UnitedHealth Group UNH

Reference table

1. What the business is, and the disclosure problem

UnitedHealth operates through four reportable segments — UnitedHealthcare (benefit plans), Optum Health (care delivery), Optum Insight (data, analytics and administration) and Optum Rx (pharmacy care services). Chief executive Stephen J. Hemsley; approximately 390,000 employees, the largest workforce in this batch by a factor of six. CIK 0000731766; NYSE-listed since 1984.

A structural note that shapes everything below: the filings in this archive are PROSE ONLY. The manifest records "tables": false for UnitedHealth, and inspection confirms it — the 10-K's results-summary table, its segment table and its medical-care-ratio series were not extracted, leaving only the narrative around them. Where other names in this batch supply a filed segment grid that settles the vendor's numbers, UnitedHealth supplies the explanatory paragraphs and not the figures. We therefore rely on the vendor's income statement for the quantitative series and on the filings for causation, and we label which is which throughout.

The vendor's segment block is unusable for the most recent year. seg_prod for FY2025 reads {"Optumhealth": 19,839,000,000, "Unitedhealthcare": 332,390,000,000} — summing to $352.229B against reported revenue of $447.567B, omitting Optum Insight and Optum Rx entirely, and showing Optum Health at $19.839B against $105.358B the previous year. That is not a business that shrank 81%; it is a broken extraction. The FY2024 block is structurally complete (UnitedHealthcare $298.208B, Optum Rx $133.231B, Optum Health $105.358B, Optum Insight $18.757B, intersegment eliminations −$150.887B) but sums to $404.667B against reported revenue of $400.278B — a 1.1% overshoot, so even the good year does not tie exactly. FY2023 is missing from the block altogether, and seg_geo is an EMPTY ARRAY.

We therefore quote FY2024 segment revenue as indicative and quote no FY2025 segment figure at all. What the FY2024 structure does establish, and it matters, is the scale of the intersegment elimination: $150.887B, or 37.7% of consolidated revenue, is business the Optum units do for UnitedHealthcare. That single number is the flywheel the knowledge base describes, quantified — more than a third of the company's gross activity is one part of UnitedHealth selling to another.

2. What actually happened — the medical care ratio, in one table

This is the most important section in the dive and the numbers are stark.

Cost of revenue rose 600 basis points as a share of revenue in two years. On FY2025's revenue base, each 100 basis points is $4.48 billion of pre-tax income. Six hundred basis points is roughly $26.9 billion — which is why net income fell $10.3 billion while revenue rose $76 billion.

The 10-K's own explanation, quoted in full because it is the most information-dense sentence in the file:

> "The MCR increased as a result of the revenue effects of the Medicare funding reductions, elevated medical cost trend, the member profile of newly added patients under value-based care arrangements, the acceleration of anticipated future losses in 2026 related to certain Optum Health value-based care contracts, decreased favorable development, the impacts of the IRA on Medicare Part D and the impacts of market morbidity changes on our individual exchange offerings, partially offset by the incremental medical costs for accommodations made to care providers in 2024 as a result of the Change Healthcare cyberattack."

Seven distinct causes, of which one is worth isolating: the acceleration of anticipated 2026 losses into the 2025 result. That is a management decision to recognise known future losses early — a conservative choice that makes FY2025 look worse and FY2026 look better, and one that materially supports the recovery consensus is modelling.

The quarterly series shows the shape of the break and the repair:

Read the operating-margin column: 8.0%, 8.6%, 7.7%, 8.3%, then 4.6%, 3.8%, 0.34%, then 8.0%, 7.1%. The break was three quarters long and the recovery has now delivered two consecutive quarters back inside the historical band. The December 2025 quarter — $380 million of operating income and a $720 million pre-tax loss on $113.2 billion of revenue — is the single most extreme quarterly figure in this entire batch.

And the March 2026 10-Q supplies the causation for the recovery in the company's own words:

> "The MCR decreased due to increased favorable reserve development, affordability initiatives and pricing trends, partially offset by expected elevated medical costs trend."

