Opinion by Judge GRABER; Concurrence by Judge WATFORD.
OPINION
GRABER, Circuit Judge:
The State of California authorizes the execution of a capital prisoner only after affording a full opportunity to seek review in state and federal courts. Judicial review ensures that executions meet constitutional requirements, but it also takes time — too much time, in Petitioner Ernest DeWayne Jones’ view. He argues that California’s post-conviction system of judicial review creates such a long period of delay between sentencing and execution that only an “arbitrary” few prisoners actually are executed, in violation of the Eighth Amendment’s prohibition against cruel and unusual punishment. Under Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989), federal courts may not consider novel constitutional theories on habeas review. That principle “serves to ensure that gradual developments in the law over which reasonable jurists may disagree are not later used to upset the finality of state convictions valid when entered.” Sawyer v. Smith, 497 U.S. 227, 234, 110 S.Ct. 2822, 111 L.Ed.2d 193 (1990). Because we conclude that Petitioner’s claim asks us to apply a novel constitutional rule, we must deny the claim as barred by Teague. Accordingly, we reverse the district court’s judgment granting relief.
FACTUAL AND PROCEDURAL HISTORY
In 1995, a jury sentenced Petitioner to death for the rape and murder of his girlfriend’s mother. The California Supreme Court affirmed the judgment in 2003, People v. Jones, 29 Cal.4th 1229, 131 Cal. Rptr.2d 468, 64 P.3d 762 (2003), and the United States Supreme Court denied cer-tiorari that same year, Jones v. California, 540 U.S. 952, 124 S.Ct. 395, 157 L.Ed.2d 286 (2003). The California Supreme Court denied Petitioner’s state habeas petition in 2009.
On direct appeal to the California Supreme Court, Petitioner presented what is commonly known as a “Lackey claim,” so named after a memorandum by Justice Stevens respecting the denial of certiorari in Lackey v. Texas, 514 U.S. 1045, 115 S.Ct. 1421, 131 L.Ed.2d 304 (1997) (mem.). Petitioner argued that the delay between imposition of sentence in 1995 and eventual execution inevitably would be so long that carrying out the sentence would vio[542]*542late the Eighth Amendment’s prohibition against cruel and unusual punishment. Relying on its precedent, the California Supreme Court rejected Petitioner’s Lackey claim. Jones, 131 Cal.Rptr.2d 468, 64 P.3d at 787; see People v. Anderson, 25 Cal.4th 543, 106 Cal.Rptr.2d 575, 22 P.3d 347, 389 (2001) (“[W]e have consistently concluded, both before and since Lackey, that delay inherent in the automatic appeal process is not a basis for concluding that either the death penalty itself, or the process leading to its execution, is cruel and unusual punishment.”).
Petitioner filed a federal habeas petition in 2010. In claim 27, Petitioner asserted the same Lackey claim that the state court had rejected, arguing that the “excessive delay” after his sentencing violates the Eighth Amendment. In 2014, the district court issued an order expressing the view that California’s post-conviction system itself may be unconstitutional. Four days later, the district court directed Petitioner to file an amended petition raising the systemic challenge and required the parties to address “petitioner’s new claim” in supplemental briefs. Consistent with the court’s order, Petitioner filed an amended federal habeas petition. In amended claim 27, Petitioner alleged that California’s post-conviction system itself violates the Eighth Amendment by creating excessive delay between sentencing and execution in capital cases generally.
After receiving briefs and holding a hearing, the district court granted relief to Petitioner on the amended claim, holding that California’s post-conviction system for capital prisoners violates the Eighth Amendment’s prohibition against cruel and unusual punishment. Jones v. Chappell, 31 F.Supp.3d 1050 (C.D.Cal.2014) (order). Although more than 900 people have been sentenced to death in California since 1978, only 13 have been executed. Id. at 1053. As of 2014, some Death-Row inmates had died of natural causes, the sentences of some had been vacated, and 748 remained on Death Row. Id. For those who are eventually executed, “the process will likely take 25 years or more.” • Id. at 1054. “[D]elay is evident at each stage of the post-conviction review process,” id. at 1056, including on direct appeal, state collateral review, and federal collateral review, id. at 1056-60. In the district court’s view, “much of the delay in California’s post[-]conviction review process is created by the State itself.” Id. at 1066.
Relying primarily on the Supreme Court’s decision in Furman v. Georgia, 408 U.S. 238, 92 S.Ct. 2726, 33 L.Ed.2d 346 (1972) (per curiam), the district court held that the “systemic delay and dysfunction” in California’s post-conviction review process was unconstitutionally “arbitrary,” because a capital prisoner’s selection for execution “will depend upon a factor largely outside an inmate’s control, and wholly divorced from the penological purposes the State sought to achieve by sentencing him to death in the first instance: how quickly the inmate proceeds through the State’s dysfunctional post-conviction review process.” Jones, 31 F.Supp.3d at 1061-63. The court concluded that, “where the State permits the post-conviction review process to become so inordinately and unnecessarily delayed that only an arbitrarily selected few of those sentenced to death are executed, the State’s process violates the Eighth Amendment. Fundamental principles of due process and just punishment demand that any punishment, let alone the ultimate one of execution, be timely and rationally carried out.” Id. at 1067.
