The Hemsley Turnaround Model

The fourth largest company in the world spent a year deliberately getting smaller. Revenue held flat. Operating earnings rose fifty five percent. Wall Street re-rated it.

By RaeAnn, Founder and Chief Executive Officer, HLTHWORKS

2.8 MILLION MEMBERS

Members UnitedHealthcare projected it would shed in 2026 across Medicare Advantage, commercial risk, Medicaid, and the ACA marketplace.

Every instinct the industry built over twenty years said growth was the answer. Hemsley did the opposite in every line of business at once, and it worked. The question for everyone else is whether the model transfers, or whether it only works when you are large enough to give up 2.8 million members and still be the fourth largest company on earth.

FIGURE	WHAT IT IS
$8 billion	One year swing at Optum Health, from $6.9 billion earned in 2024 to $1.1 billion lost in 2025
86.7 percent	Second quarter medical care ratio, improved from 89.4 percent
$3 billion	Artificial intelligence program spanning 2026 and 2027
55 percent	Operating earnings growth on revenue that did not move

What he came back to

The scale of the 2026 recovery only reads correctly against the size of the hole. Optum Health, the care delivery business that was supposed to be the strategic differentiator, earned $6.9 billion in operating income in 2024. In 2025 it lost $1.1 billion. That is an eight billion dollar swing in a single year, on revenue that itself fell from $103.5 billion to $100.1 billion.

The fourth quarter alone accounts for most of it: a $3.0 billion operating loss and a negative twelve percent margin. Across all of Optum, full year operating margin fell from 6.6 percent to 3.5 percent.

Stephen Hemsley returned as chief executive in May 2025 and restructured the leadership team. What he inherited was not a bad quarter. It was a company whose most strategically important segment had stopped making money, in a year when medical cost trend ran well above what any of its books had priced.

The turnaround did not begin with a new idea. It began with a decision about what to stop doing.

By the second quarter of 2026 the numbers had inverted. Adjusted earnings of $6.38 per share against a $4.90 consensus, a thirty percent beat and up fifty six percent year over year. Consolidated operating earnings of $7.99 billion against $5.15 billion. The medical care ratio at 86.7 percent, down from 89.4 percent. Guidance raised to $19.50 to $20.00 against an $18.47 street estimate, the buyback doubled to at least $5 billion, and debt to capital down to 41.2 percent from 43.9 percent.

And revenue essentially did not move. $112.03 billion against $111.62 billion a year earlier, a gain of four tenths of one percent. Every dollar of that earnings improvement came from somewhere other than growth.

Footprint. What came out, and what quietly went in.

The reduction happened at every level of the company simultaneously, and it is easy to miss because each piece was reported as a separate story.

LAYER	WHAT CAME OUT
Continents	All of South America. Brazil and Peru completed in 2024, Chile and Colombia with the sale of Banmedica in late 2025.
Medicare Advantage	A projected decline of 1.3 million members for 2026. More than one hundred plans withdrawn across 109 counties. Membership down roughly 965,000 since year end.
Commercial risk	A projected contraction of 1.3 to 1.4 million fully insured members.
Medicaid	A projected contraction of 565,000 to 715,000 including dual eligible enrollees, against a base of 7.4 million.
ACA marketplace	Individual enrollment down roughly one third, with the remaining book steered toward bronze and gold tiers where member mix and utilization align to plan.
Care delivery	Optum Health serving roughly 700,000 fewer value based care patients.
Legacy product	Non artificial intelligence OptumInsight products run down beginning in the first quarter.
Disclosed structure	The annual report now lists ten significant subsidiaries, where prior filings had enumerated the corporate structure at length.

THE LINE NOBODY PUT IN A HEADLINE

Against all of that, one number moved the other way. UnitedHealthcare projected growth of 550,000 to 750,000 self funded members.

That is the mechanism. Self funded business carries administrative fees rather than medical risk. The company did not shrink so much as convert: risk bearing lives out across four product lines, fee bearing lives in. Membership falls by as much as 2.8 million, revenue holds flat, and the risk underneath that flat revenue is a different animal. The chief executive of UnitedHealthcare described the Medicare Advantage reduction as a deliberate effort to reprice the book rather than defend volume. That is the doctrine in one sentence, and it applies to all four lines.

