2021 6 INTEL CORPORATION v. QUALCOMM INCORPORATED Techs., LLC, 957 F.3d 1309, 1319 (Fed. Cir. 2020) (cleaned up). The appellant in Grit, for instance, engaged in acts that not only could have but “did give rise to an infringe- ment suit”—a suit dismissed without prejudice, leaving the patentee “free to reassert those infringement claims.” Id. at 1320. Intel’s predicament here is similar. Although Qualcomm didn’t sue Intel for infringement, Qualcomm has not disputed that it mapped the ’043 patent claims to an Intel product (and only an Intel product) in a prior suit against Apple. See Appellant’s Br. 46; J.A. 4713–65; J.A. 4794–96; Intel’s Response to Qualcomm’s Motion to Dismiss, ECF No. 40 at 5 (and cited exhibits)). Like in Grit, therefore, Intel’s acts “did give rise to an infringement suit.” 957 F.3d at 1320. In Intel’s words, Qualcomm “al- ready has engaged in litigation involving that technology and this patent.” Reply Br. 29. It is of no moment that the suit wasn’t against Intel, as Intel “need not face a specific threat of infringement.” Grit, 957 F.3d at 1319 (cleaned up). Nor does it matter that the suit settled in 2019. See J.A. 4594–95. True, such settle- ments can deprive parties of standing. E.g., Apple, 17 F.4th at 1134; Apple Inc. v. Qualcomm Inc., 992 F.3d 1378, 1385 (Fed. Cir. 2021). But Intel isn’t a party to this one. In contrast, Intel was informed that “Qualcomm is not offering a covenant not to sue.” J.A. 4713. Although that refusal isn’t on its own “sufficient to create an actual con- troversy,” Prasco, LLC v. Medicis Pharm. Corp., 537 F.3d 1329, 1341 (Fed. Cir. 2008), it reinforces the analogy to Grit, where the patentee declined to stipulate that it would not reassert its previous infringement allegations, 957 F.3d at 1320 n.3. Moreover, Intel represents that it continues to sell the relevant products to Apple and at least one other customer, and that in doing so it must “address[] the ’043 patent and the risk of an infringement suit by Qualcomm.” Case: 20-1664 Document: 103 Page: 7 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 7 J.A. 4794–96; see J.A. 4785. 6 Because Intel’s risks trans- cend mere conjecture or hypothesis, see Spokeo, 578 U.S. at 339, we conclude that Intel has standing. See also Intel Corp. v. Qualcomm Inc., No. 20-1828, slip op. at 9–10 (Fed. Cir. Dec. 28, 2021). II On to the merits. We start with the parties’ dispute over the proper construction of the phrase “radio frequency input signal” in ’043 patent claims 17, 19, and 21. Before the Board, Intel said this phrase “should take its ordinary meaning of an input signal having a radio frequency.” Fi- nal Written Decision, at *7. Qualcomm disagreed, arguing that a skilled artisan reading the patent would have un- derstood the phrase to reference the radio frequency signal that is received before down-conversion: “a signal centered at a carrier frequency at which the signal was transmit- ted/received.” Id. at *6. The upshot is that Intel’s proposal (and not Qual- comm’s) covers a signal called the intermediate frequency (“IF”) signal in the two-stage “super heterodyne” architec- ture of prior-art reference Der. Unlike the one-stage “ho- modyne” architecture disclosed in the ’043 patent, receivers in super heterodyne architectures like Der demodulate a carrier signal in two stages: first to that IF signal, and then to baseband. Adopting Qualcomm’s proposal, the Board de- termined that Der didn’t anticipate claims 17, 19, and 21. Id. at *15. “We review claim construction based on intrinsic evi- dence de novo and review any findings of fact regarding ex- trinsic evidence for clear error.” SpeedTrack, Inc. v.