The district court also held that the deferential standards of review mandated by the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”), found in 28 U.S.C. § 2254(d), did not apply because [543]*543the state courts had never ruled on the systemic claim. Jones, 31 F.Supp.3d at 1067-68, 1068 n. 23. The court acknowledged that petitioners ordinarily must exhaust their claims under 28 U.S.C. § 2254(b)(1)(A), but held that Petitioner was excused from the exhaustion requirement because “circumstances exist that render [the State’s corrective] process ineffective to protect the rights of the applicant,” id. § 2254(b)(1)(B)(ii). Jones, 31 F.Supp.3d at 1067-68. In particular, “Requiring [Petitioner] to return to the California Supreme Court to exhaust his claim would only compound the delay that has already plagued his post-conviction review process.” Id. at 1068.
The court next held that Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989), which generally prohibits federal courts from announcing a new rule of constitutional law in a habeas case, did not bar Petitioner’s claim. Jones, 31 F.Supp.3d at 1068. “The rule [Petitioner] seeks to have applied here— that a state may not arbitrarily inflict the death penalty — is not new.” Id. That rule is “inherent in the most basic notions of due process and fair punishment embedded in the core of the Eighth Amendment.” Id.
The district court vacated Petitioner’s capital sentence. Id. at 1069. The court later entered partial final judgment under Federal Rule of Civil Procedure 54(b), determining that there was no just reason for delay in the entry of judgment on Petitioner’s amended claim 27. Respondent Warden Ron Davis (“the State”) timely appeals.
STANDARD OF REVIEW
‘We review the district court’s decision to grant or deny a petition for habeas corpus de novo.” Leavitt v. Arave, 383 F.3d 809, 815 (9th Cir.2004) (per curiam). Adopting the rule followed by our sister circuits, we review de novo the legal question whether Petitioner’s claim is barred by Teague. Burdine v. Johnson, 262 F.3d 336, 341 (5th Cir.2001) (en banc); O’Dell v. Netherland, 95 F.3d 1214, 1257 (4th Cir.1996) (en banc); Spaziano v. Singletary, 36 F.3d 1028, 1041 (11th Cir.1994).
DISCUSSION
The State argues that the district court .erred in granting relief on Petitionér’s amended Lackey claim because (1) Petitioner failed to exhaust state-court remedies as required by 28 U.S.C. § 2254(b)(1)(A); (2) the claim is barred by Teague because it seeks to apply a “new rule” of constitutional law; and (3) the claim fails because there is no Eighth Amendment violation, whether we review the issue de novo or under the heightened standards of § 2254(d). Because we conclude that amended claim 27 is barred by Teague, we need not and do not reach any other question.
A. We address Teague first.
As between exhaustion and Teague, we ordinarily circumstances to deny a claim on the ground that it is barred by Teague, without considering exhaustion. Two statutory provisions inform our analysis.
First, 28 U.S.C. § 2254(b)(1) provides that “[a]n application for a writ of habeas corpus ... shall not be granted unless” the petitioner has exhausted state-court remedies or has demonstrated an exception to the exhaustion requirement. (Emphasis added.) Nothing in the statute requires that we demand or analyze exhaustion if we deny the writ.
Second, § 2254(b)(2) provides affirmatively that “[a]n application for a writ of habeas corpus may be denied on the merits, notwithstanding the failure of the ap[544]*544plicant to exhaust the remedies available in the courts of the State.” We hold that Congress intended a “deni[al] on the merits” of “[a]n application for a writ of habeas corpus” to encompass the Teague inquiry.
We acknowledge that, in the abstract, the phrase “on the merits” has many potential meanings, including a narrow meaning that requires adjudication of the substantive validity of the underlying claim itself. See, e.g., Semtek Int’l Inc. v. Lockheed Martin Corp., 531 U.S. 497, 501-02, 121 S.Ct. 1021, 149 L.Ed.2d 32 (2001) (“The original connotation of an ‘on the merits’ adjudication [for purposes of Federal Rule of Civil Procedure 41(b) ] is one that actually passes directly on the substance of a particular claim before the court.” (internal quotation marks and brackets omitted)). Indeed, Congress chose words in 28 U.S.C. § 2254(d) to demonstrate precisely that meaning: “any claim that was adjudicated on the merits.” See Davis v. Ayala, — U.S. -, 135 S.Ct. 2187, 2198, 192 L.Ed.2d 323 (2015) (“Section 2254(d) thus demands an inquiry into whether a prisoner’s ‘claim’ has been ‘adjudicated on the merits’ in state court-”); Harrington v. Richter, 562 U.S. 86, 103, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011) (contrasting state-court decisions on the claim’s substantive validity with state-court decisions on other grounds).