We flag one component of that as requiring care: "increased favorable reserve development" means claims from prior periods settling for less than reserved. That is real money and it is also, by construction, non-repeating in the same form — a company that reserved too heavily in a bad quarter releases it in a good one. It should not be capitalised as run-rate margin, and the recovery is stronger if pricing and affordability are doing more of the work than reserve development is. The filing does not decompose it.

3. Cash flow, capital returns and the balance sheet

Cash generation held up far better than earnings, and that is the strongest single argument for the security. Free cash flow fell from $25.682B (FY2023) to $16.075B (FY2025) — down 37% — while net income fell 46%. On a trailing basis it has already recovered to roughly $23.66 billion, corroborated two ways from the vendor block: freeCashFlowPerShareTTM of $26.067 on ~908.1M shares gives $23.67B, and freeCashFlowYieldTTM of 6.378% on a $370.26B market capitalisation gives $23.62B. A 6.38% free-cash-flow yield on a business trading at 26.2x GAAP earnings is the clearest statement available that the reported earnings are understating the cash.

incomeQualityTTM of 1.81 — operating cash flow at 1.81x net income — reflects exactly that: depreciation, amortisation and reserve movements that suppress accounting earnings without consuming cash.

Dividend, and a vendor labelling issue. The 10-K states: "In June 2025, our Board of Directors increased the Company's quarterly cash dividend to shareholders to an annual rate of $8.84 compared to $8.40 per share, which the Company had paid since June 2024." The vendor reports lastDividend and dividendPerShareTTM of $8.95, which is ABOVE the last filing-confirmed annual rate. The most plausible explanation is a further increase in June 2026, after the most recent 10-Q in this archive (filed 2026-05-05) and not covered by any filing here. We use the filing-confirmed $8.84 rate for the yield (2.17%) and flag the $8.95 figure as a trailing blend implying an unverified mid-2026 increase. Dividend cost is roughly $8.0B a year against $23.66B of trailing free cash flow — 2.96x covered.

Buyback halved. Repurchases fell from $9.000B (FY2024) to $5.545B (FY2025) — 1.50% of market capitalisation — while free cash flow was $16.075B and the dividend cost $8.0B. That is a company conserving capital through a shock, which is correct, and it leaves roughly $10 billion a year of unallocated capacity as earnings recover. Total shareholder yield is currently about 3.67%.

The balance sheet, from the FY2025 10-K:

Total debt rose $10.95B — 16.2% — between FY2023 and FY2025 while net income halved. Net debt to trailing EBITDA (on the vendor's $25.76B trailing figure) is roughly 1.95x. Interest coverage is 5.57x, the lowest in this batch. Interest expense was $4.002B in FY2025 against operating income of $18.964B — 21% of operating profit went to lenders.

Two things a screen cannot see. First, $110.499B of goodwill and $20.474B of intangibles are 42.3% of total assets against $94.110B of shareholders' equity — a large impairment would take a meaningful share of book equity, and the forward-looking statements name "impairment of our goodwill and intangible assets" as a specific risk. Second, enterprise value is not a usable concept here. The $54.251B of long-term investments is the portfolio backing insurance reserves, not excess cash; netting it against debt would be wrong, and not netting it makes UNH look levered against peers who report differently. Combined with a 20.1% unexplained residual in the vendor's EV (see Section 7), we withhold every enterprise-value multiple.

4. Valuation — priced in or room?

At $407.71 (market cap $370.26B, ~908.1M shares):

Estimate coverage is strong — 16-19 analysts on revenue and 18-19 on EPS for FY2026 and FY2027, 13 on FY2028 EPS, 11 on FY2029. The FY2030 EPS row rests on five analysts and is excluded from conclusions.