⁶ Given Intel’s past and ongoing acts, Apple’s 2019 acquisition of “the majority of Intel’s smartphone modem business” also doesn’t negate Intel’s standing. See J.A. 4597–607. Case: 20-1664 Document: 103 Page: 8 Filed: 12/28/2021 8 INTEL CORPORATION v. QUALCOMM INCORPORATED Amazon.com, Inc., 998 F.3d 1373, 1378 (Fed. Cir. 2021) (cit- ing Teva Pharms. USA, Inc. v. Sandoz, Inc., 574 U.S. 318, 331–32 (2015)). Applying the “broadest reasonable inter- pretation” standard, 7 we affirm the Board’s construction. Claim terms are generally accorded their ordinary meaning—that is, their meaning to a skilled artisan at the time of the invention. Phillips v. AWH Corp., 415 F.3d 1303, 1312–13 (Fed. Cir. 2005) (en banc). This approach “provides an objective baseline” for our inquiry. Id. at 1313. To that end, we consult the sources available to such artisans, including “the words of the claims them- selves, the remainder of the specification, the prosecution history, and extrinsic evidence concerning relevant scien- tific principles, the meaning of technical terms, and the state of the art.” Id. at 1314 (quoting Innova/Pure Water, Inc. v. Safari Water Filtration Sys., Inc., 381 F.3d 1111, 1116 (Fed. Cir. 2004)). “Importantly,” skilled artisans are “deemed to read the claim term . . . in the context of the entire patent.” Id. at 1313. Even when seeking the “broad- est reasonable construction in light of the specification,” 37 C.F.R. § 42.100(b) (2017), we still give words “their plain meaning” unless “inconsistent with the specification and prosecution history.” Arista Networks, Inc. v. Cisco Sys., Inc., 908 F.3d 792, 796–98 (Fed. Cir. 2018) (rejecting con- struction as “overly broad, even under the broadest reason- able interpretation standard”). “Above all, the broadest reasonable interpretation must be reasonable in light of the claims and specification.” PPC Broadband, Inc. v. Corning Optical Commc’ns RF, LLC, 815 F.3d 747, 755 (Fed. Cir. 2016). Considering the disputed phrase “radio frequency in- put signal” in a vacuum, both proposals have some appeal.

⁷ Although that standard applies in this case, it has been superseded by the standard “used to construe the claim[s] in a civil action.” 37 C.F.R. § 42.100(b) (2018). Case: 20-1664 Document: 103 Page: 9 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 9 From that vantage, it’s plausible that the disputed phrase equals the sum of its parts as Intel contends—a “signal” that is “within the range defined as the radio frequency spectrum” and is an “input to a circuit or component re- ceiver part.” E.g., Appellant’s Br. 25. But the phrase could just as easily mean something more specific, referring to a particular signal in the receiving process, as Qualcomm contends. Even without considering the surrounding claim language or the rest of the patent document, we note that it is not always appropriate to break down a phrase and give it an interpretation that is merely the sum of its parts. See FCC v. AT&T Inc., 562 U.S. 397, 406 (2011) (rejecting interpretation of “personal privacy” as “simply the sum of its two words”). In any event, and decisively, our inquiry is not limited to an analysis of the phrase in isolation. See, e.g., Hockerson-Halberstadt, Inc. v. Converse Inc., 183 F.3d 1369, 1374 (Fed. Cir. 1999) (“Proper claim construction . . . demands interpretation of the entire claim in context, not a single element in isolation.”). “[A] term can be defined only in a way that comports with the instrument as a whole.” Markman v. Westview Instruments, Inc., 517 U.S. 370, 389 (1996). We therefore continue by examining the surrounding claim language. Phillips, 415 F.3d at 1314. Claim 17 below (emphasis added) has a tripartite structure: it describes (a) the signal received, (b) the signal output, and (c) the switching functionality: 17. A method comprising: (a) receiving a radio frequency input signal onto a source lead of a first transistor and onto a source lead of a second transistor; (b) outputting a baseband signal from a drain lead of the first transistor, wherein the baseband signal has a current with a magnitude; and Case: 20-1664 Document: 103 Page: 10 Filed: 12/28/2021 10 INTEL CORPORATION v. QUALCOMM INCORPORATED (c) increasing the magnitude of the current of the baseband signal by coupling the drain lead of the first transistor to a drain lead of the second tran- sistor, wherein the drain lead of the first transistor is coupled to the drain lead of the second transistor by closing a switch that is coupled to both the drain lead of the first transistor and the drain lead of the second transistor. This language points in favor of Qualcomm’s reading adopted by the Board. First, if the “radio frequency input signal” refers to just any radio frequency signal that is an input, then the word “input” makes no contribution to the claim. Omitting it would change