In our view, Congress did not intend that narrow meaning in § 2254(b)(2). Unlike § 2254(d), which refers to adjudication on the merits of a claim — “any claim that was adjudicated on the merits”— § 2254(b)(2) refers to the denial on the merits of the habeas application. That textual difference strongly suggests that Congress intended a broader meaning in § 2254(b)(2). We also find persuasive, by analogy, the Supreme Court’s holding that a civil judgment “on the merits” in federal courts connotes only that the plaintiff is barred “from returning later, to the same court, with the same underlying claim.” Semtek, 531 U.S. at 505, 121 S.Ct. 1021. Accordingly, the phrase “judgment on the merits” encompasses rulings in which the court never reached the substantive validity of the claim. Id. at 504-06, 121 S.Ct. 1021. Similarly, § 2254(b)(2) encompasses, at a minimum, the substance-like inquiry demanded by Teague.
Unlike a purely procedural bar, such as a failure to meet the statute of limitations or a lack of personal jurisdiction, Teague requires an analysis of the underlying legal theory of the claim — albeit to determine its novelty rather than its ultimate persuasiveness. Moreover, in 1996, when Congress enacted AEDPA, Congress understood that a habeas application ordinarily encompassed the Teague inquiry because Teague was settled law. See, e.g., Caspari v. Bohlen, 510 U.S. 383, 389, 114 S.Ct. 948, 127 L.Ed.2d 236 (1994) (noting that Teague “prevents a federal court from granting habeas corpus relief to a state prisoner based on a rule announced after his conviction and sentence became final” (emphasis omitted)). Indeed, the Supreme Court had emphasized that, so long as the state has not waived the defense, “the court must apply Teague.” Id.; see also Horn v. Banks, 536 U.S. 266, 272, 122 S.Ct. 2147, 153 L.Ed.2d 301 (2002) (per curiam) (summarizing, more recently, that “a federal court considering a habeas petition must conduct a threshold Teague analysis when the issue is properly raised by the state”). For those reasons, we conclude that — just as we have discretion to deny a claim on its underlying substantive validity without reaching the exhaustion issue, Runningeagle v. Ryan, 686 F.3d 758, 777 n. 10 (9th Cir.2012) — we also have discretion to deny a claim as Teague-[545]*545barred without reaching the exhaustion issue.
This case warrants the exercise of our discretion to address Teague without considering the parties’ arguments concerning exhaustion.1 In the circumstances that we face here, analyzing the exhaustion issue would serve no useful purpose. If we agreed with Petitioner’s position that he has demonstrated an exception to exhaustion, we would next consider the Teague issue. Alternatively, if we agreed with the government’s position at oral argument that our recent opinion in Andrews v. Davis, 798 F.3d 759 (9th Cir.2015), requires us to hold that Petitioner already has exhausted the claim, we also would next consider the Teague issue. Both paths lead to the same destination, so it is immaterial which fork in the road we choose. See Stokley v. Ryan, 659 F.3d 802, 807-11 (9th Cir.2011) (declining to decide, among other things, whether the petitioner exhausted state-court remedies because the petitioner was not entitled to relief in any event).
■Moreover, the very nature of Petitioner’s claim is that constitutional harm flows from the delay inherent in judicial proceedings. If we know that we must deny relief under Teague, we see nothing useful to be gained by imposing more delay unnecessarily.
The recent decision of the California Supreme Court in People v. Seumanu, 61 Cal.4th 1293, 192 Cal.Rptr.3d 195, 355 P.3d 384 (2015), also affects our decision. After the district court in the present case issued its opinion, the capital defendant in Seumanu filed a supplemental brief to the California Supreme Court “raising the same Eighth Amendment/delay issue [as Petitioner has raised here] and relying heavily on the federal [district] court’s reasoning.” Id. at 437-38. The California Supreme Court unanimously held:
[Assuming for argument the facts before the court in Jones were before this court, and further assuming that evidence of systemic delay could implicate a capital defendant’s rights under the Eighth Amendment ..., we conclude defendant has not on this record demonstrated that delays in implementing the death penalty under California law have rendered that penalty impermissibly arbitrary.
Id. at 442-43 (citation truncated). In other words, even though Petitioner may not have formally exhausted his claim by raising it personally to the state courts, we have an unusual insight into the state court’s view of Petitioner’s claim. For this reason, too, we decline to subject this federal case to further delay.
