332.
     Claim terms “are generally given their ordinary and
 customary meaning,” which is “the meaning that the term
 would have to a person of ordinary skill in the art in ques-
 tion at the time of the invention.” Phillips v. AWH Corp.,
Case: 18-2338    Document: 73      Page: 10    Filed: 09/24/2020




 10       NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD
                                                 COMPANY


 
415 F.3d 1303
, 1312–13 (Fed. Cir. 2005) (en banc) (internal
 quotation marks and citations omitted). “[T]he claims
 themselves provide substantial guidance as to the meaning
 of particular claim terms” and “the specification is the sin-
 gle best guide to the meaning of a disputed term.”
 Id.
 at 
1314–15 (internal quotation marks and citations omit-
 ted).
     As the party seeking to disturb the non-infringement
 judgment, Network-1 “must establish that [the challenged
 jury] instructions were legally erroneous, and that the er-
 rors had prejudicial effect.” SSL Servs., LLC v. Citrix Sys.,
 
769 F.3d 1073
, 1085 (Fed. Cir. 2014) (quoting Ecolab Inc.
 v. Paraclipse, Inc., 
285 F.3d 1362
, 1373 (Fed. Cir. 2002)).
     “It is well established that when an incorrect jury in-
 struction—such as an incorrect claim construction—re-
 moves from the jury a basis on which the jury could
 reasonably have reached a different verdict, the verdict
 should not stand.” Cardiac Pacemakers, Inc. v. St. Jude
 Med., Inc., 
381 F.3d 1371
, 1383 (Fed. Cir. 2004); see also
 Avid Tech., Inc. v. Harmonic, Inc., 
812 F.3d 1040
, 1042
 (Fed. Cir. 2016). An erroneous claim construction on one
 element is harmless “only if a reasonable jury would have
 been required by the evidence to find non-infringement
 even without the error.” 
Avid, 812 F.3d at 1047
.
                               B
      The district court construed “low level current” as “a
 non-data-signal current that is sufficient to begin start up
 of the access device but that is not sufficient to sustain the
 start up.” J.A. 35; see also J.A. 53–55. The district court
 explained that it was construing the disputed phrase to
 give meaning to the constituent term “low.” See J.A. 34. It
 stated that while the current must be sufficient to “begin
 start up,” the current need not “cause start up,” and thus
 it construed the term to eliminate “any implication that the
 current must be sufficient to result in a completed start-
 up.” J.A. 34 (citing ’930 patent col. 3 ll. 12–17).
Case: 18-2338     Document: 73       Page: 11     Filed: 09/24/2020




 NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD                11
 COMPANY


     On appeal, Network-1 argues that the district court
 erred by construing the phrase “low level current” to have
 a lower bound. Network-1 agrees that, in the context of the
 ’930 patent, the phrase “low level current” describes cur-
 rent that cannot “sustain start up” but argues that it was
 error for the district court to additionally require that the
 current be sufficient to begin start up. See Appellant’s
 Br. 27. Relying on extrinsic evidence related to the word
 “low,” including dictionary definitions, grammar refer-
 ences, and claim constructions adopted for unrelated pa-
 tents, Network-1 argues that the term points in one
 direction and should be construed to have a single refer-
 ence point. Network-1 asserts that the correct reference
 point with respect to the ’930 patent is the upper boundary
 requiring that the current not exceed the level needed to
 operate the device. We disagree.
     We conclude that the district court correctly construed
 “low level current.” As an initial matter, there is no dispute
 that the word “low” in the claim phrase “low level current”
 operates to limit the upper boundary of the current level.
 See Appellant’s Br. 26–29; Appellee’s Br. 36–37. Indeed,
 the ’930 patent explicitly describes a low level current with
 an upper boundary that is “unable to sustain start up.” See
 ’930 patent col. 3 ll. 14–16; see also
 id. at 
col. 3 ll. 2–13. But
 the claim phrase is not limited to the word “low,” and the
 claim construction analysis should not end just because one
 reference point has been identified. To be sure, the claim
 phrase “low level current” does not preclude a lower bound
 by use of the word “low.” Rather, in the same way the
 phrase should be construed to give meaning to the term
 “low,” the phrase must also be construed to give meaning
 to the term “current.”
     The intrinsic record of the ’930 patent confirms that the
 district court correctly construed the phrase to require a
 lower boundary of current. See 
Phillips, 415 F.3d at 1315
.
 Claim 6 recites “delivering a low level current” to an access
 device to detect whether the device can accept remote
Case: 18-2338     Document: 73      Page: 12     Filed: 09/24/2020