THE EIGHT POINT THREE BILLION DOLLAR SENTENCE

The Latin America exit states the philosophy most clearly, and it predates Hemsley’s return. The strategy shift began in 2022, Brazil and Peru closed in 2024, and Banmedica sold for roughly $1 billion in late 2025 against the $2.8 billion paid in 2018, moving seven hospitals, forty seven medical centers, and coverage for about 1.7 million members in Chile and Colombia.

Total net loss booked on the South American exit: $8.3 billion, roughly $7.1 billion from Brazil and $1.2 billion from Banmedica. International operations had represented less than three percent of revenue. A company does not write off eight point three billion dollars to escape a three percent distraction unless it has concluded that focus is worth more than the loss. Hemsley did not start that retreat. He finished it, and then applied the same arithmetic to counties, products, and lines of business.

Efficiency, and how much of it is real

Subtraction alone does not produce a fifty five percent earnings increase. The second half of the model is an operating cost program, and it is the largest one in the sector.

UnitedHealth is investing roughly $1.5 billion in artificial intelligence in 2026 inside a $3 billion program spanning 2026 and 2027. Roughly a third of the 2026 spend goes to transforming OptumInsight. Management has said it expects to reduce operating costs by close to $1 billion this year, largely driven by the same program, and has publicly claimed a two to one return.

WHERE IT SHOWS UP	WHAT HAS BEEN DISCLOSED
Scale of deployment	More than one thousand active use cases and 117 customized large language models. Of roughly 22,000 software engineers worldwide, more than eighty percent now use artificial intelligence to write code.
Prior authorization	Roughly ninety five percent of requests submitted electronically, about half processed in real time, more than ninety percent decided within one business day, with a stated target of reducing medical prior authorizations by thirty percent or more.
Pharmacy and claims	An Optum Rx capability cutting prescription approval from more than eight hours to under thirty seconds. A real time claims platform reported to have processed roughly 500 million claims.
Governance	An internal review board including medical ethicists, clinicians, technologists, and privacy and legal experts clears proposals for new uses.

THE DETAIL THAT TELLS YOU IT IS NOT A PILOT

Active use of artificial intelligence tools is tracked as an employee performance metric. Whatever one makes of that as a management practice, it answers the question every board should be asking about its own program. An organization that measures adoption at the individual level has moved past experimentation. Most healthcare organizations cannot say what percentage of their workforce uses the tools they bought.

HOW MUCH TO BELIEVE

A two to one return measured by the company reporting it is a claim rather than a finding. It is also not implausible. The disclosed use cases sit in prior authorization, claims adjudication, coding, and call handling, which are exactly the high volume, rules bound processes where automation returns are easiest to realize and to measure.

What has not been disclosed is the denominator. A cost reduction approaching a billion dollars is meaningful, and set against total operating costs at a company of this size, it does not by itself produce a fifty five percent increase in operating earnings.

The more likely explanation is that the two halves of the model compound. Subtraction removed the least profitable members, products, geographies, and entities, which raises the average margin of everything that remains. Automation then lowered the cost of serving what was left. Either one alone would have produced a good quarter. Doing both at once, in the same year, is what produced this one. That is the part most organizations miss when they try to copy this: the efficiency program and the shrinking are not two initiatives. They are one.

The part of the model nobody writes about

There is a third element to this model that receives almost no attention, because it does not appear as a segment. UnitedHealth does not only insure care and deliver care. It also banks the money moving between the two.

Optum Financial operates a Utah chartered industrial loan company, established ahead of the Medicare Modernization Act of 2003 that created health savings accounts. UnitedHealth became the first company to offer both health savings accounts and the high deductible plans that pair with them.

The scale is not modest. In a February 2026 interview, the senior vice president responsible for payer to provider payments described the business as processing more than $500 billion in healthcare payments annually, connecting roughly 2.5 million providers at the National Provider Identifier level and about 400,000 at the tax identification number level. The second figure is the one to hold onto, since it counts billing organizations rather than individual clinicians and locations. Four hundred thousand billing entities is a payments utility sitting inside a health plan.