We acknowledge that “[e]onstitu-tional issues are generally to be avoided, and ... the Teague inquiry requires a detailed analysis of federal constitutional law.” Lambrix v. Singletary, 520 U.S. 518, 524, 117 S.Ct. 1517, 137 L.Ed.2d 771 (1997). But judicial economy may outweigh constitutional-avoidance concerns. As the Supreme Court expressly held in Lambrix, “[j]udicial economy might counsel giving the Teague question priority, for example, if it were easily resolvable against the habeas petitioner, whereas [a non-constitutional issue] involved complicated issues of state law.” Id. at 525, 117 [546]*546S.Ct. 1517. Indeed, the Court there skipped a non-constitutional issue of state law to address Teague. Id. As in Lambrix, judicial economy favors deciding the Teague issue here. See also Lyons v. Stovall, 188 F.3d 327, 344 (6th Cir.1999) (concluding that, “in the interest of judicial economy, we will excuse the lack of exhaustion because Petitioner’s evidentiary claim is barred under the doctrine of Teague v. Lane, and thus dispositive of this case”); Fisher v. Texas, 169 F.3d 295, 303 (5th Cir.1999) (holding that, because “Fisher’s claim is barred by Teague, judicial efficiency makes it appropriate to dispose of Fisher’s claim without requiring additional litigation”).
Similar reasoning applies to the consideration of comity. “[T]he main purpose of exhaustion is to protect principles of comity between state and federal courts.” Greene v. Lambert, 288 F.3d 1081, 1088 (9th Cir.2002). “Comity ... dictates that when a prisoner alleges that his continued confinement for a state court conviction violates federal law, the state courts should have the first opportunity to review this claim and provide any necessary relief.” O’Sullivan v. Boerckel, 526 U.S. 838, 844, 119 S.Ct. 1728, 144 L.Ed.2d 1 (1999). “This rule of comity reduces friction between the state and federal court systems by avoiding the ‘unseemliness’ of a federal district court’s overturning a state court conviction without the state courts having had an opportunity to correct the constitutional violation in the first instance.” Id. at 845, 119 S.Ct. 1728 (brackets omitted). Those concerns are reduced greatly when we deny the application for habeas relief — there is no risk of “unseemliness” because we do not disturb the state court’s judgment.
Nevertheless, we have recognized that comity still plays a role in our discretionary determination to deny an unexhausted claim, at least when we deny that claim as unpersuasive: “[T]he principle of comity counsels in favor of a standard that limits a federal court’s ability to deny relief under § 2254(b)(2) to circumstances in which it is perfectly clear that the petitioner has no hope of prevailing. A contrary rule would deprive state courts of the opportunity to address a colorable federal claim in the first instance and grant relief if they believe it is warranted.” Cassett v. Stewart, 406 F.3d 614, 624 (9th Cir.2005). For the reasons discussed below, we conclude that it is “perfectly clear” that Petitioner cannot prevail.2 Id.
B. Teague bars Petitioner’s claim.
Subject to two exceptions, Teag-ue prohibits the application of a “new rule” on collateral review.3 Schriro v. Summerlin, 542 U.S. 348, 351-52, 124 S.Ct. 2519, 159 L.Ed.2d 442 (2004). “A new rule is defined as a rule that was not dictated by [547]*547precedent existing at the time the defendant’s conviction became final.” Whorton v. Bockting, 549 U.S. 406, 416, 127 S.Ct. 1173, 167 L.Ed.2d 1 (2007) (internal quotation marks and ellipsis omitted). “And a holding is not so dictated ... unless it would have been apparent to all reasonable jurists.” Chaidez v. United States, — U.S. -, 133 S.Ct. 1103, 1107, 185 L.Ed.2d 149 (2013) (internal quotation marks omitted). We must ask “whether a state court considering the defendant’s claim at the time his conviction became final would have felt compelled by existing precedent to conclude that the rule he seeks was required by the Constitution.” Caspari, 510 U.S. at 390, 114 S.Ct. 948 (brackets omitted). Petitioner’s conviction became final in 2003.
1. Petitioner seeks to apply a “new . rule. ”
Petitioner contends that the Supreme Court’s decision in Furman v. Georgia, 408 U.S. 238, 92 S.Ct. 2726, 33 L.Ed.2d 346 (1972), dictates the relevant rule to apply here. He formulates the relevant rule as the district court did: “[A] state may not arbitrarily inflict the death penalty.” Jones, 31 F.Supp.3d at 1068.