 12       NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD
                                                 COMPANY


 power. ’930 patent col. 4 ll. 49–67 (emphasis added); see
 also
 id. at 
col. 2 ll. 8–14. The express language of the claim
 thus requires that at least some level of current be deliv-
 ered. Even Network-1 admits that the term “current” nec-
 essarily requires some flow of electric charge because “[i]f
 there is no flow, there is no ‘current.’” See Appellant’s Br.
 33–34.
      The specification of the ’930 patent discloses the lower
 boundary of current. It explains that an access device ca-
 pable of accepting remote power will signal a varying volt-
 age level and teaches “[t]he varying level is created by the
 remote power supply beginning to start up but the low cur-
 rent level is unable to sustain the start up.” ’930 patent
 col. 3 ll. 14–16 (emphasis added); see also
 id. at 
col. 3 ll. 2–
 13. Thus, consistent with the district court’s construction,
 the ’930 patent teaches that the delivered “low level cur-
 rent” is current that is sufficient to begin start-up.
     Accordingly, we conclude that the district court did not
 err in its construction of “low level current.”
                                C
      The district court construed “main power source” as “a
 DC power source,” and thereby excluded AC power sources
 from its construction. J.A. 50–51. 7 In reaching its decision,
 the court relied on HP’s expert testimony that a network
 device would either fail to function or would be damaged if
 it received AC power. J.A. 50–51. The court therefore con-
 cluded that embodiments using AC power are inoperable
 and construed “main power source” to exclude such embod-
 iments. J.A. 50–51 (“A construction that renders the
 claimed invention inoperable should be viewed with ex-
 treme skepticism.” (quoting Talbert Fuel Sys. Pats. Co. v.



      7  “DC” refers to direct current and “AC” refers to al-
 ternating current.
Case: 18-2338    Document: 73      Page: 13     Filed: 09/24/2020




 NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD             13
 COMPANY


 Unocal Corp., 
275 F.3d 1371
, 1376 (Fed. Cir. 2002))); see
 also J.A. 30–31.
     On appeal, Network-1 argues that the claim phrase
 “main power source” should be construed consistent with
 the ordinary meaning of “power source” to include both AC
 and DC power sources. See Appellant’s Br. 46. Network-1
 acknowledges that certain embodiments with AC power
 sources require power from the main power source to be
 converted to DC power before being received by the net-
 work device, but Network-1 argues that nothing in the
 ’930 patent precludes such embodiments. To the contrary,
 Network-1 asserts that the intrinsic record supports a con-
 struction that includes both types of power sources. We
 agree and reverse the district court’s construction of main
 power source.
     We conclude that the correct construction of “main
 power source” includes both AC and DC power sources.
 There is no dispute that the ordinary meaning of “power
 source” includes both AC and DC power sources. Appel-
 lant’s Br. 45–49; see Appellee’s Br. 47. And neither the
 claims nor the specification of the ’930 patent require a de-
 parture from this ordinary meaning. Starhome GmbH v.
 AT & T Mobility LLC, 
743 F.3d 849
, 857 (Fed. Cir. 2014)
 (“In the absence of an express intent to impart a novel
 meaning to claim terms, an inventor’s claim terms take on
 their ordinary meaning.”).
      Claim 6 recites a “main power source” that is “con-
 nected to supply power to the data node” and that “de-
 liver[s] a low level current . . . to the access device.”
 ’930 patent col. 4 ll. 56–57, ll. 60–61; see also
 id. at 
col. 1
 ll. 64–65, col. 2 ll. 8–9. Nowhere do the claims suggest that
 the “main power source” should be limited to DC power
 sources.
     The specification likewise never expresses a preference
 for DC power sources, much less a suggestion that DC
 power is an “essential” feature of the main power source.
Case: 18-2338     Document: 73      Page: 14    Filed: 09/24/2020




 14       NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD
                                                 COMPANY