Per the data contract we tested est.ebitAvg and est.ebitdaAvg and they PASS the arithmetic screens — ebitAvg is below ebitdaAvg in every forward year, no negative rows, no fabrication signature. Two other estimate defects are present, however. sgaExpenseAvg reads exactly 0 in every year including historical ones — a zeroed field, the defect class the contract warns hides trends. And the FY2023 estimate row is badly wrong against actuals: epsAvg of $15.280 and netIncomeAvg of $14.103B against reported FY2023 diluted EPS of $23.86 and net income of $22.381B — a 36% understatement. The FY2024 row, by contrast, reconciles on an adjusted basis ($27.607 against a well-known adjusted figure near $27.66) even though it is 78% above the $15.51 of GAAP EPS. This establishes the basis question definitively: the est block is struck on ADJUSTED earnings, and every forward multiple in this dive is therefore on the adjusted basis and labelled as such.

The single most important context for the forward multiple. FY2027 consensus adjusted EPS is $22.415. UnitedHealth reported $23.86 of GAAP diluted EPS in FY2023 — three years before that. Consensus does not have the company back above its 2023 earnings until FY2028. This is not a growth forecast; it is a recovery forecast, and it is being made about a company whose revenue in the meantime grew from $371.6B to a projected $485.4B. Re-earning a margin you previously held on a 31% larger revenue base is a materially different proposition from growing into a new one, and it is the core of the case.

Peer context. The vendor peer set is half appropriate. Cigna ($72.42B), CVS Health ($133.23B), Elevance ($81.98B) and Humana ($43.48B) are the genuine comparables; Abbott, AstraZeneca, Merck, Novo Nordisk, Novartis and Thermo Fisher are not — they are life-sciences companies grouped by sector code. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The scale fact worth recording: UnitedHealth's $370.26B market capitalisation exceeds the combined $331.11B of Cigna, CVS, Elevance and Humana.

4a. What today's price assumes (the inversion)

At $407.71 — 20.7x FY2026E and 18.2x FY2027E adjusted — the price embeds the following falsifiable claims:

• Adjusted EPS recovers from $18.64 trailing to $19.715 (FY2026), $22.415 (FY2027) and $26.218 (FY2028). (Consensus-derived; 18, 19 and 13 analysts.) That is 15.7% compound growth and it is entirely margin, since consensus revenue grows 2.9% and 5.8% in the same years and DECLINES 0.4% in FY2026. The nearest check is 2026-10-27, where consensus is $4.08 — the same figure delivered a year earlier and 36% below the $6.38 just reported. That is a strikingly low bar.
• The medical care ratio keeps falling and the improvement is not mostly reserve development. (Filing-confirmed that it fell in Q1 2026; our concern about the composition.) This is the most fragile assumption in the price. The 10-Q names "increased favorable reserve development" first among the drivers, and reserve releases are self-limiting.
• The Optum Health value-based-care losses the 10-K pulled forward into 2025 do not recur. (Filing-confirmed that they were accelerated; our assumption that the acceleration was complete.) If the 2026 losses exceed what was pre-recognised, the FY2026 consensus breaks directly.
• The Department of Justice matter costs nothing material. (Filing-confirmed as unresolved and unquantified.) The 10-Q states the company "cannot reasonably estimate the outcome which may result from this matter given its procedural status." No provision is reflected in any figure in this dive, and none can be — but nor is any discount applied by the market at 18.2x.
• The market pays 18-21x forward adjusted earnings. (Our number.) At 15x FY2027E the stock is $336; at 25x it is $560. The multiple band is worth $224 of share price on unchanged estimates — 55% of the current price.

4b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: adjusted EPS growth (+17.1%, from FY2027E $22.415 to FY2028E $26.218) + multiple drift (COMPRESSION, from 18.2x to roughly 17x on the respective forward year, −6.6%) + shareholder yield (+3.67%) ≈ +13% to +15%.

Nearly all of the expected return is earnings recovery, and the multiple is a modest headwind. Our base of $445 assumes mild multiple compression to 19.9x FY2027E from the 20.7x currently paid on FY2026E — not a de-rating thesis, simply that a business whose earnings are recovering rather than growing does not deserve an expanding multiple as the recovery matures.

Sensitivity, stated openly. On the FY2027 consensus of $22.415: 15x gives $336, 18x gives $403 (essentially spot), 20x gives $448, 22x gives $493, 25x gives $560. The street's $473.89 implies 21.1x. The entire disagreement between us and the street is one multiple point.