In Furman, 408 U.S. at 239-40, 92 S.Ct. 2726, the Supreme Court considered capital sentences imposed under Georgia’s and Texas’ criminal statutes. In a short per curiam opinion joined by five justices, the Court held that “the imposition and carrying out of the death penalty in these cases constitute cruel and unusual punishment in violation of the Eighth and Fourteenth Amendments.” Id. Each of the five concurring justices filed a separate opinion in support of the judgments, but each opinion received only one vote — the author’s. Id. at 240-57, 92 S.Ct. 2726 (Douglas, J., concurring); id. at 257-306, 92 S.Ct. 2726 (Brennan, J., concurring); id. at 306-10, 92 S.Ct. 2726 (Stewart, J., concurring); id. at 310-14, 92 S.Ct. 2726 (White, J., concurring); id. at 314-74, 92 S.Ct. 2726 (Marshall, J., concurring). Justices Brennan and Marshall thought that the death penalty is unconstitutional in all its applications. Id. at 305, 92 S.Ct. 2726 (Brennan, J., concurring); id. at 359, 92 S.Ct. 2726 (Marshall, J., concurring). The other three justices focused primarily on the fact that the state statutes provided no guidance to the fact-finder as to when the death penalty is appropriate, thus raising the possibility of discriminatory and arbitrary imposition. See, e.g., id. at 253, 92 S.Ct. 2726 (Douglas, J., concurring) (“[W]e deal with a system of law and of justice that leaves to the uncontrolled discretion of judges or juries the determination whether defendants committing these crimes should die or be imprisoned. Under these laws no standards govern the selection of the penalty. People live or die, dependent on the whim of one man or of 12.”); id. at 310, 92 S.Ct. 2726 (Stewart, J., concurring) (“I simply conclude that the Eighth and Fourteenth Amendments cannot tolerate the infliction of a sentence of death under legal systems that permit this unique penalty to be so wantonly and so freakishly imposed.”); id. at 311, 92 S.Ct. 2726 (White, J., concurring) (stating that he' would hold unconstitutional “capital punishment statutes under which (1) the legislature authorizes the imposition of the death penalty for murder or rape; (2) the legislature does not. itself mandate the penalty in any particular class or kind of case (that is, legislative will is not fiustrat-ed if the penalty is never imposed), but delegates to judges or juries the decisions as to those cases, if any, in which the penalty will be utilized; and (3) judges and juries have ordered the death penalty with such infrequency that the odds are now. very much against imposition and execu[548]*548tion of the penalty with respect to any convicted murderer or rapist”).
Four years later, in Gregg v. Georgia, 428 U.S. 153, 96 S.Ct. 2909, 49 L.Ed.2d 859 (1976), the Supreme Court held that an amended version of Georgia’s criminal statutes survived constitutional scrutiny. The Court described its decision in Fur-man as concerning primarily the earlier statutes’ lack of guidance given to the fact-finder in determining whether to impose the death penalty. See id. at 188-89, 96 S.Ct. 2909 (plurality opinion) (“Furman held that [the death penalty] could not be imposed under sentencing procedures that created a substantial risk that it would be inflicted in an arbitrary and capricious manner.... Furman mandates that where discretion is afforded a sentencing body on a matter so grave as the determination of whether a human life should be taken or spared, that discretion must be suitably directed and limited so as to minimize the risk of wholly arbitrary and capricious action.”); id. at 220-21, 96 S.Ct. 2909 (White, J., concurring) (“In Furman, this Court held that as a result of giving the sentencer unguided discretion to impose or not to impose the death penalty for murder, the penalty was being imposed discriminatorily, wantonly and freakishly, and so infrequently that any given death sentence was cruel and unusual.” (footnotes omitted)). The Court in Gregg held that the amended Georgia statutes — which provided new substantive standards to guide the fact-finder’s selection of punishment and new procedures, such as bifurcated guilt and penalty trials — met the concerns expressed in Furman. Id. at 206-07, 96 S.Ct. 2909 (plurality opinion); id. at 208, 220-26, 96 S.Ct. 2909 (White, J., concurring).
We have held that Teague bars a delay-based Lackey claim founded on the Supreme Court’s decisions in Furman and Gregg. In Smith v. Mahoney, 611 F.3d 978, 998 (9th Cir.2010), the petitioner cited Furman and Gregg in support of his argument that “his four sentences in combination with his twenty-five years on death row satisfied any need for retribution and deterrence and that any penalty beyond such punishment violates the Eighth Amendment.” We noted that we previously had determined, “in the context of AEDPA, that ‘[t]he Supreme Court has never held that execution after a long tenure on death row is cruel and unusual punishment.’ ” Id. (alteration in original) (quoting Allen v. Omoski, 435 F.3d 946, 958 (9th Cir.2006)). We concluded that “a state court considering [the petitioner’s] Eighth Amendment claim at the time his conviction became final would not have felt compelled by existing precedent to conclude that the rule sought was required by the Constitution.” Id. at 998-99.