 See GE Lighting Sols., LLC v. AgiLight, Inc., 
750 F.3d
 1304
, 1309–10 (Fed. Cir. 2014). Indeed, far from limiting
 the plain language of the claims to a DC power source, the
 specification states that the power source may be a “con-
 ventional main power supply,” which according to the ordi-
 nary meaning would include both AC and DC power
 sources. ’930 patent col. 2 ll. 52–53. This reading of the
 specification is further confirmed by the ’930 patent’s dis-
 closure of a preferred embodiment, which is shown in Fig-
 ures 1 and 3. See MBO Labs., Inc. v. Becton, Dickinson &
 Co., 
474 F.3d 1323
, 1333 (Fed. Cir. 2007) (“A claim inter-
 pretation that excludes a preferred embodiment from the
 scope of the claim is rarely, if ever, correct.” (internal cita-
 tion omitted)). In Figure 1, power source 16 is identified as
 a “conventional” power supply, and in Figure 3, main power
 supply 70 is specifically depicted as a traditional AC power
 outlet. ’930 patent col. 2 ll. 52–53, Fig. 3.
     Despite the express disclosures of the ’930 patent, the
 district court construed “main power source” to exclude AC
 power sources on the basis of expert testimony that receipt
 of AC power by a network device would render it inopera-
 ble. J.A. 30–31, 50–51. This is error for two reasons. First,
 even though the network device cannot receive AC power,
 the record establishes that “data nodes” or network
 switches were commonly used to convert AC power to DC
 power as needed to power the network device. Indeed, we
 need not look beyond the specification of the ’930 patent.
 In the preferred embodiment, as shown in Figure 1, a data
 node exists between main power source 16 and the remote
 equipment. And in Figure 3, an 8-port switch is shown con-
 nected to the AC power outlet such that the AC power is
 converted to DC power before reaching the network device.
 ’930 patent col. 2 l. 29, col. 3 ll. 59–60, Figs. 1 & 3. Even
 HP’s expert Dr. Neikirk explained that Figure 3 illustrates
 “‘the main power supply 70’ receiving power from a typical
 AC wall socket, and in turn suppl[ying] operational power
 via the ‘Main Power Distribution Bus’ to the data
Case: 18-2338    Document: 73      Page: 15    Filed: 09/24/2020




 NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD            15
 COMPANY


 switches.” J.A. 1049 ¶ 35. Nothing in the claims of the
 ’930 patent precludes the conversion of AC power to DC
 power, and it was error for the district court to add such a
 limitation.
      Second, the district court erred by adding a limitation
 to the claims to carve out certain inoperable embodiments.
 To be sure, we have stated that we skeptically view a con-
 struction that renders all embodiments inoperable, but we
 have instructed that it is nonetheless improper to add lim-
 itations to constructions to exclude only certain inoperable
 embodiments. Cordis Corp. v. Medtronic AVE, Inc.,
 
511 F.3d 1157
, 1174 (Fed. Cir. 2008). In this case, the dis-
 trict court’s construction not only excluded inoperable em-
 bodiments that do not convert AC to DC power, but also
 excluded operable embodiments, like the preferred embod-
 iment. See MBO 
Labs., 474 F.3d at 1333
.
      Accordingly, we conclude that the district court erred
 in its construction of “main power source.”
                               D
     We now consider the effect of the district court’s erro-
 neous construction of “main power source” on the jury’s ver-
 dict of non-infringement. Network-1 argues that if the
 district court erred in its construction of either claim term,
 that error was prejudicial, and it is entitled to a new trial
 on infringement. HP argues in response that to the extent
 either term was misconstrued, such error was harmless
 and the judgment of non-infringement should not be dis-
 turbed. We conclude that Network-1 established that it
 was prejudiced by the district court’s erroneous claim con-
 struction of “main power source” and is entitled to a new
 trial on infringement.
     At trial, HP contested infringement based on the ele-
 ments “low level current” and “main power source.” With
 respect to “main power source,” the record shows that the
 jury was instructed to apply the erroneous construction,
Case: 18-2338    Document: 73      Page: 16    Filed: 09/24/2020