Company materials carry somewhat different versions of these figures across pages and are not dated, so treat the magnitude as reliable and the precision as approximate. The order of magnitude is not in question.

READ THE PRODUCT DESCRIPTION CAREFULLY

Optum Bank offers working capital loans to health systems and provider organizations. The marketing describes them as helping providers smooth cash flow and cover gaps in receiving reimbursement, with funds delivered as soon as the same day. It also states that offers are informed by the borrower’s claim payment history.

Follow the structure. A provider experiences a cash flow gap created partly by payer adjudication and payment timing. A lender affiliated with one of the largest of those payers offers to close it, priced on claims data the enterprise already holds. Nothing about that is unlawful, and the product may well be useful. It is simply a position no other participant in the system occupies.

AND NOBODY OUTSIDE CAN SIZE IT

Here is what makes this worth a section rather than a footnote. Optum Financial is not a reported segment. Its results are consolidated inside one of the three Optum businesses and never broken out. There is no public line item, no separate margin, and no way for an investor, a regulator, or a provider counterparty to know what it earns.

Economically it is a float business. The bank holds deposits it pays little on and earns prevailing rates on them, and rates have been elevated since 2023. Add payment processing fees, interchange on benefit cards, and interest on working capital lending. That is a rate sensitive earnings stream carrying essentially no medical risk.

WHY THAT PROFILE MATTERS RIGHT NOW

A company deliberately shedding risk bearing members wants exactly this kind of earnings. It does not move with medical trend, it does not depend on a benchmark rate set by CMS, and it does not require winning an annual enrollment period. Whether it contributed materially to the 2026 recovery is unknowable from the outside, and that is the observation. A banking operation large enough to move half a trillion dollars sits inside the fourth largest company in the world and reports no separate results.

This is why the vertical integration question is not going away. In January 2026 the chief executives of five major insurers were questioned in back to back congressional hearings about vertically integrated business models, with breakup legislation raised by explicit analogy to the separation of commercial and investment banking.

For a health system or medical group, the practical point is more immediate. When you negotiate with a payer that also owns a care delivery arm, a pharmacy benefit manager, a claims technology platform, and a bank that lends against your receivables, you are not negotiating with a counterparty. You are negotiating inside an ecosystem, and every term you concede in one venue is visible in the others.

What Wall Street is actually buying

Shares rose more than seven percent on the print. Goldman Sachs called the beat material. UBS described the company as making good progress with more to go. The most cautious major voice was a single Hold.

But what analysts are underwriting is neither the artificial intelligence program nor revenue growth. It is margin recovery in a business that had lost margin, specifically the path back at Optum Health from under three percent toward the eight percent range it once produced. Management raised the Optum Health outlook to at least $2.2 billion, UnitedHealthcare’s operating profit forecast to at least $12 billion, and reaffirmed long term earnings growth of thirteen to sixteen percent.

TWO THINGS THE COVERAGE IS NOT LEADING WITH

Roughly $860 million of the quarter was net favorable prior period medical reserve development. That is a legitimate and routine accounting event. It is also not operations, and it will not repeat on demand. Strip it out and the quarter is still strong and materially less spectacular.

Commercial medical cost trend is still running above eleven percent. The open question for 2027 is whether the guidance floor holds once that trend works fully into estimates. The Medicare Advantage book has been repriced. The commercial book is where the next pressure lands.

The company’s own executives have said the 2027 Medicare Advantage rate increase still falls short of expected medical trend. UnitedHealth is not exempt from the arithmetic driving exits across the sector. It is simply large enough to absorb the repricing and still report a record quarter.

Shrink to grow

The uncomfortable conclusion is that most health plans will have to run some version of this, and most of them will have to run it without the balance sheet that makes it survivable.

The premium factors have moved down and the cost curve has moved up. That arithmetic does not distinguish between a plan with 50 million members and one with 50,000. What differs is the ability to absorb the transition. UnitedHealth could book an $8.3 billion loss, shed 2.8 million members, and remain the fourth largest company in the world. A regional plan doing proportionally the same thing is fighting for its existence.

