OPINION

KETHLEDGE, Circuit Judge.

Craig Crockett's law firm signed an adhesion contract with LexisNexis that contained an arbitration clause. Eventually the parties had a billing dispute. The arbitration clause provided that any arbitration with respect to LexisNexis's charges must occur in the city where LexisNexis is located. That provision and others made arbitration of Crockett's individual claims economically unfeasible, so Crockett filed an arbitration demand on behalf of himself and a putative class of other LexisNexis customers. But the arbitration clause says nothing about classwide arbitration, and the Supreme Court has recently made clear that we must interpret arbitration clauses according to their terms. We therefore agree with the district court that the arbitration clause does not permit the classwide arbitration that Crockett seeks here.

I.

LexisNexis (a business division of Reed Elsevier) provides legal-research services, primarily on-line. In 2007, Craig Crockett and his former law firm — Dehart & Crockett, P.C. — subscribed to a LexisNexis Subscription Plan. The Plan allowed subscribers unlimited access to certain legal databases for a flat, monthly fee. Subscribers could access other databases for an additional fee. According to Crockett, LexisNexis told subscribers that a warning sign — such as a dollar ($) sign — would display if the subscriber was about to use a database outside of the Plan.

Several years after signing up for the Plan, Crockett complained to LexisNexis that his firm was being charged additional fees without any warning that the firm was using a database outside the Plan. LexisNexis allegedly insisted on payment of the additional fees anyway. Soon thereafter, Dehart & Crockett dissolved. Crockett then formed the Crockett Firm and entered into a LexisNexis subscription agreement that is materially identical to the Plan.

The Plan contains an arbitration clause. In 2010, Crockett filed an arbitration demand with the American Arbitration Association against LexisNexis on behalf of himself and two putative classes. One class comprised law firms that were charged additional fees by LexisNexis. The other class comprised clients onto whom such fees were passed. The demand sets forth state-law claims for fraud, negligent misrepresentation, breach of contract, negligence, gross negligence, unjust enrichment, and violation of the New York Consumer Protection Act. Crockett sought damages in excess of $500 million.

In response, LexisNexis sued Crockett in a federal district court in Ohio, seeking a declaration that the Plan's arbitration clause does not authorize class arbitration. LexisNexis also sought an injunction barring Crockett from proceeding with classwide arbitration. In an opinion that thoroughly canvassed the caselaw, the district court granted summary judgment in favor of LexisNexis on its declaratory claim and dismissed the injunctive claim without prejudice.

This appeal followed.

II.

We review the district court's grant of summary judgment de novo. Grden v. Leikin Ingber & Winters PC, 643 F.3d 169, 171 (6th Cir.2011).

A.

1.

Crockett first argues that an arbitrator, rather than the district court, should have decided whether the Plan's arbitration clause authorizes classwide arbitration. "[A]rbitrators derive their authority to resolve disputes only because the parties have agreed in advance to submit such grievances to arbitration." AT & T Techs., Inc. v. Commc'n Workers of Am., 475 U.S. 643, 648-49, 106 S.Ct. 1415, 89 L.Ed.2d 648 (1986). Thus, an arbitrator has authority to answer the question whether an agreement provides for classwide arbitration — a question we refer to here as "classwide arbitrability" — only if the parties have authorized the arbitrator to answer that question. See First Options of Chicago, Inc. v. Kaplan, 514 U.S. 938, 943, 115 S.Ct. 1920, 131 L.Ed.2d 985 (1995). As to this much, the law is clear.

Less clear is the showing necessary for a court to decide that the parties have authorized an arbitrator to determine classwide arbitrability. On the one hand, courts presume that so-called "gateway disputes" are "for judicial determination unless the parties clearly and unmistakably provide otherwise." Howsam v. Dean Witter Reynolds, Inc., 537 U.S. 79, 83, 123 S.Ct. 588, 154 L.Ed.2d 491 (2002) (internal quotation marks and alterations omitted). Gateway disputes include "whether the parties have a valid arbitration agreement at all or whether a concededly binding arbitration clause applies to a certain type of controversy." Green Tree Fin. Corp. v. Bazzle, 539 U.S. 444, 452, 123 S.Ct. 2402, 156 L.Ed.2d 414 (2003) (plurality opinion). These matters are important enough that courts "hesitate to interpret silence or ambiguity" as grounds for giving an arbitrator the power to decide them, because "doing so might too often force unwilling parties to arbitrate a matter they reasonably would have thought a judge, not an arbitrator, would decide." First Options, 514 U.S. at 945, 115 S.Ct. 1920.