The bull case at $560 requires multiple expansion to 25x and is therefore the weaker leg of the two extremes — we say so. The bear case at $336 requires no earnings failure at all, only a de-rating to 15x on unchanged consensus, and it lands 2.7% below the 200-day moving average of $345.20.

4c. Variant perception (where we differ, what would surprise)

• We differ from the street on level and we are below it, but the gap is small and the direction is the same. Consensus $473.89 across 52 analysts, of whom 43 are buyers — 83%, the most one-sided distribution in this batch. Our base of $445 is 6.1% below. On a name this consensual, the honest statement is that our edge is not directional.
• Where we do differ is on the composition of the recovery, and it is a specific, checkable disagreement. The Q1 2026 10-Q lists "increased favorable reserve development" FIRST among the drivers of the medical-care-ratio improvement, ahead of affordability initiatives and pricing. Reserve releases are a one-way valve: a company that over-reserved in a catastrophic quarter releases it in the following ones, and the release ends. Consensus is extrapolating 15.7% compound EPS growth from a base that includes an unquantified amount of it. Watchable number: the medical care ratio disclosed in the 2026-10-27 release, and specifically whether the company attributes the improvement to pricing rather than to development.
• We think the acceleration of 2026 Optum Health losses into the 2025 result is under-discussed and is a POSITIVE. The 10-K says explicitly that the FY2025 MCR increase included "the acceleration of anticipated future losses in 2026 related to certain Optum Health value-based care contracts." Losses recognised in 2025 cannot be recognised again in 2026. That is a known headwind removed from the FY2026 consensus base, and it makes the $19.715 estimate more achievable than the raw margin arithmetic suggests.
• We think the knowledge base's most recent claim is already wrong, and we record it rather than ignoring it. On 2026-07-19 — three days after the Q2 print that beat by 29.1% — the lane's only 2026 entry reads: "UNH is 'a dead company' — bearish on the insurers." A five-word thesis published seventy-two hours after a 29% beat, on a stock that has risen 71% in twelve months, is a claim we are prepared to grade against and take the other side of.
• Positive surprise that would force a re-rate: a Q3 print materially above the $4.08 consensus with the medical care ratio improvement attributed to pricing rather than reserve development; a court ruling adopting the Special Master's March 2025 recommendation for summary judgment in the company's favour; or the buyback returning toward the $9B FY2024 level, which at current prices would retire roughly 2.4% of the shares annually.
• Negative surprise that would break the thesis: any quarter in which operating margin falls back below 6%; an adverse ruling or settlement in the False Claims Act matter; a risk-adjustment data validation audit producing a material retrospective adjustment, which the 10-Q names as a live possibility; or a goodwill impairment against the $110.499B balance. Note the asymmetry in that last item: goodwill is 42.3% of total assets and 117% of shareholders' equity.

Synthos fair values

All three anchors are multiples of the FY2027 consensus ADJUSTED EPS of $22.415 (19 analysts — the best-covered forward year), cross-checked against FY2026E and FY2028E.

• Bear ~$336 — 15.0x FY2027E. Cross-checks: 17.0x FY2026E; 12.8x FY2028E; 2.7% below the 200-day average of $345.20; a 2.63% dividend yield. The scenario: reserve development exhausts, the medical care ratio stalls above 80%, the Department of Justice matter turns adverse, and the multiple returns to where distressed managed care trades. −17.6%. Note the free-cash-flow yield at that price would be 7.7%.
• Base ~$445 — 19.9x FY2027E. Cross-checks: 22.6x FY2026E; 17.0x FY2028E; 0.79x FY2027E revenue; a 5.31% free-cash-flow yield on trailing cash generation. Sensitivity: 18x gives $403 — essentially spot — and 22x gives $493. The street's $473.89 implies 21.1x. The scenario: the recovery lands roughly on consensus, the pulled-forward Optum Health losses do not recur, the legal matters resolve without material cost, and the multiple compresses mildly as the recovery matures. +9.1%, plus 3.67% of shareholder yield.
• Bull ~$560 — 25.0x FY2027E. Cross-check: 21.4x FY2028E. This bull case REQUIRES multiple expansion and that is its weak leg. The scenario: the medical care ratio returns fully to the 75-76% band of 2020-2023, adjusted EPS reaches the FY2028 consensus HIGH of $29.06 a year early, the Special Master's recommendation is adopted, and the buyback returns to $9B. +37.4%.