nothing, given the mixer’s “receiving” of the signal “onto a source lead.” See Antonin Scalia & Bryan A. Garner, Reading Law 176 (2012) (“Be- cause legal drafters should not include words that have no effect, courts avoid a reading that renders some words al- together redundant.”); see also Intel, No. 20-1828, slip op. at 13 (collecting cases). Second, part (a) of the claim doesn’t strictly parallel part (b), raising the question why part (a) doesn’t instead say “inputting a radio frequency signal” if indeed “a radio frequency input signal” is just a radio fre- quency signal that is input. These linguistic clues suggest that “radio frequency input signal,” to the relevant audi- ence, refers to the signal entering the device as a whole, not (as Intel proposes) to any radio frequency signal entering any component. The specification provides further support for the Board’s reading. It consistently and repeatedly uses “radio frequency input signal” to reference the carrier frequency signal received at the antenna, amplified by the low-noise amplifiers, and received at the mixer’s transistors for direct down-conversion to baseband. For example, Figure 2 (see annotated versions above) shows a signal 32 that’s received at antenna 16 and that, Qualcomm has not disputed, is the carrier frequency signal sent over the air. Indeed, the pa- tent in the context of Figure 1 expressly calls signal 32 “the Case: 20-1664 Document: 103 Page: 11 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 11 carrier signal” that is “downconvert[ed]” to “baseband.” Id. at col. 5 ll. 43–58. 8 And then in its subsequent discussion of Figure 2, the patent explains that “[r]adio frequency in- put signal 32 is amplified by LNAs 17–19 and then received onto the source leads of transistors 39–40,” after which point “a baseband signal current 84 is output from the drain lead [of the] first transistor 39.” Id. at col. 10 l. 64–col. 11 l. 8 (emphasis added); see also id. at col. 3 ll. 52–58 (similar); id. at col. 3 ll. 18–30 (similar). That dis- cussion also references and parallels the flowchart of Fig- ure 11, which as its first step recites “receive a radio frequency input signal onto a source lead of a first transis- tor and onto a source lead of a second transistor” followed by direct down-conversion of that signal to baseband. Id. at Fig. 11 (capitalization normalized). We find this con- sistency compelling. Intel objects that the Board’s construction limits the or- dinary meaning of the claims to the patent’s embodiments, an approach we’ve “expressly rejected.” Info-Hold, Inc. v. Applied Media Techs. Corp., 783 F.3d 1262, 1267 (Fed. Cir. 2015). Just because the patent doesn’t “specifically use the term ‘RF input signal’ to describe the input into a second- stage mixing circuit,” Intel says, that doesn’t limit the pa- tent “to homodyne receiver architectures” and exclude “su- per heterodyne” ones. Appellant’s Br. 26–27. But this presupposes the point it purports to prove. It takes for granted that Intel has the ordinary meaning right. In this case, however, the question is what the contextually correct meaning is, not whether anything affirmatively limits an undisputed ordinary meaning.

⁸ Elsewhere, the patent also suggests that the inven- tion could be used in reverse to “modulate baseband data” up to a radio frequency signal “transmitted from the wire- less device” (i.e., a carrier signal). ’043 patent col. 11 ll. 56–61 (emphasis added). Case: 20-1664 Document: 103 Page: 12 Filed: 12/28/2021 12 INTEL CORPORATION v. QUALCOMM INCORPORATED And although the patent doesn’t disclose any super het- erodyne embodiments, it does (once) mention intermediate frequencies in a way that favors the Board’s construction. As background, it states: “Receivers for wireless communi- cation systems typically require low noise amplifiers (LNAs) followed by double-balanced mixers to pre-amplify incoming signals and to down-convert those signals to an appropriate intermediate frequency (IF) or baseband fre- quency.” ’043 patent col. 1 ll. 11–15. This statement doesn’t describe an intermediate frequency as a species of “radio frequency input signal.” Rather, it uses a distinct label, “intermediate frequency (IF).” Generally speaking, “[a] word or phrase is presumed to bear the same meaning throughout a text; a material variation in terms suggests a variation in meaning.” Scalia & Garner, supra, at 170; see PPC, 815 F.3d at 752 (noting this canon is “employed in both statutory interpretation and claim construction”). Here we discern such a material variation, suggesting that “intermediate frequency (IF)” means something different from “radio frequency input signal.” Further, beyond indicating a material variation in terms, the structure of the patent’s “IF” passage also hints that an intermediate frequency isn’t covered by the claimed “radio frequency input signal.” By describing both the “in- termediate frequency (IF)”and the “baseband frequency” as results of down-converting “incoming” signals, the passage parallels the claim’s structure—which reflects “receiving” an incoming “radio frequency input signal” and outputting the result of down-conversion: a “baseband signal.” Id. at claim17. As Intel’s counsel observed, the claim structure “juxtapos[es]” the “radio frequency input signal” with the “baseband.” Oral Arg. at 2:30–50, No. 20-1664. 