On the other hand, "the law reverses the presumption[,]" id., with respect to what we refer to here as "subsidiary questions." Subsidiary questions "grow out of the dispute and bear on its final disposition[,]" John Wiley & Sons, Inc. v. Livingston, 376 U.S. 543, 557, 84 S.Ct. 909, 11 L.Ed.2d 898 (1964); and they include, for example, issues related to "waiver, delay," or "whether a condition precedent to arbitrability has been fulfilled." Howsam, 537 U.S. at 84-85, 123 S.Ct. 588 (quotation marks and citation omitted). Once a court decides that the parties have agreed to resolve a particular dispute through arbitration, it follows almost a fortiorari — absent clear language to the contrary in the parties' agreement — that they would have agreed to have an arbitrator decide these subsidiary questions as well.

So we must determine whether classwide arbitrability is a gateway question or a subsidiary one. The Supreme Court faced this same issue in Bazzle, though only a plurality of the Justices agreed upon its resolution. There, the plurality concluded that classwide arbitrability is merely a subsidiary question (as we use that term here) because it concerns not whether the parties "agreed to arbitrate a matter[,]" but rather "what kind of arbitration proceeding the parties agreed to." 539 U.S. at 452, 123 S.Ct. 2402 (emphasis omitted). Crockett urges us to adopt the same reasoning and conclusion here.

Although the Supreme Court's puzzle of cases on this issue is not yet complete, the Court has sorted the border pieces and filled in much of the background. As an initial matter, the Court has pointedly observed that "only the plurality" in Bazzle decided whether classwide arbitrability is a gateway question. Stolt-Nielsen S.A. v. AnimalFeeds Int'l Corp., 559 U.S. 662, 680, 130 S.Ct. 1758, 176 L.Ed.2d 605 (2010). And just last Term, the Court flatly stated that it "has not yet decided whether the availability of class arbitration" is a gateway question. Oxford Health Plans LLC v. Sutter, ___ U.S. ___, 133 S.Ct. 2064, 2068 n. 2, 186 L.Ed.2d 113 (2013). Thus, the issue before us — whether classwide arbitrability is presumptively for an arbitrator to decide, or presumptively for a judge — remains an open one.

But for Crockett the caselaw is even worse than that — for recently the Court has given every indication, short of an outright holding, that classwide arbitrability is a gateway question rather than a subsidiary one. The Court has stated that "it cannot be presumed the parties consented to [classwide arbitration] by simply agreeing to submit their disputes to an arbitrator." Stolt-Nielsen, 559 U.S. at 685, 130 S.Ct. 1758. Indeed, for several reasons, the Court has characterized the differences between bilateral and classwide arbitration as "fundamental." Id. at 686, 130 S.Ct. 1758; AT&T Mobility LLC v. Concepcion, ___ U.S. ___, 131 S.Ct. 1740, 1750, 179 L.Ed.2d 742 (2011) (same). First, arbitration's putative benefits — "lower costs, greater efficiency and speed," et cetera — "are much less assured" with respect to classwide arbitration, "giving reason to doubt the parties' mutual consent" to that procedure. Stolt-Nielsen at 685, 130 S.Ct. 1758; see also Concepcion, 131 S.Ct. at 1751 (stating that "the switch from bilateral to class arbitration sacrifices the principal advantage of arbitration — its informality — and makes the process slower, more costly, and more likely to generate procedural morass than final judgment"). Second, "[c]onfidentiality becomes more difficult" in classwide arbitrations, id. at 1750 — thus "potentially frustrating the parties' assumptions when they agreed to arbitrate." Stolt-Nielsen, 559 U.S. at 686, 130 S.Ct. 1758. Third, "the commercial stakes of class-action arbitration are comparable to those of class-action litigation" — indeed, Crockett seeks an award of $500 million here — "even though the scope of judicial review is much more limited[.]" Id. at 686-87, 130 S.Ct. 1758. And then there are the due-process concerns: once an arbitration is expanded classwide, "[t]he arbitrator's award no longer purports to bind just the parties to a single arbitration agreement, but adjudicates the rights of absent parties as well." Id. at 686, 130 S.Ct. 1758. Consequently, the absent parties "must be afforded notice, an opportunity to be heard, and a right to opt out of the class." Concepcion, 131 S.Ct. at 1751. Indeed, "where absent class members have not been required to opt in, it is difficult to see how an arbitrator's decision to conduct class proceedings could bind absent class members who have not authorized the arbitrator to decide on a classwide basis which arbitration procedures are to be used." Oxford Health, 133 S.Ct. at 2071-72 (Alito, J., concurring). Thus, in sum, "[a]rbitration is poorly suited to the higher stakes of class litigation." Concepcion, 131 S.Ct at 1752.