Base is 9.1% above spot; asymmetry roughly 2.12:1 to the upside (17.6% down, 37.4% up). That is the best ratio in this batch and it is the primary reason for the Buy — Tactical rather than a Watch. What holds the position size down is the unquantifiable legal exposure and a knowledge-base lane with no named speaker in it.

5. Knowledge base — ten claims, seven from one afternoon, and not one named speaker

Raw entity hits: 10. Text matches: 0. Discarded for collision: 0. Independent voice-days: 4.

The search covered UNH, UnitedHealth, UnitedHealth Group, United Health and Optum across the entity field of all 51,928 distilled claims. Stance: 7 bullish, 3 bearish, 0 neutral.

Two structural problems and both are severe.

First, seven of the ten claims — 70% — come from a single business_breakdowns episode dated 2025-07-12, chunked into seven distilled claims. One source, one day, one connected argument about the same company. The lane is therefore four voice-days deep, not ten claims deep.

Second, and more serious: NOT ONE claim in this lane carries a named speaker. All ten are attributed to a channel only — business_breakdowns (7), all_in (1), real_vision (1), jordi_visser (1). Under the 4-lane attribution policy that is the weakest form of sourcing we accept, and it applies to 100% of this lane, which is the only name in this batch where that is true. No speaker_role of management appears, so the half-weight management discount does not arise.

The 2025-07-12 episode, which is the substance of the lane, is genuinely good and it is two-sided within itself.

> bullish · conviction 88 · principle

> "UNH's flywheel — UnitedHealthcare insurance feeding Optum care/data delivery — gives underwriting edge and operational cost control peers can't replicate."

> bullish · conviction 85 · thesis

> "Market repeatedly misprices UNH as just an insurer; regulatory noise and leadership change obscure a mispriced integrated healthcare giant."

> bullish · conviction 83 · thesis

> "Revenue quadrupled from $110B to $400B since 2014 while free cash flow per share compounded 13%, well ahead of benchmarks — durable capital-efficient growth."

> bullish · conviction 82 · thesis

> "Optum, long overlooked, is the underappreciated profit driver and future growth engine; a $100B revenue business that would be Fortune 50 standalone."

> BEARISH · conviction 65 · fact

> "Elevated senior medical utilization and Medicare Advantage scrutiny are pressuring a key growth driver, driving the recent share collapse."

The flywheel claim is corroborated by the one hard number the vendor segment block does supply: $150.887 billion of intersegment eliminations in FY2024, 37.7% of consolidated revenue. More than a third of UnitedHealth's gross activity is Optum selling to UnitedHealthcare. That is the flywheel, measured.

The bearish claim in the same episode identified the exact mechanism that then produced the December 2025 quarter — elevated senior utilisation and Medicare Advantage scrutiny — and it did so six months before the $10-million quarter. A source that states the bull case and the specific bear mechanism in the same sitting is more useful than one that does only either, and we credit it accordingly.

The two independent bear claims are structural rather than cyclical:

> 2025-09-09 · BEARISH · conviction 75 · thesis · channel all_in · no named speaker

> "Vertically-integrated insurers/PBMs game an opaque system and face antitrust risk; separating formularies from PBMs would collapse their rent-extraction model."

> 2026-07-19 · BEARISH · conviction 62 · thesis · channel jordi_visser · speaker_role independent

> "UNH is 'a dead company' — bearish on the insurers."

The first is the serious version of the bear case and it is the mirror image of the flywheel claim: the same vertical integration that produces the cost advantage is the thing a regulator would unwind, and Optum Rx was $133.231B of gross revenue in FY2024. The second is a five-word assertion published three days after a 29.1% earnings beat, on a stock that has since risen. We record it, we grade it, and we take the other side.