Smith arguably controls here. Although the conviction in Smith became final. in 1986 and Petitioner’s conviction became final in 2003, both convictions became final well after the Supreme Court’s decisions in Furman and Gregg. Both Petitioner here and the petitioner in Smith asserted that Furman created a constitutional rule holding that extended delay in carrying out an execution violates the Eighth Amendment because it serves no retributive or deterrent purpose. We are bound by Smith to conclude that Teague bars Petitioner’s claim to the extent that his claim is the same as the petitioner’s claim in Smith.
On the other hand, both the district court and the parties have portrayed Petitioner’s claim as different than an ordinary Lackey claim like the one discussed in Smith. An ordinary Lackey claim focuses on the delay experienced by the petitioner personally, without regard to the fate of others; and it asserts the legal theory that [549]*549his continued imprisonment on Death. Row does not meet the purposes of “retribution and deterrence.” Smith, 611 F.3d at 998. Petitioner’s amended claim 27, by contrast, focuses on the delay inherent in the system itself and on the fate of capital prisoners generally, without particular regard to Petitioner’s personal situation; and it asserts the legal theory that the delay in carrying out executions among all capital prisoners represents a form of arbitrary infliction of the death penalty. In short, the parties argue that, although both types of claims — the ordinary individualized claim and Petitioner’s systemic claim— stem from Furman and Gregg, the claims present materially different legal theories.
Our recent decision in Andrews casts some doubt on that conclusion. See 798 F.3d at 789-90 (“[T]he state argues that a Lackey claim is an individual challenge, ... while [the district court’s opinion in] Jones was based on the theory that the California system itself creates the constitutional infirmity_ We disagree.”). But, for purposes of the Teague analysis, we need not decide whether the claims differ. As we explain below, even if Petitioner’s claim rests on a legal theory different than the theory advanced in Smith, Teague bars it. In particular, we assume that, although Smith rejected as Teague-barred the theory that delay undermines the purposes of “retribution and deterrence,” Smith did not address the theory that systemic delays have led to results that are unconstitutionally “arbitrary.” 4
We next consult the Supreme Court’s guidance on formulating the relevant “rule” for Teague purposes. In Sawyer, 497 U.S. at 229, 110 S.Ct. 2822, the Court considered whether it previously had announced a new rule for purposes of the Teague analysis when it decided Caldwell v. Mississippi, 472 U.S. 320, 105 S.Ct. 2633, 86 L.Ed.2d 231 (1985). The petitioner cited the Court’s pre-Caldwell cases “in support of the argument that Caldwell was ‘rooted’ in the Eighth Amendment command of reliable sentencing.” Sawyer, 497 U.S. at 235-36, 110 S.Ct. 2822. The Court agreed that those earlier cases stood “for the general proposition that capital sentencing must have guarantees of reliability.” Id. at 235, 110 S.Ct. 2822.
But the Court rejected the petitioner’s attempt to define the rule in such a broad fashion: “In petitioner’s view, Caldwell was dictated by the principle of reliability in capital sentencing. But the test would be meaningless if applied at this level of generality.” Id. at 236, 110 S.Ct. 2822. Instead, the Court considered the context of Caldwell and asked whether the rule— conceived at a more specific level — was “dictated by existing law at the time petitioner’s [conviction] became final.” Id. at 237, 110 S.Ct. 2822.
Similarly, in Beard v. Banks, 542 U.S. 406, 408, 124 S.Ct. 2504, 159 L.Ed.2d 494 (2004), the Court considered whether it had announced a new rule for Teague purposes when it decided Mills v. Maryland, 486 U.S. 367, 108 S.Ct. 1860, 100 L.Ed.2d 384 (1988). The court of appeals in Beard had considered decisions such as Lockett v. Ohio, 438 U.S. 586, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978), and “distilled the rule that the ‘Eighth Amendment prohibits any [550]*550barrier to the sentencer’s consideration of mitigating evidence.’ ” Beard, 542 U.S. at 409, 124 S.Ct. 2504. So formulated, the court of appeals had concluded that Lock-ett compelled the result in Mills and, accordingly, that Mills did not announce a new rule within the meaning of Teague. Beard, 542 U.S. at 410, 124 S.Ct. 2504.