 16       NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD
                                                 COMPANY


 that Network-1 elicited testimony showing that the main
 power source for certain HP products originates at an AC
 input, J.A. 2291, 2056, and that HP relied on evidence of
 non-infringement, at least in part, based on the fact that
 these products are not within the scope of the ’930 patent
 because they use an AC input, J.A. 2056–57, 2065–66. The
 jury found that HP did not infringe the asserted claims of
 the ’930 patent. Because the general verdict form does not
 indicate the basis for the jury’s decision, where the evi-
 dence shows that HP relied on the district court’s erroneous
 construction for its argument that it does not infringe the
 asserted claims, we conclude that Network-1 has estab-
 lished that it was prejudiced by the claim construction.
 J.A. 69–70.
      HP argues that the erroneous claim construction was
 harmless because HP presented conclusive evidence that
 no accused product meets the claim limitation “delivering
 a low level current from said main power source.” Appel-
 lee’s Br. 25–26 (emphases in original). More particularly,
 HP argues that the record shows that its products deliver
 detection current from a PoE chip, not an AC or DC power
 source. But HP’s argument relies on an element that is not
 in the claims—namely, that current be delivered directly
 from the main power source to the access device. Indeed,
 the district court expressly rejected HP’s attempt to import
 such a limitation during claim construction. See J.A 39–40.
 And in any case, Network-1 presented evidence at trial
 showing that HP’s products practice this claim limitation.
 See, e.g., J.A. 1790–95, 2480–81, 2724. We therefore reject
 HP’s argument that the non-infringement verdict may be
 affirmed on appeal.
      Accordingly, we affirm the district court’s construction
 of “low level current,” reverse the district court’s construc-
 tion of “main power source,” and remand to the district
 court for proceedings consistent with this opinion.
Case: 18-2338    Document: 73     Page: 17    Filed: 09/24/2020




 NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD           17
 COMPANY


                   II. Statutory Estoppel
     We now turn to HP’s cross-appeal. HP argues that the
 district court erroneously granted JMOL with respect to
 the ’930 patent’s validity based on its determination that
 HP was estopped under 35 U.S.C. § 315(e) from presenting
 obviousness challenges as a consequence of its joinder to
 the Avaya IPR. We agree that HP is not estopped. We
 therefore vacate the district court’s judgment and remand
 to the district court for proceedings consistent with this
 opinion.
                              A
     We review “decisions on motions for JMOL, motions for
 a new trial, and evidentiary rulings under the law of the
 regional circuit.” SSL 
Servs., 769 F.3d at 1082
. The Fifth
 Circuit reviews decisions on motions for JMOL “de novo,
 reapplying the JMOL standard.” Summit 6, LLC v. Sam-
 sung Elecs. Co., 
802 F.3d 1283
, 1293 (Fed. Cir. 2015) (citing
 Ford v. Cimarron Ins. Co., 
230 F.3d 828
, 830 (5th Cir.
 2000)). In the Fifth Circuit, a jury verdict may only be re-
 versed by JMOL if substantial evidence does not support
 the verdict. Versata Software, Inc. v. SAP Am., Inc.,
 
717 F.3d 1255
, 1261 (Fed. Cir. 2013). “Thus, a JMOL may
 only be granted when, ‘viewing the evidence in the light
 most favorable to the verdict, the evidence points so
 strongly and overwhelmingly in favor of one party that the
 court believes that reasonable jurors could not arrive at
 any contrary conclusion.’”
 Id. (quoting Dresser–Rand Co.
 v. Virtual Automation, Inc., 
361 F.3d 831
, 838 (5th Cir.
 2004)).
                              B
     Following trial, the district court granted Network-1’s
 motion for JMOL on validity of the ’930 patent. The district
 court concluded that as a result of HP’s joinder to the
 Avaya IPR, HP was estopped under 35 U.S.C. § 315(e) from
 raising obviousness challenges not based on the Fisher
Case: 18-2338    Document: 73     Page: 18    Filed: 09/24/2020




 18       NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD
                                                 COMPANY