One older claim is worth noting for its framing rather than its content:

> 2023-03-22 · bullish · conviction 75 · thesis · channel real_vision · no named speaker

> "Regulated industries (healthcare, education) that adopt AI cut costs while keeping pricing power — margins go insane; UnitedHealth saves billions yet maintains pricing."

That claim has aged badly — margins did the opposite over the following two years, falling 450 basis points at the operating line — and the 10-Q now names "risks and uncertainties associated with our increasing use of artificial intelligence and other emerging technologies" as a forward-looking risk factor rather than an opportunity. We record the miss.

Conclusion. The knowledge base has a positively-signed but structurally thin view of UnitedHealth: one good episode, two structural bear claims, one stale bull claim that has been falsified, and one very recent bear claim that appears to be wrong already. Zero named speakers anywhere. We use the flywheel argument to support the exponential score and the antitrust claim in the variant perception. It does not move the fair value, and its thinness is a direct input to the small position size.

6. Technicals

• Price $407.71. −6.56% from the 52-week high of $436.35; +69.19% above the 52-week low of $240.98. Position within the annual range: 85th percentile.
• But the drawdown that matters is longer: max_dd_from_peak reads −34.79% against a pct_from_hi of −6.56%. The two differ because the peak is a multi-year high, not a 52-week one — this stock is still 34.8% below where it traded before the shock, which implies a prior peak near $625. That is not a defect; it is the correct reading of a security that halved and has recovered most but not all of the way.
• Moving averages, split: −0.9% below a 50-day average of $411.31, which has flattened; +18.1% above a 200-day average of $345.20, which is rising steeply.
• RSI 43.4 — the weakest reading in this batch and consistent with a stock consolidating a very large move. MACD +0.69 — marginally positive.
• Relative performance: 3-month +9.97% vs SPY +7.59% and QQQ +7.67%; 6-month +42.76% vs SPY +11.09% and QQQ +15.71%; 12-month +71.47% vs SPY +24.26% and QQQ +30.80%. A 47-point outperformance over twelve months on a 0.633-beta stock is the signature of a recovery, not a trend.
• Sentiment: 43 buy, 7 hold, 2 sell across 52 analysts — 82.7% buyers, the most one-sided distribution of the twelve names here. Consensus target $473.89 (+16.2%), median $491 (+20.4%), high $529 (+29.7%), low $373 (−8.5%).

Today's move

UNH closed 2026-08-04 at $407.71, down 1.84% or $7.65 from a $415.36 previous close. It opened at $415.555, traded $406.67 to $416.00, and closed within 0.3% of the day's LOW on 5.25M shares against a 6.70M average — light volume, weak close. No company-specific news is in this file for the date; the last company event was the 2026-07-16 second-quarter release, nineteen days earlier, which beat adjusted EPS consensus by 29.1%. The shares are lower than they were the day after that beat. That is a stock digesting a very large recovery, and it is a better entry than the same stock 6% higher would have been.

7. Data integrity — what we rejected from the vendor file and why

Ten findings. UnitedHealth's file has the most damaged segment data in this batch, one enterprise-value residual large enough to trigger the contract's withholding rule, and a quarterly gross-margin series that is simply wrong.

1. seg_prod FY2025 omits two of four segments and understates revenue by $95.3B — REJECTED. The block reads {"Optumhealth": 19,839,000,000, "Unitedhealthcare": 332,390,000,000}, summing to $352.229B against reported revenue of $447.567B — a 21.3% shortfall. Optum Insight and Optum Rx are absent entirely, and Optum Health at $19.839B is 81% below its own FY2024 figure of $105.358B. No FY2025 segment figure appears anywhere in this dive.

2. seg_prod FY2024 does not tie either, and FY2023 is missing — flagged, used as indicative only. FY2024's five lines (UnitedHealthcare $298.208B, Optum Rx $133.231B, Optum Health $105.358B, Optum Insight $18.757B, eliminations −$150.887B) sum to $404.667B against reported revenue of $400.278B — a 1.1% overshoot. FY2023 is absent from the block altogether. We quote the FY2024 structure to establish the shape of the business and the $150.887B elimination, and draw no precise conclusion from it.