The Supreme Court rejected that formulation: “The generalized Lockett rule (that the sentencer must be allowed to consider any mitigating evidence) could be thought to support the Court’s conclusion in Mills and [a second case]. But what is essential here is that it does not mandate the Mills rule.” Beard, 542 U.S. at 414, 124 S.Ct. 2504; see also id. at 416, 124 S.Ct. 2504 (rejecting as insufficient, for purposes of the Teague analysis, that “the Lockett principle — conceived of at a high level of generality — could be thought to support the Mills rule”). Instead, the Court asked whether, considering the more specific rules announced in Lockett and Mills, “reasonable jurists could have concluded that the Lockett line of cases did not compel Mills.” Beard, 542 U.S. at 416, 124 S.Ct. 2504; see also Gray v. Netherland, 518 U.S. 152, 169, 116 S.Ct. 2074, 135 L.Ed.2d 457 (1996) (rejecting as too broad the formulation of a rule that “a capital defendant must be afforded a meaningful opportunity to explain or deny the evidence introduced against him at sentencing”: “the new-rule doctrine ‘would be meaningless if applied at this level of generality’ ” (quoting Sawyer, 497 U.S. at 236, 110 S.Ct. 2822)); Gilmore v. Taylor, 508 U.S. 333, 344, 113. S.Ct. 2112, 124 L.Ed.2d 306 (1993) (rejecting as too broad the formulation of a rule that “the right to present a defense includes the right to have the jury consider it, and that confusing instructions on state law which prevent a jury from considering an affirmative defense therefore violate due process”: “the level of generality at which [the habeas petitioner] invokes this line of cases is far too great to provide any meaningful guidance for purposes of our Teague inquiry”).
With that guidance in mind, we must reject Petitioner’s proposed formulation of the rule: “[A] state may not arbitrarily inflict the death penalty.” We agree with Petitioner that Furman and Gregg “articulate a general Eighth Amendment standard that the death penalty is unconstitutional if imposed arbitrarily.” Andrews, 798 F.3d at 790. But the Supreme Court precedent discussed above does not allow us to define the “rule” for Teague purposes at such a high level of generality. Just as the Supreme Court rejected the proposed rule in Sawyer — -“reliability in sentencing” — even though its prior cases supported that general proposition, we must reject as too broad a rule that prohibits “arbitrariness” in imposing the death penalty. Although Furman condemned one specific form of arbitrariness related to the death penalty, it does not necessarily follow that Furman dictates the result in all other challenges to the death penalty under the banner of “arbitrariness.” Instead, we must examine whether, in 2003, reasonable jurists would have been compelled to conclude that the Eighth Amendment prohibited not only the form of arbitrariness prohibited by Furman, but also the form of arbitrariness alleged by Petitioner.
In Maynard v. Cartwright, 486 U.S. 356, 362, 108 S.Ct. 1853, 100 L.Ed.2d 372 (1988), the Court wrote: “Since Furman, our cases have insisted that the channeling and limiting of the sentencer’s discretion in imposing the death penalty is a fundamental constitutional requirement for sufficiently minimizing the risk of wholly arbitrary and capricious action.” The rule that Petitioner seeks to establish differs from Furman in two important respects. First, unlike the prisoners in Furman, Pe[551]*551titioner does not allege arbitrariness at sentencing. Instead, he alleges that the State “arbitrarily” determines when to carry out a lawfully and constitutionally imposed capital sentence. Second, Petitioner does not contend that the State has granted unfettered discretion to a fact-finder to decide on an execution date. Nor does he contend that the State intentionally chooses an execution date through a truly random selection process, such as a lottery. Instead, he contends that the delays in processing capital prisoners’ statutorily guaranteed appeals are long, such that few prisoners are ever actually executed, a result that Petitioner describes as “arbitrary” because it is hard to predict which prisoners in fact will be executed. In sum, Petitioner asks us to apply a rule that a state’s post-sentencing procedures are unconstitutionally “arbitrary” if they produce long delays resulting in few actual executions and uncertainty as to which prisoners will be executed.
Furman did not dictate such a rule. In 2003, reasonable jurists could have differed as to whether Furman applied to challenges to the delays caused by a state’s post-sentencing procedures. As an initial matter, we know of no other case in the four decades since Furman was decided that has invalidated a state’s post-sentencing procedures as impermissibly arbitrary under the Eighth Amendment, strongly suggesting that the rule is novel. See Sawyer, 497 U.S. at 236, 110 S.Ct. 2822 (“It is beyond question that no case prior to Caldwell invalidated a prosecutorial argument as impermissible under the Eighth Amendment.”). We have little doubt, of course, that Furman and Gregg would inform the analysis of Petitioner’s claim, but Teague requires much more: “Even were we to agree with [the petitioner’s] assertion that [Furman and Gregg ] inform, or even control or govern, the analysis of his claim, it does not follow that they compel the rule that [he] seeks.” Saffle v. Parks, 494 U.S. 484, 491, 110 S.Ct. 1257, 108 L.Ed.2d 415 (1990).
