PBM profit already left the spread — pass-through promises are cheap

One-line summary: The spread-compression bear case on CVS/CI/UNH targets a pool the PBMs already vacated — rebates fell from ~50% of PBM revenue in 2012 to under 15% by 2023, while specialty pharmacy rose to 35% and admin fees to 22%, and the residual rebate economics were re-routed through offshore GPOs that federal law does not reach until 2028-29.

The insight

When Optum announced on 2026-07-16 that ~95% of its PBM clients would be on 100% pass-through arrangements by end-2026, the natural read was: the largest PBM is conceding the spread model, therefore the spread-compression thesis is confirmed. That read is backwards on three counts.

1. The promise covers the smallest pool. PBM revenue composition moved decisively between 2012 and 2023: rebates and related cash streams fell from ~50% of revenue to under 15%, while specialty pharmacy rose 16%→35% and administrative fees 5%→22%. A 100% pass-through promise on rebates is a promise about the smallest and fastest-shrinking of the three pools. Compression bites something that is already <15% of revenue and falling.

2. Optum is the last mover, not the forcer. The sequence runs the other way: Express Scripts (CI) announced ClearNetwork in November 2023; CVS announced CostVantage and TrueCost with cash cards in H1 2024 and PBM launches in 2025 — and CVS's earlier Guaranteed Net Cost model already passed through "100 percent of rebates to plan sponsors," with clients representing >75% of CVS Caremark commercial lives having adopted two or more elements by 2025. OptumRx's own Cost Clarity model targets full implementation only by 2028; the "95% by end-2026" claim is an acceleration of that. "Who is forced to follow" is the wrong question — CVS and Cigna moved first (2023), UNH moved last (2026). Nobody is being forced to follow Optum.

The motive is on the record from a former regulator rather than inferred — Scott Gottlieb on CVS's move: "as congress is set to act on rebates, then the benefits of paying pharmacies based on complex contracts...is starting to diminish." This is voluntary pre-emption ahead of anticipated congressional action, with no explicit mandate.

3. Pass-through relocates the profit; it does not eliminate it. Two independently-cited mechanisms:

• Cost-plus formulas hide the margin in the inputs. CVS CostVantage reimburses on an internally-computed acquisition cost index, an undisclosed percentage markup negotiated per payer, and a flat dispensing fee. Drug Channels' verdict: cost-plus "shift[s] compensation for prescriptions from a spread-based model to a service-based model," but "PBM profits relocate rather than disappear" — CVS's vertical integration means it "could use its other businesses [Cordavis] to establish acquisition cost benchmarks that enable higher reimbursement to its pharmacies." Because markups are negotiated per payer, "different payers and plans will pay different amounts for the same prescription — replicating current problems rather than solving them."
• The GPO layer converts a regulated rebate into an unregulated fee. Ascent Health Services (Cigna/Evernorth, 2019, Switzerland), Zinc Health Services (CVS, 2020, Minnesota) and Emisar Pharma Services (UnitedHealth/Optum, 2021, Ireland) sit between the PBMs and manufacturers. Drug Channels lists the reasons the big PBMs built them, and one is explicit: "there is no parallel effort to alter GPO safe harbor rules" — a stated regulatory hedge. Manufacturers paid an estimated $7.6 billion in fees to rebate aggregators in 2022 alone.

The hedge has a shelf life, and it is dated. The CAA 2026 names the workaround — it requires remittance of 100% of "rebates, fees, alternative discounts, and other remuneration received from manufacturers, GPOs, and rebate aggregators," with a Part D bona fide service fee model (flat, FMV, not varying with drug price or formulary decisions). But effective 2028 (Part D) and 2028-2029 (commercial). The hedge is a 2028 problem, not a 2026 one. Mintz flags a live ambiguity: "The statute does not carve out PBM-affiliated entities; however, it remains unclear whether affiliated GPOs trigger separate compliance obligations."

Retention is enforced-against, so it is real rather than theoretical: the Illinois AG recovered $45 million from CVS Caremark in 2024 for retaining Zinc-associated rebate funds owed to clients, and a federal judge in Rhode Island declined to dismiss that state's suit against all three PBMs.

Evidence

All from 2026-07-17-autoresearch-pbm-pass-through-preemption-profit-pool-migration.