9 There- fore, although the passage may not explicitly distinguish 9 https://oralarguments.cafc.uscourts.gov/de- fault.aspx?fl=20-1664_10072021.mp3. Case: 20-1664 Document: 103 Page: 13 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 13 IF from RF, see Oral Arg. at 30:15–32:37; Final Written De- cision, at *7, it treats intermediate frequencies as akin to baseband signals—which map to the “outputting” phrase of part (b) rather than the “radio frequency input signal” of part (a). Indeed, this framing illuminates how, rather than be- ing limited to homodyne architectures as Intel suggests, the Board’s construction can indeed cover super hetero- dyne architectures. The first stage would “receiv[e]” a “ra- dio frequency input signal” that, per the Board’s construction, is centered at a carrier frequency, and the second stage would “output[]” a baseband signal. Id. at claim 17. In between would be the work of down-conver- sion, first to an intermediate frequency and then to the “output[]” baseband signal of the claim. Id. The effect of the Board’s construction is not, therefore, that all super heterodyne architectures are excluded, but more modestly that the intermediate signal of such an architecture simply isn’t the “radio frequency input signal” of the claims. In sum, while Intel’s interpretation may have superfi- cial appeal, Qualcomm’s better reflects the usage of “radio frequency input signal” in the intrinsic record. It prevents the word “input” from being redundant in the claim and comports with the claim’s tripartite structure. And it re- flects the specification’s repeated use of the disputed phrase to reference incoming carrier signals before down- conversion. See Arista, 908 F.3d at 798 (construing “broad- cast” to mean “a transmission to one or more devices using a multicast address” based on “the specification’s con- sistent focus on broadcasting via a multicast address”). Be- cause the Board’s construction gives effect to the contextually appropriate meaning of “radio frequency in- put signal,” we affirm it. III Next we turn to obviousness, addressing first the orig- inally challenged claims and second the substitute claims. Case: 20-1664 Document: 103 Page: 14 Filed: 12/28/2021 14 INTEL CORPORATION v. QUALCOMM INCORPORATED The ultimate judgment of obviousness under 35 U.S.C. § 103 is a legal determination based on underlying factual inquiries. “[T]he scope and content of the prior art are to be determined; differences between the prior art and the claims at issue are to be ascertained; and the level of ordi- nary skill in the pertinent art resolved.” Graham v. John Deere Co. of Kan. City, 383 U.S. 1, 17 (1966). Also, “[s]uch secondary considerations as commercial success, long felt but unsolved needs, failure of others, etc., might be utilized to give light to the circumstances surrounding the origin of the subject matter sought to be patented.” Id. at 17–18. In cases where “the question is whether a patent claim- ing the combination of elements of prior art is obvious,” “it can be important to identify a reason that would have prompted a person of ordinary skill in the relevant field to combine the elements in the way the claimed new inven- tion does.” KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 417–18 (2007). That, too, is a factual inquiry. ZUP, LLC v. Nash Mfg., Inc., 896 F.3d 1365, 1371 (Fed. Cir. 2018). We review the Board’s factual findings for substantial evi- dence and its legal conclusions de novo. Donner Tech., LLC v. Pro Stage Gear, LLC, 979 F.3d 1353, 1358 (Fed. Cir. 2020). Both obviousness issues on appeal involve “whether there was an apparent reason to combine” the prior-art el- ements. KSR, 550 U.S. at 418. In KSR (decided over 10 years ago now), the Supreme Court rejected the “teaching, suggestion, or motivation” (“TSM”) test that we inherited from our predecessor court. Id. at 407, 418. Under that test, “a patent claim [was] only proved obvious if some mo- tivation or suggestion to combine the prior art teachings [could] be found in the prior art, the nature of the problem, or the knowledge of a person having ordinary skill in the art.” Id. at 407 (cleaned up). The Court acknowledged that the TSM test “captured a helpful insight” by addressing the intuition that “a patent composed of several elements is not proved obvious merely by demonstrating that each of its Case: 20-1664 Document: 103 