dissenting.
The facts of this case are regrettable. Flagstar Bank contracted with Blue Cross Blue Shield of Michigan (“Blue Cross”) to administer Flagstar’s self-insured health-benefits plan. Among the “[sjervices”— the Contract between Flagstar and Blue Cross describes them as such — that Blue Cross agreed to provide were “[ejstablishing, arranging, and maintaining provider networks ... through contractual arrangements with preferred participating hospitals, physicians, and other health care providers!.]” One of the years in which Blue Cross agreed to provide these “[s]ervices” was 2004. In the latter half of that year, however, Blue Cross apparently concluded that its own wholly owned subsidiary, the Blue Care Network (“BCN”) HMO, was not sufficiently profitable.
So, in approximately September of that year, Blue Cross circled back to many of its participating hospitals and renegotiated the rates that those hospitals would charge not only BCN, but also self-insured plans like the Flagstar Plan. Indeed that was the idea: The decrease in BCN’s rates would be “budget neutral” for each hospital, Blue Cross explained, because Blue Cross would agree to a commensurate increase in the rates paid by the self-insured plans that Blue Cross represented. Many hospitals agreed to the deal. The rate increases were also made retroactive to January 1 of that year — about nine months before the deals were signed — which means the Flagstar Plan (and others) was billed anew *749for services it had already paid for. Thus, in a nutshell, Blue Cross lowered rates for its own subsidiary by effectively raising them for Flagstar and other self-insured plans. The letter agreements between Blue Cross and the hospitals spell out these facts in black and white.
But that does not mean that Flagstar knew about the deals. To the contrary, Blue Cross has admitted (in interrogatory responses in this case) that it never told Flagstar it had raised the Plan’s rates in order to lower them for its own subsidiary. And it appears that Flagstar was otherwise clueless about the change, because Blue Cross did not provide backup data for the bottom-line charges it sent Flagstar each month.
No one disputes that these facts would amount to a breach of fiduciary duty, in the ERISA sense of the term, if Blue Cross’s duties to Flagstar with respect to “[establishing, arranging, and maintaining provider networks” were indeed fiduciary. Whether they were, therefore, is the principal question presented here. The statute provides:
[A] person is a fiduciary with respect to a plan to the extent (i) he exercises any discretionary authority or discretionary control respecting management of such plan or exercises any authority or control respecting management or disposition of its assets ... or (iii) he has any discretionary authority or discretionary responsibility in the administration of such plan.
29 U.S.C. § 1002(21)(A). We “employ! ] a functional test to determine fiduciary status!,]” rather than focus on titles (e.g., “trustee”) or on limiting language in the parties’ agreement. Briscoe v. Fine, 444 F.3d 478, 486 (6th Cir.2006).
Whether Blue Cross functioned as a fiduciary when it established and maintained provider networks for Flagstar depends on how one characterizes their agreement. DeLuca says — and I think no one disagrees — that the function of negotiating rates with provider hospitals surely would have been fiduciary in nature had the Plan’s trustees kept that function in-house; and in DeLuca’s view, the Contract merely delegated that function from the trustees to Blue Cross. He therefore contends that Blue Cross was acting as a fiduciary when, as part of the services it provided under the Contract, it negotiated rates for the Plan. In contrast, Blue Cross argues that it actually provided a product — off-the-shelf access to its provider network at whatever rates Blue Cross cared to negotiate with them — rather than services.
The difference matters because, while selling a product tends not to create fiduciary duties under ERISA, providing services quite frequently does. And that is especially true for discretionary services that directly impact a plan’s finances. The nub of this case, therefore, is which conception of the parties’ agreement is right.
I do not think this issue is one we can fairly decide — at least in Blue Cross’s favor — as a matter of law. Each side has legitimate points to make about it. On the one hand, there is the Contract’s own description of Blue Cross’s obligation: “[ejstablishing, arranging, and maintaining provider networks!.]” (Emphasis added.) Those words describe conduct rather than a commodity. And the Contract itself characterizes these things as “Services” that “BCBSM [was] To Provide!.]” I do not think that characterization is one that Blue Cross can brush off as a matter of law, particularly given that the Contract itself recites that it “represents the entire understanding and agreement of the parties!.]” In construing a contract, the words matter; and the words here point clearly in the direction of services.