• The revenue-mix migration (the load-bearing fact) — Forbes/Joshua Cohen (2026-06-01): rebates and related cash streams ~50% → under 15% of PBM revenue (2012→2023); specialty pharmacy 16% → 35%; administrative fees 5% → 22%.
• The pre-emption sequence — Express Scripts ClearNetwork, November 2023 (flat pharmacy fee + 15% fee across 65,000 retail pharmacies); CVS CostVantage/TrueCost, cash cards H1 2024, PBM launches 2025; CVS Guaranteed Net Cost already passed through "100 percent of rebates to plan sponsors", with >75% of CVS Caremark commercial lives on two or more elements by 2025 (⚠ search-layer only — Fierce Healthcare 403'd). OptumRx Cost Clarity targets full implementation "by 2028" (⚠ search synthesis).
• The stated motive — Scott Gottlieb via PharmExec: "as congress is set to act on rebates, then the benefits of paying pharmacies based on complex contracts...is starting to diminish." PharmExec's framing: voluntary preemption ahead of anticipated congressional action, with no explicit regulatory mandate.
• Profit relocates — Drug Channels: cost-plus "shift[s] compensation... from a spread-based model to a service-based model" but "PBM profits relocate rather than disappear"; CVS "could use its other businesses [Cordavis] to establish acquisition cost benchmarks that enable higher reimbursement to its pharmacies"; "different payers and plans will pay different amounts for the same prescription — replicating current problems rather than solving them"; clients get only "some sort of audit rights."
• The GPO hedge, stated — Drug Channels ("Five (or Maybe Six?) Reasons the Largest PBMs Operate GPOs"): reasons include new revenue streams from manufacturers ("contracting entity administrative fees, prescription data services, data portals, enterprise fees," supplementing typical PBM admin fees of "3% to 5% of the WAC list price value"); reduced transparency and pass-through obligations (GPO "businesses and finances are less transparent to plan sponsors," so "a smaller share of those revenues are passed through"); tax efficiencies (Switzerland/Ireland "transfer pricing and rebate accounting using other countries' lower corporate tax rates"); and the regulatory hedge — reforms target rebates and spread while "there is no parallel effort to alter GPO safe harbor rules."
• The GPO entities — Ascent Health Services (Cigna/Evernorth, 2019, Switzerland), Zinc Health Services (CVS, 2020, Minnesota), Emisar Pharma Services (UnitedHealth/Optum, 2021, Ireland).
• The fee scale — manufacturers paid an estimated $7.6 billion to rebate aggregators in 2022 (⚠ search layer over Frier Levitt; primary not retrieved).
• The mechanic — Buchanan Ingersoll (via search synthesis; page 403'd): "If what used to be a rebate is reclassified as a GPO 'administrative fee,' the money moves to a different bucket — one clients can't claim and regulators haven't regulated, and one that sits offshore." Economic substance of rebate retention "can remain intact even as contractual promises appear satisfied."
• Enforcement precedent — Illinois AG recovered $45 million from CVS Caremark in 2024 over Zinc-associated rebate retention (Grokipedia); a federal judge in Rhode Island declined to dismiss that state's suit against all three PBMs.
• The shelf life — Mintz on the CAA 2026: 100% remittance of remuneration from "manufacturers, GPOs, and rebate aggregators", quarterly within 90 days, GPO/aggregator pass-through within 45 days, auditable by a plan-selected auditor the PBM cannot fund; Part D bona fide service fee model — compensation "only in the form of a bona fide service fee" that "does not vary based on drug price, Rebates, coverage or formulary decisions." Effective 2028 plan year (Part D) and 2028-2029 (commercial). Medicaid explicitly excluded (GAO study only).
• The proximate trigger the earnings call didn't state — FTC insulin cases against all three PBM-GPO pairs resolved or near-resolved in the weeks before the call; the 8th Circuit dismissed the PBMs' countersuit against the FTC on 2026-07-01 (Healthcare Dive); Optum Rx "accepted a proposed settlement in June 2026" with terms undisclosed.