Page: 15 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 15 elements was, independently, known in the prior art.” Id. at 418. “Helpful insights, however, need not become rigid and mandatory formulas.” Id. at 419. Therefore, the Court instead adopted an “expansive and flexible approach.” Id. at 415. No longer would “[t]he obviousness analysis . . . be confined by a formalistic conception of the words teaching, suggestion, and motivation, or by overemphasis on the im- portance of published articles and the explicit content of issued patents.” Id. at 419. We assess Intel’s challenges to the Board’s obviousness determinations under the principles in KSR. We reject the first, (A) that the Board didn’t satisfactorily explain why a skilled artisan would not have combined Der with Valla. We accept the second, (B) that a skilled artisan would have had reason to combine Burgener with Der, because the Board’s decision rejecting that rationale is not supported by substantial evidence. A We begin with the Board’s determination that Intel didn’t prove claims 6, 17–19, and 21 were obvious in light of Der and Valla (nor, for the same reason, substitute claims 27, 28, and 31 in light of Der, Valla, and Burgener). Before the Board, Intel argued that a skilled artisan would have “modified Valla to include two parallel passive mixers having outputs coupled by a gain control block con- sisting of switches . . . to achieve the advantages of Der and Valla.” Final Written Decision, at *19. Qualcomm disa- greed, reasoning that “the transistors from Der would im- pair the low impedance that Valla seeks for its amplifier.” Id. at *20. The Board agreed with Qualcomm. As the Board explained, “Valla discusses the importance of a pas- sive mixer with a ‘low impedance’ load” for obtaining the system’s benefits, calling this a “key feature.” Id. at *21. Along the way, it rejected Intel’s riposte: that in “triode mode” Der’s transistors “would have very low impedance Case: 20-1664 Document: 103 Page: 16 Filed: 12/28/2021 16 INTEL CORPORATION v. QUALCOMM INCORPORATED and would have no adverse impact” on Valla’s circuit. Id. at *20 (quoting J.A. 1834 ¶ 61). On appeal, Intel doesn’t argue that the Board’s deter- mination lacks substantial-evidence support. 10 Rather, it says the Board didn’t “apply the proper mode of legal anal- ysis [of] weigh[ing] the putative disadvantage . . . against the combination’s undisputed benefits.” Reply Br. 18. As an initial matter, we agree that “simultaneous advantages and disadvantages . . . do[] not necessarily obviate motiva- tion to combine.” Medichem, S.A. v. Rolabo, S.L., 437 F.3d 1157, 1165 (Fed. Cir. 2006). In KSR itself, for example, ev- idence indicating that a prior-art product was “bulky, com- plex, and expensive” wasn’t on its own enough to dismiss it as “too flawed to upgrade.” 550 U.S. at 425–26. But here we conclude the Board’s analysis is adequate, applying the proper standard. Specifically, the Board noted that Intel’s expert didn’t disagree with Qualcomm’s expert as to what would happen in the non-triode modes. Final Written Decision, at *21–22. The thrust of both experts’ testimony, the Board explained, is that “although there are some conditions . . . that result in Der’s transistor M9 operating in the triode mode . . . over most of the operating conditions the transistor is in a dif- ferent mode.” Id. at *22. Therefore, the Board concluded, a skilled artisan would have lacked motivation to combine Valla with Der because that “would impair Valla’s need for low impedance during the majority of operation.” Id. (em- phasis added). On this record, we’re satisfied that the Board weighed the competing evidence regarding the

¹⁰ Oral Arg. at 14:06–32 (Q: “So you’re not making a substantial-evidence challenge?” A: “No your Honor, we’re not.”); Oral Arg. at 15:34–49 (Q: “So, just to be clear, if we don’t find that the Board applied the incorrect legal test for obviousness, you’re not saying their conclusion lacks sub- stantial evidence?” A: “That is correct, your Honor.”). Case: 20-1664 Document: 103 Page: 17 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 17 relevant tradeoffs and concluded, based on the stated im- portance of low impedance as “key” to Valla, that negating this benefit during “the majority of the operation” would have outweighed any reason to combine. 11 B Last, we conclude that substantial evidence does not support the Board’s determination that a skilled artisan would have lacked reason to combine Der with Burgener (in the Der-Burgener-Razavi combination) to achieve sub- stitute claims 27, 28, and 31. Those claims depend on sub- stitute claim 26, which adds a single limitation to claim 1 (additions italicized; deletions bracketed): 26. (Proposed Substitute for Claim 1) A device com- prising: (a) an amplifier having an output lead; (b) a first transistor having a source lead, a drain lead and a gate lead; (c) a second transistor having a source lead, a drain lead and a gate lead, wherein the output lead of the amplifier is coupled to the source lead of the first