Importantly, there is a “simple and logical difference” between Furman’s rule prohibiting unfettered discretion by a jury deciding whether to impose the death penalty and a rule prohibiting systemic lengthy delays resulting from a state’s post-sentencing procedures in the carrying out of that sentence when permissibly imposed. Id. at 490, 110 S.Ct. 1257; see id. (“There is a simple and logical difference between rules that govern what factors the jury must be permitted to consider in making its sentencing decision and rules that govern how the State may guide the jury in considering and weighing those factors in reaching a decision.”). We and other courts previously have rejected a foundation of Petitioner’s proposed rule — that delay in resolving post-conviction proceedings has constitutional significance: “It would indeed be a mockery of justice if the delay incurred during the prosecution of claims that fail on the merits could itself accrue into a substantive claim to the very relief that had been sought and properly denied in the first place.” McKenzie v. Day, 57 F.3d 1461, 1466 (9th Cir.1995); see, e.g., Chambers v. Bowersox, 157 F.3d 560, 570 (8th Cir.1998) (“We believe that delay in capital cases is too long. But delay, in large part, is a function of the desire of our courts, state and federal, to get it right, to explore exhaustively, or at least sufficiently, any argument that might save someone’s life.... [W]e do not see how the present situation even begins to approach a constitutional violation.” (footnote omitted)); Seumanu, 192 Cal.Rptr.3d 195, 355 P.3d at 442 (“[S]uch delays are the.product of a constitutional safeguard, not a constitutional defect, because they assure careful review of the defendant’s conviction and sentence.” (internal quota[552]*552tion marks and brackets omitted)).5 Although the rule sought by Petitioner concerns not simply the delay that he has experienced personally but the allegedly “arbitrary” systemic results caused by delay, we cannot conclude, in light of the existing precedent rejecting the constitutional significance of delay, that Petitioner’s proposed rule would have been “apparent to all reasonable jurists” in 2003. Lambrix, 520 U.S. at 528, 117 S.Ct. 1517. Similarly, one might reasonably conclude that systemic delays of the sort alleged by Petitioner are not “arbitrary” in the ordinary sense of the word. See Seumanu, 192 Cal.Rptr.3d 195, 355 P.3d at 442 (“[Allowing each case the necessary time, based on its individual facts and circumstances, to'permit this court’s careful examination of the claims raised is the opposite of a system of random and arbitrary review.”). In sum, we conclude that Petitioner seeks to apply a “new rule,” which Teague prohibits.
2. Neither of Teague’s exceptions applies.
Petitioner contends, in the alternative, that Teague’s first exception— for substantive rules — applies.6 See Summerlin, 542 U.S. at 351, 124 S.Ct. 2519 (“New substantive rules generally apply retroactively.”). In particular, he argues that his proposed new rule is substantive because it would “prohibit imposition of a certain type of punishment for a class of defendants because of their status.” Sawyer, 497 U.S. at 241, 110 S.Ct. 2822 (emphasis added). For example, the Supreme Court has held that the Eighth Amendment prohibits the execution of a capital prisoner who is insane, Ford v. Wainwright, 477 U.S. 399, 410, 106 S.Ct. 2595, 91 L.Ed.2d 335 (1986), or intellectually disabled, Atkins v. Virginia, 536 U.S. 304, 321, 122 S.Ct. 2242, 153 L.Ed.2d 335 (2002). Courts have held that those rules fall within the exception for substantive rules. E.g., Penry v. Lynaugh, 492 U.S. 302, 329-30, 109 S.Ct. 2934, 106 L.Ed.2d 256 (1989), abrogated in other part by Atkins, 536 U.S. at 321, 122 S.Ct. 2242; Davis v. Norris, 423 F.3d 868, 879 (8th Cir.2005).
Petitioner does not assert that he fits into one of the traditionally recognized classes of persons whose “status” is an intrinsic quality, such as insanity or intellectual disability. Instead, Petitioner argues that he — and all California capital prisoners — belong to a class of persons with the “status as individuals whose sentence ‘has been quietly transformed’ from one of death to one of grave uncertainty and torture and one that ‘no rational jury or legislature could ever impose: life in prison, with the remote possibility of death.’ ” Pet’r’s Br. at 54 (emphasis omitted) (quoting Jones, 31 F.Supp.3d at 1053). Petitioner’s expansive description, of this exception finds no support in the cases. Nor is it supported by logic. Under Petitioner’s view, almost any procedural rule could be characterized as substantive merely by defining the petitioner as belonging to a class of persons with the “status” of those whose convictions or sentences were obtained through an uneonsti-[553]*553tutional procedural rule. We reject Petitioner’s unconventional interpretation of the exception for substantive rules.
CONCLUSION
Many agree with Petitioner that California’s capital punishment system is dysfunctional and that the delay between sentencing and execution in California is extraordinary. But “the purpose of federal habeas corpus is to ensure that state convictions comply with the federal law in existence at the time the conviction became final, and not to provide a mechanism for the continuing reexamination of final judgments based upon later emerging legal doctrine.” Sawyer, 497 U.S. at 234, 110 S.Ct. 2822. Because Petitioner asks us to apply a novel constitutional rule, we may not assess the substantive validity of his claim.
REVERSED.