 system. J.A. 91. According to the district court, in contrast
 to the Fisher system, which was not a patent or printed
 publication that HP “reasonably could have raised” in the
 IPR, HP could have reasonably raised its remaining inva-
 lidity arguments during the IPR—i.e., the Fisher patents,
 Woodmas, and Chang. J.A. 88–89, 91 (quoting 35 U.S.C.
 § 315(e)(2)). In reaching this conclusion, the district court
 specifically rejected HP’s argument that it could not have
 raised new grounds in the Avaya IPR because it was a
 joined party. The district court stated that “the fact that
 HP sought joinder with Avaya’s IPR does not mean that HP
 could not have reasonably raised different grounds from
 those raised by Avaya.” J.A. 91. The district court further
 stated that allowing HP to raise arguments “that it elected
 not to raise during the IPR would give it a second bite at
 the apple and allow it to reap the benefits of the IPR with-
 out the downside of meaningful estoppel.” J.A. 91 (internal
 quotations omitted). The district court therefore granted
 JMOL on validity without considering the merits of HP’s
 invalidity arguments. J.A. 91 & n.6.
     HP argues that, in granting Network-1’s motion for
 JMOL on invalidity, the district court misapplied the es-
 toppel provision under 35 U.S.C. § 315(e)(2). Specifically,
 HP argues that no validity ground that it raised at trial
 “reasonably could have [been] raised” through its joinder to
 the Avaya IPR. See J.A. 88–91. We agree with HP. HP’s
 joinder to the Avaya IPR and the estoppel consequences of
 that joinder are governed by the America Invents Act
 (“AIA”), which established IPR proceedings. According to
 the AIA, under 35 U.S.C. § 315(c), HP was permitted to join
 the Avaya IPR “as a party” even though HP was time-
 barred under § 315(b) from bringing its own petition. But,
 as we held in Facebook, Inc. v. Windy City Innovations,
 LLC, the joinder provision does not permit a joining party
 to bring into the proceeding new grounds that were not al-
 ready instituted. Facebook, Inc. v. Windy City Innovations,
 LLC, __ F.3d __, No. 18-1400, 
2020 WL 5267975
, at *9–10
Case: 18-2338    Document: 73     Page: 19    Filed: 09/24/2020




 NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD             19
 COMPANY


 (Fed. Cir. Sept. 4, 2020). Rather, it may only join the al-
 ready-instituted proceeding as a party.
 Id.
     Following a final 
written decision in an IPR, the AIA
 provides for statutory estoppel under 35 U.S.C. § 315(e) to
 limit the invalidity challenges that an IPR petitioner may
 bring in a separate action involving the same patent
 claims. With respect to district court actions, § 315(e)(2)
 states:
     CIVIL ACTIONS AND OTHER PROCEEDINGS.—The peti-
     tioner in an inter partes review of a claim in a pa-
     tent under this chapter that results in a final
     written decision under section 318(a) . . . may not
     assert in . . . a civil action arising in whole or in
     part under section 1338 of title 28 . . . that the
     claim is invalid on any ground that the petitioner
     raised or reasonably could have raised during that
     inter partes review.
 35 U.S.C. § 315(e)(2) (emphases added). Thus, according to
 the statute, a party is only estopped from challenging
 claims in the final written decision based on grounds that
 it “raised or reasonably could have raised” during the IPR.
 Because a joining party cannot bring with it grounds other
 than those already instituted, that party is not statutorily
 estopped from raising other invalidity grounds.
     In this case, the Board instituted two grounds in the
 Avaya IPR, which challenged claims 6 and 9 of the ’930 pa-
 tent based on Matsuno and De Nicolo. HP did not timely
 petition for IPR but relied on the joinder exception to the
 time bar under § 315(b). HP first filed a motion to join the
 Avaya IPR with a petition requesting review based on
 grounds not already instituted. The Board correctly denied
 HP’s request. The Board, however, granted HP’s second
 joinder request, which petitioned for only the two grounds
 already instituted. When the Board reached a final written
 decision in the Avaya IPR, because HP was a petitioner in
 that proceeding, HP was statutorily estopped from raising
Case: 18-2338    Document: 73      Page: 20    Filed: 09/24/2020




 20       NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD
                                                 COMPANY


 invalidity grounds based on Matsuno and De Nicolo
 against claims 6 and 9 in a district court action. HP, how-
 ever, was not estopped from raising other invalidity chal-
 lenges against those claims because, as a joining party, HP
 could not have raised with its joinder any additional inva-
 lidity challenges. Thus, contrary to the district court’s sug-
 gestion, permitting HP to challenge the asserted claims of
 the ’930 patent as obvious over the Fisher patents, Wood-
 mas, and Chang does not give HP a “second bite at the ap-
 ple” to challenge the ’930 patent, J.A. 91, because HP could
 not have raised such a challenge in the Avaya IPR.
     Accordingly, we conclude that HP was not statutorily
 estopped under § 315(e) from challenging the asserted
 claims of the ’930 patent based on the Fisher patents,
 Woodmas, and Chang, which were not raised in the Avaya
 IPR and which could not have reasonably been raised by
 HP. We therefore vacate the district court’s JMOL decision
 on validity with respect to estoppel.