¹¹ The Board also said Intel relied “at least in part[] on impermissible hindsight” because it “provide[d] no sup- port—whether from Valla, Der, or some other source” for a benefit also identified in the patent. Final Written Deci- sion, at *22. We’re skeptical that this necessarily trans- lates to hindsight. There’s another possibility: that without “seek[ing] out precise teachings,” Intel merely took “account of the inferences and creative steps” a skilled ar- tisan would employ. KSR, 550 U.S. at 418. At any rate, Intel doesn’t press a substantial-evidence challenge (see above), so we affirm the Board’s Der-Valla determination. Case: 20-1664 Document: 103 Page: 18 Filed: 12/28/2021 18 INTEL CORPORATION v. QUALCOMM INCORPORATED transistor and to the source lead of the second tran- sistor; [[and]] (d) a first switch, wherein the drain lead of the first transistor is coupled to the drain lead of the second transistor through the first switch when the first switch is closed, and wherein an oscillating signal is present on the gate lead of the first transistor and on the gate lead of the second transistor; and (e) a second switch, wherein the drain lead of the second transistor is coupled to ground through the second switch when the second switch is closed. J.A. 3763 (alterations in original). Intel relied on Burgener to disclose this added limita- tion. Burgener describes as background a “prior art switch” that includes “a switching transistor” and “a shunt- ing transistor.” J.A. 1873. Via that switch, Burgener states, “RF signals are either routed from an RF input node” to “an RF output node,” or else are “shunted to ground through the shunting transistor M2 7.” J.A. 1873. From there, Intel argued that a skilled artisan would have had reason to combine Burgener’s switch with Der’s re- ceiver architecture (and the teachings of Razavi)—render- ing these substitute claims unpatentable. Intel’s asserted rationale proceeds in two steps. First, a skilled artisan would have sought to improve energy efficiency by turning off a portion of Der’s circuit when not in use. Second, doing so would have triggered a known problem to which Bur- gener’s switch was a predictable solution. The Board disagreed for three reasons: (1) that Intel’s energy-efficiency theory was too generic, (2) that Burgener described the switch as prior art with shortcomings, and (3) that the combination would render Der unsuitable for its intended purpose. Under applicable legal principles, none of the Board’s reasons are supported by substantial evidence. Case: 20-1664 Document: 103 Page: 19 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 19 20 INTEL CORPORATION v. QUALCOMM INCORPORATED Of course, “generic” might also have a different sense: conclusory. Certainly, “[c]onclusory expert testimony does not qualify as substantial evidence,” TQ Delta, LLC v. Cisco Sys., Inc., 942 F.3d 1352, 1358–59 (Fed. Cir. 2019), as demonstrated in ActiveVideo. But unlike ActiveVideo, this isn’t a case “where the motivation of increased effi- ciency is asserted so generically as to be legally insuffi- cient.” Huawei Techs. Co. v. Iancu, 813 F. App’x 505, 510 (Fed. Cir. 2020) (describing ActiveVideo). The testimony in ActiveVideo bore “no relation to any specific combination of prior art elements . . . from specific references” and didn’t explain why a skilled artisan would have combined them “in the way the claimed invention does,” 694 F.3d at 1328, but here Intel’s expert indicated precisely how and why a skilled artisan would have combined the references. Recall that Der has a super heterodyne architecture. Der’s second stage “employs two parallel mixers . . . a main mixer and a variable-gain mixer,” along with “a gain con- trol circuit at the output of the variable-gain mixer.” Ap- pellee’s Br. 42 (cleaned up). According to Intel’s expert, a skilled artisan “would have recognized that, in cases where gain control is not required (e.g. when the gain of the ‘main mixer’ alone is sufficient), it would have been advantageous to disable the variable-gain mixer to save power.” J.A. 1840 ¶ 70 (capitalization normalized). Specifically, such an artisan “would have combined the mixer circuit of Der Figure 9(b) with the shunt transistor M2 7 of Case: 20-1664 Document: 103 Page: 21 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 21 Burgener,” as shown in Intel’s annotated Figure 9(b) below (which incorporates Burgener’s switches into Der’s circuit): J.A. 3802. As Intel’s expert explained, “one way to disable the variable-gain mixer in Der would be to turn off transis- tor switches M7-M10 to shut off the input signal into the gain control circuit, by setting Der’s VG+ and