Crockett's argument does not fit this puzzle. Gateway questions are fundamental to the manner in which the parties will resolve their dispute — whereas subsidiary questions, by comparison, concern details. And whether the parties arbitrate one claim or 1,000 in a single proceeding is no mere detail. Unlike the question whether, say, one party to an arbitration agreement has waived his claim against the other — which of course is a subsidiary question — the question whether the parties agreed to classwide arbitration is vastly more consequential than even the gateway question whether they agreed to arbitrate bilaterally. An incorrect answer in favor of classwide arbitration would "forc[e] parties to arbitrate" not merely a single "matter that they may well not have agreed to arbitrate[,]" Howsam, 537 U.S. at 84, 123 S.Ct. 588, but thousands of them. We therefore hold that the question whether an arbitration agreement permits classwide arbitration is a gateway matter, which is reserved "for judicial determination unless the parties clearly and unmistakably provide otherwise." Id. at 83, 123 S.Ct. 588 (internal quotations and alterations omitted).

2.

Crockett cannot make that showing here. The Plan's arbitration clause provides, in relevant part:

The clause also provides: "Issues of arbitrability will be determined in accordance and solely with the federal substantive and procedural laws relating to arbitration[.]"

This language does not clearly and unmistakably assign to an arbitrator the question whether the agreement permits classwide arbitration. Instead it does not mention classwide arbitration at all. It is true that the clause provides that "any controversy ... arising out of or in connection with this Order" shall be resolved by binding arbitration; and one might argue that the question whether an arbitrator should decide classwide arbitrability is a "controversy ... arising ... in connection with" Crockett's order. That, indeed, was the interpretation that the plurality gave to analogous language in Bazzle. See 539 U.S. at 448, 123 S.Ct. 2402 (plurality opinion). But given the total absence of any reference to classwide arbitration in this clause, the agreement here can just as easily be read to speak only to issues related to bilateral arbitration. Thus, at best, the agreement is silent or ambiguous as to whether an arbitrator should determine the question of classwide arbitrability; and that is not enough to wrest that decision from the courts. Stolt-Nielsen, 559 U.S. at 684-85, 130 S.Ct. 1758. We therefore agree with the district court that the question whether Crockett and LexisNexis agreed to arbitrate must "be decided by the court, not the arbitrator." AT & T Techs., 475 U.S. at 649, 106 S.Ct. 1415. And so we turn to that question next.

B.

The principal reason to conclude that this arbitration clause does not authorize classwide arbitration is that the clause nowhere mentions it. A second reason, as the district court correctly observed, is that the clause limits its scope to claims "arising from or in connection with this Order," as opposed to other customers' orders. Crockett responds that the arbitration clause refers to the AAA's Commercial Rules, which themselves incorporate the AAA's Supplemental Rules for Class Arbitration. But the Supplemental Rules expressly state that one should "not consider the existence of these Supplementary Rules, or any other AAA rules, to be a factor either in favor of or against permitting the arbitration to proceed on a class basis." Crockett also responds that the agreement does not expressly exclude the possibility of classwide arbitration, which is true enough. But the agreement does not include it either, which is what the agreement needs to do in order for us to force that momentous consequence upon the parties here.

The Supreme Court has made clear that "[a]n implicit agreement to authorize class-action arbitration" should not be inferred "solely from the fact of the parties' agreement to arbitrate." Stolt-Nielsen, 559 U.S. at 685, 130 S.Ct. 1758. That, at bottom, is the inference that Crockett asks us to make here. The agreement in this case does not provide for classwide arbitration.

C.

Crockett's remaining argument is that, if read not to permit classwide arbitration, the arbitration clause is unconscionable. The clause is indeed as one-sided as Crockett says: the clause favors LexisNexis at every turn, and as a practical matter makes it economically unfeasible for Crockett or any other customer to assert the individual claims that Crockett seeks to assert here. The clause provides that any arbitration of any dispute concerning LexisNexis's charges must occur in Dayton, Ohio, where LexisNexis is headquartered. The customer must pay his own legal fees, even if the arbitrator concludes that LexisNexis's charges were improper. And unlike many corporations that require arbitration of disputes with their customers, LexisNexis makes its customer split the tab for the arbitrator's fee.

The idea that the arbitration agreement in this case reflects the intent of anyone but LexisNexis is the purest legal fiction. But all of these things — the one-sided nature of the arbitration clause, and its adhesive nature — were also present in American Express Co. v. Italian Colors Restaurant, ___ U.S. ___, 133 S.Ct. 2304, 186 L.Ed.2d 417 (2013). And there the Supreme Court held that, all of those concerns notwithstanding, the absence of a class-action right does not render an arbitration agreement unenforceable. Id. at 2309 (The solution to Crockett's problem is likely a market solution; as the district court observed, Westlaw's agreement with its customers lacks any arbitration clause, much less a clause of the sort at issue here.) Under Italian Colors, therefore, the agreement here is not unconscionable.

The district court's judgment is affirmed.