                               C
     Because we conclude that the district court erred in
 granting JMOL of validity for Network-1 on the basis that
 HP was statutorily estopped from raising certain invalidity
 challenges, we vacate that decision.
     HP argues that if we vacate the district court’s JMOL
 on validity, substantial evidence supports the jury’s verdict
 of invalidity, and thus the jury’s verdict should be rein-
 stated. In response, Network-1 argues that even if we re-
 verse on estoppel, we cannot resolve this case on appeal
 because there is an outstanding new-trial motion that the
 district court must decide. Indeed, the district court failed
 to conditionally rule on Network-1’s motion for a new trial
 on validity even though it was required to do so by Federal
 Rule of Civil Procedure 50(c). See J.A. 78–91, 9260–65. We
 decline to consider Network-1’s motion in the first instance.
Case: 18-2338    Document: 73      Page: 21    Filed: 09/24/2020




 NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD            21
 COMPANY


 We therefore remand to the district court for further pro-
 ceedings consistent with this opinion.
                    III. Claim Broadening
     On cross-appeal, HP also argues that claim 6 and the
 other asserted claims are invalid under 35 U.S.C. § 305 be-
 cause Network-1 improperly broadened claim 6 through
 the addition of claim 15 and 16 in the ’401 reexamination.
 We disagree and affirm the district court’s judgment that
 the asserted claims were not improperly broadened.
                               A
      A patentee is not permitted to enlarge the scope of a
 patent claim during reexamination. 35 U.S.C. § 305. The
 broadening inquiry under § 305 involves two steps: (1) “an-
 alyz[ing] the scope of the claim prior to reexamination” and
 (2) “compar[ing] it with the scope of the claim subsequent
 to reexamination.” Creo Prods., Inc. v. Presstek, Inc., 
305
 F.3d 1337
, 1344 (Fed. Cir. 2002). A claim “is broader in
 scope than the original claims if it contains within its scope
 any conceivable apparatus or process which would not have
 infringed the original patent.” See Predicate Logic, Inc. v.
 Distributive Software, Inc., 
544 F.3d 1298
, 1303 (Fed. Cir.
 2008).
     “Whether amendments made during reexamination
 enlarge the scope of a claim is a matter of claim construc-
 tion,” Creo 
Prods., 305 F.3d at 1344
, which we review de
 novo, while giving deference to subsidiary factual determi-
 nations, Teva 
Pharms., 574 U.S. at 331
.
                               B
     HP argues that dependent claims 15 and 16 added dur-
 ing the ’401 reexamination resulted in improper claim
 broadening of claim 6 and asserted dependent claims. In
 relevant part, prior to reexamination, claim 6 of the
 ’930 patent was construed in two separate district court ac-
 tions to require the “secondary power source” to be
Case: 18-2338    Document: 73     Page: 22    Filed: 09/24/2020




 22       NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD
                                                 COMPANY


 physically separate from the “main power source.” See J.A.
 59–62; see also J.A. 40–42. Subsequently, during the ’401
 reexamination, Network-1 added claims 15 and 16, which
 depended from claim 6 and respectively added the limita-
 tions that the secondary power source “is the same source
 of power” and “is the same physical device” as the main
 power source. ’930 patent, Ex Parte Reexamination Certif-
 icate, col. 1 ll. 39–44.
     After claims 15 and 16 issued at the conclusion of the
 ’401 reexamination, HP moved in the underlying district
 court action for summary judgment of invalidity under
 35 U.S.C. § 305 for improper claim broadening. Network-1
 subsequently filed a statutory disclaimer under 35 U.S.C.
 § 253 of claims 15 and 16. See J.A. 5075. Ultimately, the
 district court denied HP’s motion, finding that claim 6 had
 not been improperly broadened. J.A. 59–62; see also
 J.A. 40–42. In the same order, the district court also con-
 strued the claim term “secondary power source” consistent
 with the earlier district court actions to require that the
 “secondary power source be physically separate from the
 driving points of the main power source.” J.A. 52–53; see
 also J.A. 32–33.
      We do not agree that claim 6 is invalid for improper
 broadening based on the addition of claims 15 and 16. Our
 broadening inquiry begins and ends with claim 6. Claim 6
 was not itself amended during the ’401 reexamination.
 And as HP admits, “[t]he district court’s construction [after
 reexamination] is consistent with how the [earlier district]
 courts construed this term [pre-reexamination].” Appel-
 lee’s Br. 15. Neither party appeals that construction.
 There can be no dispute, therefore, that the scope of claim 6
 was not changed as a result of the ’401 reexamination.
 Where the scope of claim 6 has not changed, there has not
 been improper claim broadening, and HP’s argument fails.
 See Creo 
Prods., 305 F.3d at 1344
.
Case: 18-2338    Document: 73     Page: 23    Filed: 09/24/2020




 NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD           23
 COMPANY


     Furthermore, our precedent is clear that “dependent
 claims cannot broaden an independent claim from which
 they depend.” Enzo Biochem Inc. v. Applera Corp., 
780
 F.3d 1149
, 1156–57 (Fed. Cir. 2017). Thus, even were de-
 pendent claims 15 and 16 broader than unamended, inde-
 pendent claim 6, the remedy would not be to find claim 6
 invalid as broadened, but to invalidate added claims 15 and
 16. See MBO Labs., Inc. v. Becton, Dickinson & Co., 
602
 F.3d 1306
, 1319 (Fed. Cir. 2010). We need not determine
 whether claims 15 and 16 are invalid because they are not
 asserted and because those claims have already been can-
 celed through Network-1’s statutory disclaimer.
     Despite the clarity of our caselaw, HP principally relies
 on ArcelorMittal France v. AK Steel Corp., 
786 F.3d 885
 (Fed. Cir. 2015), to argue that claim 6 was improperly
 broadened and should be invalidated. In that case, we con-
 cluded that the patentee had improperly broadened inde-
 pendent claim 1 through reissue by adding a number of
 dependent claims. See 
ArcelorMittal, 786 F.3d at 890
. HP
 quotes our explanation that one of the new dependent
 claims had “the practical effect of expanding the scope of
 claim 1 to cover claim scope expressly rejected by a previ-
 ous claim construction ruling.” Appellee’s Br. 70 (quoting
 
ArcelorMittal, 786 F.3d at 890
). HP also emphasizes that
 in ArcelorMittal, we held invalid for improper broadening
 under 35 U.S.C. § 251 not only the newly added claims but
 also claim 1 itself. Appellee’s Br. 70. HP argues that the
 facts in ArcelorMittal are “nearly identical” to this case,
 and that claim 6 is invalid because claims 15 and 16 have
 “the practical effect of expanding” the scope of claim 6.
 Id.
 HP is wrong.
     ArcelorMittal is inapposite. In that case, the patentee
 had stipulated that all reissued claims, including claim 1,
 were broader than the original claims. 
ArcelorMittal, 786
 F.3d at 890
. Thus, in ArcelorMittal, there was no dispute
 that the claims had been broadened. Furthermore, we did
 not hold, as HP suggests, see Appellee’s Br. 70–71, that a
Case: 18-2338     Document: 73    Page: 24      Filed: 09/24/2020




 24        NETWORK-1 TECHNOLOGIES, INC. v. HEWLETT-PACKARD
                                                  COMPANY


 dependent claim added during reissue (or reexamination)
 may broaden and therefore invalidate an unamended, in-
 dependent claim. To the contrary, we rejected “the argu-
 ment that a defective reissue application invalidates . . .
 [the] original claims carried over from the original applica-
 tion.” 
ArcelorMittal, 786 F.3d at 891 
(quoting Hewlett–
 Packard Co. v. Bausch & Lomb, Inc., 
882 F.2d 1556
, 1566
 (Fed. Cir. 1989)).
     Accordingly, we affirm the district court’s conclusion
 that claim 6 and the other asserted claims are not invalid
 due to improper claim broadening.
                         CONCLUSION
    We have considered the parties’ remaining arguments
 and find them unpersuasive. For the reasons described
 above, we affirm-in-part, reverse-in-part, vacate, and re-
 mand.
      AFFIRMED-IN-PART, REVERSED-IN-PART,
           VACATED, AND REMANDED
                            COSTS
      The parties shall bear their own costs.