Implications

• The spread-compression bear case on CVS/CI/UNH does not clear this project's bar through 2027. Compression is real and arriving, but it bites a pool that is already <15% of revenue and shrinking. See the contradiction recorded on drug-pricing-reform-pbm-spread-compression.
• The real structural risk is mis-located. It is not spread transparency — it is divestiture law aimed at specialty pharmacy, the pool that grew to 35% and which Forbes calls "one of the industry's major profit engines." The Patients Before Monopolies Act (bipartisan, reintroduced ~May 2026) would force PBM-owning conglomerates to divest retail pharmacies within a year; Arkansas HB 1150 is the state version (enjoined since 2025-07-28). Low probability, high impact, no dated catalyst — a watch item, not a chain.
• The dated catalyst that matters is 2028, not 2026: when the CAA's GPO-remittance and bona-fide-service-fee provisions bite. Anything before that is pre-emption theatre on a shrinking pool.
• Reads across to legislative-divergence-base-rate: PBMs building GPOs because "there is no parallel effort to alter GPO safe harbor rules" is Brian Potter's divergence mechanism running deliberately rather than accidentally — regulated parties routing around a statute into an unregulated adjacent capability. Divergence is not always an accident; sometimes it is engineered.
• The binding-now forcing function is elsewhere — see state-nadac-floor-reimbursement-mandates.

Contradictions / tensions

• ⚠ The FTC-consent → 07-16-announcement causal link is the autoresearch's inference, not a sourced claim — and it is the single highest-value thing to resolve. Optum's settlement terms are undisclosed, so it is possible the consent agreement requires the pass-through — which would flip the finding from "voluntary pre-emption" to "mandated," and materially change the read. The timing is suggestive; the mechanism is not sourced.
• ⚠ The 2012→2023 revenue-mix figures carry this entire concept and are effectively single-sourced. They appear in Forbes and in a search synthesis over the same underlying material, and could not be traced to the primary study. They are also 2023 data being used to argue about 2026.
• ⚠ The $7.6B GPO fee figure (2022) comes from the search layer over Frier Levitt, not a fetched page.
• Source conflicts on the FTC settlement timeline, unresolved: Healthcare Dive says Express Scripts finalized Feb 2025 and Caremark agreed to a proposed settlement March 2026; the search layer says Express Scripts settled Feb 2026 ("projected up to $7 billion," "reshoring Ascent from Switzerland") and Caremark's is a July 2026 announcement ("up to $8.5 billion in consumer savings over the next decade"). The FTC's own press-release URL is dated 2026/07. ftc.gov 403'd on three separate attempts, so nothing could be verified against the primary. The Ascent-reshoring claim would partially close the offshore leg if true — snippet-only, unverified.
• Do Optum's and Caremark's consent orders reach the affiliated GPOs (Emisar, Zinc), or only rebate practices? This determines whether the hedge survives to 2028.
• No PBM discloses GPO fee revenue separately, so "100% pass-through" is unauditable from outside — and the Mintz-flagged ambiguity stays live.

Open questions

• Does Optum's FTC consent order require the pass-through? If yes, "voluntary pre-emption" is wrong and the concept's framing needs revision. Terms undisclosed.
• Can the 50%→15% / 16%→35% / 5%→22% mix figures be traced to a primary study, and is there a 2026 update?
• What share of CVS/CI/UNH EBIT is GPO-routed? Still unsized — "the same gap the existing page flagged in June, now sharper because we know which pool to size."
• Does any consent order or rule reach affiliated GPOs before 2028?

Valuation snapshot

Last refreshed 2026-07-20 (pre-open; marks are the Friday 2026-07-17 close, markets closed over the weekend). Price fills tagged twelvedata.

✎ All three of this page's suggested tickers are now marked — today's Twelve Data pull is broader than the 07-17 run's 77-symbol set, so CVS, CI and UNH carry prices this run where they were blank before.

Sector read-across from the pull: XLF $56.75 (+0.34%) — financials closed green in the defensive rotation; managed care is not in XLF, so this is directional context only, not a proxy.