*750But, on the other hand, there is the reality that Blue Cross negotiates rates for its self-insured customers in gross, rather than individually for each of them. That does tend to make what Blue Cross provided here look less like negotiations and more like off-the-shelf access to providers and rates. The majority seems to regard this fact as dispositive. In doing so, however, I think the majority makes an assumption that we should not make for purposes of summary judgment.
The assumption is that if, in fact, Blue Cross agreed to negotiate on Flagstar’s behalf, it could only have conducted those negotiations on behalf of Flagstar alone. (Blue Cross obviously did not negotiate for Flagstar alone, so the majority concludes that it did not agree to negotiate for Flags-tar at all.) I see no basis for that assumption. The Contract nowhere prohibits Blue Cross from negotiating on behalf of all of its client plans at once. So far as the Contract is concerned, Blue Cross’s obligation was simply to establish, arrange, and maintain provider networks; and if Blue Cross discharged that obligation at the same time it discharged the same obligation to other plans, the terms of the Contract afforded Flagstar no reason to complain. So the possibility remains that Blue Cross agreed to provide what the Contract says it agreed to provide: services. And I otherwise see no basis in the record to decide, as a matter of law, that Flagstar agreed to buy a product rather than services.
Not surprisingly, then, Blue Cross’s principal arguments on appeal concern not the record, but policy. Blue Cross says we would disrupt its “business model” — and thus, we are told, the health-insurance market in Michigan — if we deemed it an ERISA fiduciary when it negotiates rates for ERISA health-benefit plans. The argument assumes that ERISA itself might forbid Blue Cross from negotiating on behalf of its client plans in gross if, in doing so, Blue Cross acts as a fiduciary. But here again I see no basis for the assumption. If, as Blue Cross continually emphasizes in its brief, Blue Cross obtains better rates for its client plans by negotiating for them in gross, I do not think the statute’s standard of care for fiduciaries (set forth in 29 U.S.C. § 1104) would require Blue Cross to negotiate on behalf of the plans individually, with worse results. The fiduciary’s duty above all is one of loyalty; and I see no breach of that duty in banding together with other plans to obtain a better result for all. What the statute would require, of course, is that Blue Cross refrain from self-dealing — which is exactly what DeLuca says happened here.
Another part of Blue Cross’s business model — as the letter deals themselves illustrate — is to negotiate rates for its wholly owned subsidiary, BCN, at the same time it negotiates rates for its client plans. Blue Cross suggests that this part of its model would go out the window if it were deemed a fiduciary when it negotiates rates for the plans. That marginal loss of leverage, Blue Cross suggests, would be a bad thing for the plans. I believe the facts of this case suggest the contrary. Sometimes loyalty is more important than leverage.
More fundamentally, I reject the unspoken premise of the preceding two arguments, which is that we should be acutely concerned about Blue Cross’s business model in the first place. Cases have consequences, and we should be mindful of them. But our task in this case is not to divine the business model that best serves the plans’ interests and those of everyone else; our task, instead, is the comparatively simple one of determining whether the letter deals violated ERISA. The wisdom *751of business models can be determined elsewhere.
Thus, to summarize: The record here would allow a jury to find that Blue Cross agreed to provide services rather than a product. Those services — “[ejstablishing, arranging, and maintaining provider networks ... through contractual arrangements” with hospitals and other healthcare providers — are highly discretionary and have a direct impact on the Plan’s bottom line. Thus, if Blue Cross indeed provided those services, it was an ERISA fiduciary when it did so. And a jury could surely find that Blue Cross breached its fiduciary duties when it made the letter deals. Summary judgment should not have been granted as to DeLuca’s claim under § 1104 of the statute.
The majority also affirms the district court’s grant of summary judgment as to DeLuca’s claim under § 1106(b)(2). They do so for the same reason that they affirm as to the § 1104 claim: Blue Cross, in their view, was not acting as a fiduciary when it negotiated rates for the Plan. See Maj. Op. at 748. But DeLuca says that reasoning should not apply to a § 1106(b)(2) claim. As an initial matter, it is undisputed that Blue Cross is a fiduciary in its role as claims processor for the Plan. And in DeLuea’s view, § 1106(b)(2) (unlike § 1104) expressly allows a fiduciary to be held liable for certain actions taken in its non-fiduciary capacity.