3. seg_geo is an EMPTY ARRAY. There is no geographic disclosure of any kind in the vendor file, and the filings in this archive are prose-only so none can be recovered. We state the gap rather than estimating around it.

4. enterpriseValueTTM carries a 20.1% unexplained residual — WITHHELD per the data contract. The vendor's $412.120B less market capitalisation of $370.260B implies net debt of $41.860B. The most recent filed balance sheet (2025-12-31) shows total debt of $78.389B less cash of $24.365B and short-term investments of $3.756B — net debt of $50.268B. The residual is $8.41B, or 20.1% of the vendor's implied figure, above the 15% threshold at which the renderer withholds the multiple. No June-2026 balance sheet exists in either the vendor file or the archive to reconcile it. We quote no EV multiple in this dive.

5. Enterprise value is conceptually inappropriate for a health insurer regardless — stated as a second, independent reason. The $54.251B of long-term investments is a regulated investment portfolio backing insurance reserves, not distributable cash. Netting it against debt overstates liquidity; ignoring it makes the leverage look worse than peers who classify differently. evToSalesTTM (0.916), evToEBITDATTM (16.00), netDebtToEBITDATTM (1.625), evToOperatingCashFlowTTM (15.25) and evToFreeCashFlowTTM (17.45) are all rejected as lenses.

6. netDebt omits $3.756B of short-term investments — corrected. The vendor's $54.024B is total debt less cash and equivalents only. Corrected: $50.268B. The same omission appears in the FY2024 and FY2023 rows.

7. The quarterly gross-margin series is internally impossible — REJECTED. costOfRevenue for the June 2026 quarter reads $75.358B on $112.032B of revenue (67.3%), against $86.312B on $111.721B (77.3%) in the immediately preceding quarter and $91.604B on $111.616B (82.1%) in the year-ago quarter. A 1,000-basis-point sequential improvement in cost of revenue at a health insurer does not happen, and the operating income figures ($7.991B against $8.990B) move the opposite way. The quarterly gross-profit and gross-margin lines are discarded; the revenue, operating income, pre-tax income, net income and EPS lines are used and reconcile to the earnings calendar. ratios_ttm.grossProfitMarginTTM of 22.5% inherits the same contamination.

8. earn_cal reports the SAME revenue figure for two different quarters — flagged. Both the 2026-04-21 and 2026-07-16 entries read revenueActual: 112,032,000,000. The income statement shows $111.721B for the March quarter and $112.032B for June. The March-quarter revenue in earn_cal is a duplicate of June's and is wrong by $311M. The EPS actuals ($7.23 and $6.38) are distinct and appear correct.

9. The FY2023 estimate row is 36% below actuals and sgaExpenseAvg is zero in every year — flagged; basis established. FY2023 epsAvg of $15.280 and netIncomeAvg of $14.103B against reported diluted EPS of $23.86 and net income of $22.381B. By contrast the FY2024 row ($27.607 epsAvg) reconciles on an ADJUSTED basis while sitting 78% above the $15.51 of GAAP EPS. This establishes definitively that the est block is struck on adjusted earnings, and every forward multiple in this dive is labelled accordingly. sgaExpenseAvg reads exactly 0 in all eight years — a zeroed field, reported as such. est.ebitAvg and est.ebitdaAvg PASS the arithmetic screens in every year and no defect signature is present, but they are not used because the EV multiples they would feed are withheld.

10. lastDividend: 8.95 exceeds the last filing-confirmed annual rate — flagged as unverifiable. The 10-K states the Board raised the annual rate to $8.84 in June 2025 from $8.40. The vendor's $8.95 is above that and implies a further increase around June 2026 that no filing in this archive covers — the most recent 10-Q was filed 2026-05-05 and the three 8-Ks in the archive are cover pages only. We use $8.84 for the forward yield (2.17%) and flag the $8.95 figure.