Forward-looking outcomes (12-month)

Bull case (for CVS/CI/UNH — i.e. the spread-compression bear case fails, as this concept argues) — the pass-through headlines keep landing on a pool that is already gone, and the market keeps mistaking them for compression: the load-bearing fact is the revenue-mix migration — rebates and related cash streams fell from ~50% of PBM revenue in 2012 to under 15% by 2023, while specialty pharmacy rose 16%→35% and administrative fees 5%→22%. A 100% pass-through promise on rebates is a promise about the smallest and fastest-shrinking of the three pools. Meanwhile the profit relocates rather than disappears: Drug Channels' verdict on cost-plus is that it "shift[s] compensation... from a spread-based model to a service-based model" but "PBM profits relocate rather than disappear." And the offshore GPO layer (Ascent/Switzerland, Zinc/Minnesota, Emisar/Ireland) converts a regulated rebate into an unregulated fee — built, on Drug Channels' account, partly because "there is no parallel effort to alter GPO safe harbor rules." Implied price: unsized — no marks. Cited: 2026-07-17-autoresearch-pbm-pass-through-preemption-profit-pool-migration.

Base case — nothing breaks before 2028, and the sector trades on managed-care fundamentals rather than PBM reform: the dated federal catalyst is 2028 (Part D) / 2028-2029 (commercial), when the CAA 2026's remittance and bona-fide-service-fee provisions bite — and those do name the workaround, requiring remittance of remuneration from "manufacturers, GPOs, and rebate aggregators." The hedge has a shelf life; it is just not a 2026 or 2027 problem. Anything before that is pre-emption theatre on a shrinking pool. Implied price: unsized. Cited: 2026-07-17-autoresearch-pbm-pass-through-preemption-profit-pool-migration.

Bear case — the real risk is mis-located by the whole debate, and it is divestiture: not spread transparency but divestiture law aimed at specialty pharmacy — the pool that grew to 35% and which Forbes calls "one of the industry's major profit engines." The Patients Before Monopolies Act (bipartisan, reintroduced ~May 2026) would force PBM-owning conglomerates to divest retail pharmacies within a year; Arkansas HB 1150 is the state version (enjoined since 2025-07-28). Low probability, high impact, no dated catalyst. The nearer-dated pressure is state law — see state-nadac-floor-reimbursement-mandates, where Frier Levitt puts the incidence of NADAC-floor costs "primarily [on] PBMs and health plans." Implied price: unsized. Cited: 2026-07-17-autoresearch-pbm-pass-through-preemption-profit-pool-migration.

Currently undervalued vs base case? No view on the names — but the concept delivers a clear directional verdict, which is what it exists for.

The verdict: the spread-compression bear case on CVS/CI/UNH does not clear this project's bar through 2027. Compression is real and arriving, but it bites a pool that is already <15% of revenue and shrinking, and the profit has demonstrably relocated into cost-plus markups and offshore GPO fees that federal law does not reach until 2028-29. If you were short these names on the 2026-07-16 Optum announcement, you were short the wrong thing. See the contradiction recorded on drug-pricing-reform-pbm-spread-compression.

⚠ Three caveats that keep this from being a long call rather than merely "not a short":

• The trigger may be mandated, not voluntary — and that flips the finding. "The FTC-consent → 07-16-announcement causal link is the autoresearch's inference, not a sourced claim." Optum's settlement terms are undisclosed, so the consent agreement may require the pass-through. The timing is suggestive; the mechanism is not sourced. This is the single highest-value thing to resolve.
• The load-bearing numbers are effectively single-sourced and stale. The 2012→2023 mix figures "carry this entire concept," could not be traced to the primary study, and are 2023 data being used to argue about 2026.
• "100% pass-through" is unauditable from outside. No PBM discloses GPO fee revenue separately, and no source sizes what share of CVS/CI/UNH EBIT is GPO-routed. The relocation is evidenced; its magnitude is not.

Catalyst path:

• Disclosure of Optum's FTC consent-order terms — decides whether "voluntary pre-emption" (this concept's framing) or "mandated" (which would require revising it) is correct.
• CVS / CI / UNH Q2 2026 earnings (late July–early August) — the first calls after the 07-16 announcement; watch for any GPO fee-revenue disclosure, and for whether the pass-through is characterised as voluntary or required.
• Patients Before Monopolies Act floor action / Arkansas HB 1150 appeal — the divestiture tail. Low probability, high impact, currently undated.

Related

• drug-pricing-reform-pbm-spread-compression — the hypothesis this concept contradicts
• state-nadac-floor-reimbursement-mandates — the forcing function that actually binds now
• legislative-divergence-base-rate — divergence engineered on purpose
• unh · patrick-conway — the 2026-07-16 announcement that triggered this research