Section 1106(b)(2) provides in relevant part: “A fiduciary with respect to a plan shall not ... in his individual or in any other capacity act in any transaction involving the plan on behalf of a party (or represent a party) whose interests are adverse to the interests of the plan[.]” (Emphasis added.) DeLuca reads the italicized language to mean that Blue Cross can be hable for actions taken “in [its] individual or in any other capacity,” so long as the challenged actions meet the other requirements of the subsection.
I have considerable sympathy for DeLuca’s reading of this language. The subsection says that fiduciaries shall not take certain actions even in their individual or other non-fiduciary capacities. If a fiduciary then takes such an action — even in its individual capacity — a natural reading of the subsection is that it can be held liable for doing so.
But there is caselaw to the contrary. In Pegram v. Herdrich, 530 U.S. 211, 120 S.Ct. 2143, 147 L.Ed.2d 164 (2000), the Supreme Court stated that, “[i]n every case charging breach of ERISA fiduciary duty, then, the threshold question is ... whether that person was acting as a fiduciary (that is, was performing a fiduciary function) when taking the action subject to complaint.” Id. at 226, 120 S.Ct. 2143 (emphasis added). But Pegram was only a § 1104 case, so that statement is pure dicta as to § 1106(b)(2). A similar statement by our court, however, cannot be so characterized. In Hunter v. Caliber System, Inc., 220 F.3d 702 (6th Cir.2000), we said that, “by its own terms, § 1106 applies only to those who act in a fiduciary capacity.” Id. at 724. The Hunter court characterized that statement as a holding (albeit an alternative one), and I cannot fairly recast it as dicta. It is binding precedent for our circuit.
That said, the statement strikes me as Orwellian as applied to § 1106(b)(2). In the plainest conceivable English, the section bars fiduciaries from taking certain actions even in their individual capacities; and yet, we are told, the section “applies only to those who act in a fiduciary capacity.” Hunter, 220 F.3d at 724. Perhaps I am missing something. Perhaps the subsection requires not only that the fiduciary *752act in a non-fiduciary capacity on one side of a proscribed transaction, but that it also act in its fiduciary capacity on the other. (If so, to my knowledge no court has explained why that is so.) Or perhaps the statement just stands as a caution against overlong opinions with numerous alternative holdings.
That caution applies with special force to cases interpreting large and complex statutes like ERISA. Loose language in one case hardens into a holding in another, and other courts follow suit. Eventually the caselaw takes on a life of its own, often lived at variance with the rules laid down in the statute itself. We encountered precisely this scenario in Central States, SE and SW Areas Pension Fund v. Int. Comfort Products, LLC, 585 F.3d 281 (6th Cir. 2009), cert. denied, — U.S. -, 131 S.Ct. 223, 178 L.Ed.2d 134 (2010), which was another ERISA case where the case-law had diverted from the statute’s terms. See id. at 287 (“The mere accumulation of contrary precedent in three other circuits does not, in our view, give us license to disregard the plain language of [29 U.S.C.] § 1392(a)”). And the problem is compounded here because the Supreme Court’s dicta in Pegram is likely causing all the circuit courts to break one way. Perhaps eventually the Court will take a § 1106(b)(2) ease and decide whether the subsection means what it seems clearly to say.
In the meantime, we have Hunter’s holding that § 1106(b)(2) imposes liability only for actions taken by a fiduciary qua fiduciary. For the reasons already explained, the majority and I disagree as to whether DeLuca can prove that Blue Cross acted as a fiduciary when it negotiated the Plan’s rates: they say as a matter of law that Blue Cross did not, I say a jury could find otherwise. Hence I think De-Luca can prove this element of his § 1106(b)(2) claim. I also think he can prove the claim’s other elements. Notably, I think the letter agreements were transactions “involving the plan[.]” 29 U.S.C. § 1106(b)(2). In my view, transactions that have only some incidental effect upon a plan^ — raising the cost of paper clips, when the plan happens to buy them — do not “involve” the plan. But, at the same time, the statute’s text does not require that the plan have been a party to the challenged transaction. What the statute requires, I think, is something in between: a transaction that in some fashion acts directly upon a plan, even if the plan is not a party to it. In my view, “budget-neutral” letter deals that achieve that status expressly by raising rates on this Plan and others meet that requirement. Thus I would reverse the district court’s entry of summary judgment on this claim as well, and remand both claims for trial.
For these reasons, I respectfully dissent.