So the sequencing matters more for everyone else than it did for UnitedHealth. The realistic path is a deliberate rather than forced contraction through 2027 and 2028, with the return to growth positioned for 2028 or 2029, and two things built while smaller that make the growth possible when it arrives.

THE TWO THINGS TO BUILD WHILE YOU ARE SMALLER

Efficiency that survives the recovery. Cost taken out under pressure comes back when pressure lifts, unless the work itself was redesigned. The difference between a hiring freeze and an operating model change shows up in the first good year, not the bad one.

Contract rates set from a defensible position. A smaller, denser footprint where you hold real provider alignment negotiates better than a broad thin one. Concentration is leverage. The plans that emerge with better unit economics will be the ones that used the contraction to renegotiate rather than merely survive it.

WHAT DOES NOT TRANSFER

Three parts of this model are not available to most organizations.

The write off capacity. Very few organizations can take an eight billion dollar loss as a strategic decision rather than an existential one.
The conversion path. Shedding risk membership into a growing self funded book requires an administrative services business at scale. Most plans do not have one to grow into.
The adjacent economics. Care delivery, pharmacy benefit management, claims technology, and banking each carry margin that a standalone insurer does not have access to when the insurance line is compressed.

The line worth holding onto

The industry spent two decades treating membership as the objective and scale as the strategy. UnitedHealth spent 2025 and 2026 demonstrating that at sufficient size, membership is a variable and focus is the strategy. Revenue flat, earnings up fifty five percent, 2.8 million fewer members, and one fewer continent.

Every plan reading this will be asked by its board why it cannot do the same thing. The honest answer is that most of it is available and the hardest part is not. Anyone can shrink. Very few can shrink deliberately, on a schedule, while rebuilding the cost structure underneath, and arrive on the other side with better economics than they started with. That is the model. The subtraction is the visible part and the least difficult.

Sources

Financial results. UnitedHealth Group second quarter 2026 results and Form 8-K, July 16, 2026; first quarter 2026 results and Form 8-K; fourth quarter and full year 2025 segment reporting; second quarter 2026 earnings call transcript; CNBC; Benzinga; 24/7 Wall St.; Investing.com; MarketBeat.
Membership guidance. UnitedHealth Group fourth quarter 2025 and first quarter 2026 earnings calls; Becker’s Hospital Review and Becker’s Payer Issues coverage of both; Fierce Healthcare.
Latin America divestitures. Reuters; CNBC; Becker’s Payer Issues; Yahoo Finance reporting on the Banmedica transaction and the aggregate net loss on the South American exit.
Corporate structure. UnitedHealth Group annual report subsidiary exhibits; the Sunlight Report published July 16, 2025 by the Center for Health and Democracy with funding from Arnold Ventures, compiled from state Schedule Y filings; Becker’s Payer Issues; The American Prospect.
Artificial intelligence program. UnitedHealth Group first and second quarter 2026 earnings calls; Modern Healthcare; STAT; Healthcare Finance News; The National CIO Review; company statements regarding the OptumInsight transformation.
Payments and banking. Optum Financial and Optum Bank published product and partner materials, which are largely undated and carry some variation in stated figures across pages; interview with Brian Andrews, Senior Vice President of payer to provider payments at Optum Financial, published February 2026; The American Prospect reporting on industrial loan company charters in healthcare; contemporaneous coverage of the January 2026 congressional hearings on vertical integration.
Scale. Fortune Global 500, released July 28, 2026; Fortune 500, released June 3, 2026.

Figures reflect company disclosures and published reporting available at the time of writing. Guidance is forward looking and subject to revision. Statements regarding disclosed subsidiary counts reflect filing requirements governed by materiality thresholds and do not by themselves establish a change in underlying corporate structure. This article is analysis provided for general informational purposes and is not investment, legal, or actuarial advice.

ALSO IN THE STANDARD

Article 10. Medicare Advantage Mass Exit for 2027. The cost curve moved up as the premium factors moved down.

Article 12. Centene’s turnaround, powered by Sarah London.

Article 13. The Health Plan AI Maturity Model survey and findings.

HLTHWORKS builds artificial intelligence governance, scalability, sustainability, oversight, audit readiness, and a mature enterprise solution for medical groups, health systems, health plans, payers, and the vendors, plus investors serving them.