VG- to a low voltage (e.g. 0V).” Id. (capitalization normalized and em- phasis omitted). This explanation for the how isn’t re- motely conclusory. J.A. 1840 ¶ 70. Nor is Intel’s explanation for the why. For one thing, Intel’s expert pointed out that Der expressly highlights power consumption as a consideration. J.A. 1840 ¶ 70 (cit- ing J.A. 1178 (quantifying power consumption)). For an- other, Intel’s expert explained that the idea of “disabl[ing] the variable-gain mixer to save power” flows directly from the observation that the variable-gain mixer isn’t always needed. J.A. 1840 ¶ 70 (capitalization normalized). That, in turn, follows from Der’s statement that the “gain-control circuit . . . can add, subtract, or null signal currents from the variable-gain mixer.” J.A. 1840 ¶ 70 (quoting J.A. 1182 (emphasis added)). In Intel’s words, “the gain-control cir- cuit will block all, some, or none of the output signal cur- rent from the variable-gain mixer, whichever is appropriate to achieve the desired level of gain.” Appel- lant’s Br. 10. Consistent with that observation, Qual- comm’s expert acknowledged at least one scenario in which Case: 20-1664 Document: 103 Page: 22 Filed: 12/28/2021 22 INTEL CORPORATION v. QUALCOMM INCORPORATED “the output of the variable gain mixer would not contribute anything to the output.” J.A. 1909–10. And if that weren’t enough, Der expressly contemplates the idea of turning off unused circuits. As Intel noted to the Board, “although Der does not expressly describe turning off the variable gain mixer, it does describe turning off the calibration circuit.” Final Written Decision, at *26. The facts of KSR are instructive here. The patent in KSR claimed a design for an adjustable pedal with an elec- tronic sensor mounted on its fixed pivot point. 550 U.S. at 425. This, the Court concluded, was an obvious improve- ment to a prior-art adjustable pedal (disclosed in the Asano reference) with a sensor (like those disclosed in other ref- erences). Id. at 425–26. Just as KSR examined whether a skilled artisan “would have seen a benefit to upgrading As- ano with a sensor,” Intel argued that a skilled artisan would have “seen a benefit” to upgrading Der with Bur- gener’s switch. Id. at 424. Intel explained that doing so would improve energy efficiency for situations in which the variable-gain mixer is not in use. Far from being conclu- sory, Intel “fit the teachings of multiple patents together like pieces of a puzzle.” Id. at 420. The Board’s critique of this rationale as impermissibly “generic” is not supported by substantial evidence. INTEL CORPORATION v. QUALCOMM INCORPORATED 23 isolation of the transistor switches M7-M10, even when those transistors are off, some of the local oscillator (‘LO’) signal (applied to the gates of transistors M3A-M6A) would still feed through and appear at the output of the variable- gain mixer, degrading the signal quality.” J.A. 1840 ¶ 70 (capitalization normalized). Using the prior-art switch of Burgener, with its “shunting transistor,” Intel’s expert ex- plained, would “allow for improved signal quality . . . by shunting any LO feedthrough current to ground.” J.A. 1840 ¶ 71. Intel’s expert opined, accordingly, that this combination “would have involved nothing more than use of a known technique (adding a shunt switch to an RF switch) to improve a similar device (e.g., the overall mixer of Der) in the same way (providing a shunt path to ground to improve isolation).” J.A. 1840 ¶ 72. Here again, KSR is our guide. In KSR, a key question was “where to attach the sensor” to Asano’s pedal—specif- ically, whether a skilled artisan “starting with Asano would have found it obvious to put the sensor on a fixed pivot point.” 550 U.S. at 424–25. One prior-art reference taught “putting the sensor on the pedal device.” Id. at 425. Another (Smith) taught putting it “not on the pedal’s foot- pad but instead on its support structure,” and a third (Rixon) noted that a “wire-chafing” problem arises when placing a sensor on a moving part of a pedal. Id. From this constellation of references, a skilled artisan “would know to place the sensor on a nonmoving part of the pedal struc- ture” and that “[t]he most obvious nonmoving point on the structure from which a sensor can easily detect the pedal’s position is a pivot point” like the one in Asano—arriving at the claimed invention. Id. Just as the known wire-chafing problem in KSR had an obvious solution (placing the sensor on a nonmoving point), so too here undisputed evidence shows that incorporating Burgener’s switch was an “obvious solution” to the “known [feedthrough] problem.” Id. at 420; see generally J.A. 1873 (Burgener discussing feedthrough problem and shunting Case: 20-1664 Document: 103 Page: 24 Filed: 12/28/2021 24 INTEL CORPORATION v. QUALCOMM INCORPORATED solution). There is therefore little difference between this circumstance and the one in KSR, where “the prior art was replete with patents indicating that a fixed pivot point was an ideal mount for a sensor.” 550 U.S. at 420. Intel’s ex- pert explained that combining Der with Burgener “results in the LO feedthrough being reduced,” and “predictable re- sults and benefits are obtained as described in Burgener.” J.A. 1840 ¶ 72 (emphasis omitted). To use KSR’s words, this is “the predictable use of prior art elements according to their established functions.” 550 U.S. at 417. The Board saw things differently. Noting that “Bur- gener discusses problems associated with [the] prior art de- signs” (namely “insertion loss, switch isolation, and switch compression”), and that for that reason Burgener “provides improvements,” the Board reasoned that although a skilled artisan “could have selected the less effective prior art RF switches over the improved switches that are the subject of Burgener, [Intel] has not sufficiently demonstrated” why such an artisan “would have selected using elements from the less effective prior art designs.” Final Written Decision, at *27. “Without such an explanation,” the Board said, “the only reasonable inference is that [Intel] focused on Bur- gener’s description of prior art switches using impermissi- ble hindsight.” Id. Not so. Our caselaw is clear. It’s not necessary to show that a combination is “the best option, only that it be a suitable option.” PAR Pharm., Inc. v. TWI Pharms., Inc., 773 F.3d 1186, 1197–98 (Fed. Cir. 2014). And “if a technique has been used to improve one device, and a person of ordinary skill in the art would recognize that it would improve similar devices in the same way, us- ing the technique is obvious unless its actual application is beyond his or her skill.” KSR, 550 U.S. at 417. That’s what we have here. Given the facts and reasoning of KSR, it is readily apparent that the Board’s hindsight inference lacks substantial evidence. Qualcomm describes Intel’s theory as fixing a “problem of their own making.” Appellee’s Br. 52. But the same can Case: 20-1664 Document: 103 Page: 25 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 25 be said of the known wire-chafing problem in KSR, which arose by virtue of combining an adjustable pedal with an electronic sensor. 550 U.S. at 425. As the KSR Court noted, “[i]n automotive design, as in many other fields, the interaction of multiple components means that changing one component often requires the others to be modified as well.” Id. at 424. So too here. Like modifying Asano in a way that avoided the wire-chafing problem in KSR, modi- fying Der with Burgener in a way that avoids the feed- through problem doesn’t neutralize Intel’s rationale. 26 INTEL CORPORATION v. QUALCOMM INCORPORATED 550 U.S. at 420. Indeed, one of the points on which the Court in KSR disagreed with our decision below in that case was our reliance on the primary purposes of prior-art references. We had pointed out that Asano “was designed to solve the constant ratio problem,” that Rixon “suffered from the problem of wire chafing but was not designed to solve it,” and that Smith “did not relate to adjustable ped- als and did not necessarily go to the issue of motivation to attach the electronic control on the support bracket of the pedal assembly.” Id. at 414 (internal quotation marks omitted). That was misguided. “Regardless of Asano’s pri- mary purpose, the design provided an obvious example of an adjustable pedal with a fixed pivot point” such that “[t]he idea that a designer hoping to make an adjustable electronic pedal would ignore Asano because Asano was de- signed to solve the constant ratio problem makes little sense.” Id. at 420–21. Finally, the Board didn’t explain how disabling Der’s variable-gain mixer would hinder its purpose in use cases where “achieving the ‘desired output’ requires that the output of the variable-gain mixer be ‘null[ed]’—a possibility that Der itself contemplates.” Ap- pellant’s Br. 45–46 (quoting J.A. 1182) (alteration in origi- nal). Accordingly, the Board’s analysis lacks substantial ev- idence under KSR. We vacate the Board’s decision as to substitute claims 27, 28, and 31 and remand with instruc- tions that Intel demonstrated sufficient reason for combin- ing Der with Burgener. The Board should resolve any remaining disputes regarding that combination on re- mand. E.g., Appellee’s Br. 52 n.11. CONCLUSION We have considered the parties’ remaining arguments but find them unpersuasive. We affirm the Board’s claim construction and decision as to claims 6, 17–19, and 21. We vacate the Board’s determination as to substitute claims 27, 28, and 31 and remand for further proceedings. Case: 20-1664 Document: 103 Page: 27 Filed: 12/28/2021 INTEL CORPORATION v. QUALCOMM INCORPORATED 27 AFFIRMED-IN-PART, VACATED-IN-PART, AND REMANDED COSTS The parties shall bear their own costs.