11. quote.yearHigh/yearLow disagree materially with tech.hi52/lo52 — we use tech. The quote block reports $461.62 / $239.50; the computed block reports $436.35 / $240.98. The high discrepancy is 5.8% — the largest in this batch — and it changes the drawdown read from −11.7% to −6.6%. We use tech throughout. Separately, tech.max_dd_from_peak of −34.79% correctly differs from pct_from_hi of −6.56% because peak is a multi-year maximum; per the data contract this is NOT a defect and it is the more informative number here.

Not defects, correctly reported and independently confirmed: the implied share count of ~908.1M is consistent with the 906.0M diluted weighted average with no share-class complication; the quarterly revenue, operating income and EPS series reconcile to the earn_cal actuals on the adjusted basis; freeCashFlowPerShareTTM ($26.067) and freeCashFlowYieldTTM (6.378%) cross-check to the same $23.6-23.7B figure from two directions; incomeQualityTTM of 1.81 correctly reflects operating cash flow well above net income; and effectiveTaxRateTTM of 14.5% is consistent with the disclosed effect of tax benefits on depressed pre-tax income.

Vendor rating — one artefact, two correct signals. The composite reads B / 3 with debtToEquityScore of 1. Total debt of $78.389B against $94.110B of shareholders' equity is a 0.83x ratio, which is unremarkable for a regulated insurer and the sub-score of 1 is harsh — but unlike the cases where such a score traced to an undefined input, here the input is defined and the score is merely aggressive. priceToEarningsScore of 2 and priceToBookScore of 2 are reasonable at 26.2x GAAP earnings and 3.53x book.

Non-equity tripwire — checked and passed. UNH is common stock, NYSE-listed, one class. Beta 0.633; a 52-week range of $240.98 to $436.35 — an 81% spread, the second-widest in this batch; a variable, rising dividend; volume of 5.25M shares (roughly $2.14B of turnover). This is common equity.

8. Insiders — routine director grants only

Every insider transaction in this file was filed 2026-07-06 for transactions dated 2026-07-01, and every one is coded A-Award at a price of $0.

The reading. This is the quarterly non-employee director equity grant, awarded on the first day of the quarter at a zero cost basis, in small and unequal amounts consistent with a fixed dollar value divided by the prevailing price. It is a calendar event, not a decision.

What the file contains is therefore: zero open-market purchases, zero open-market sales, and no officer transactions of any kind. Neutral, and reported as neutral. For a company that has just reported a 29% earnings beat after a period in which its stock halved, the absence of any executive purchase is a mild disappointment and no more than that. The residual holdings are small — the largest director position shown is 15,462 shares, roughly $6.3M — which is normal for outside directors and carries no information.

9. Verdict, kill-criteria and flip conditions

Buy — Tactical.

What makes this actionable: an earnings base that is still 40% below its own 2023 level with consensus not recovering it until FY2028, so the case requires re-earning rather than growing; a documented, filing-verified recovery in progress — quarterly operating income of $380M → $8,990M → $7,991M, a 10-Q statement that the medical care ratio decreased, and a June quarter beating adjusted EPS consensus by 29.1%; roughly $23.66 billion of trailing free cash flow, a 6.38% yield, covering the dividend 2.96 times with roughly $10 billion a year of unallocated capacity; a 2.12:1 bull-to-bear asymmetry, the best in this batch; a 0.633 beta; and a specific, filing-disclosed positive nobody discusses — the FY2025 result already absorbed "the acceleration of anticipated future losses in 2026 related to certain Optum Health value-based care contracts."

What holds the size down: an unresolved False Claims Act action the Department of Justice has pursued since February 2017 alleging improper Medicare risk-adjustment submissions, on which the company states it "cannot reasonably estimate the outcome"; a 10-Q listing fourteen named federal agencies conducting investigations, audits or reviews, with risk adjustment data validation audits explicitly capable of "retrospective adjustments to payments"; $110.499 billion of goodwill — 42.3% of total assets and 117% of shareholders' equity; interest expense consuming 21% of FY2025 operating income; a recovery whose first-named driver is reserve development, which is self-limiting; and a knowledge-base lane of ten claims, seven from one day, with not a single named speaker.

The distinction that matters. We are not claiming an informational edge on UnitedHealth — 83% of the street is already positive and our fair value sits below theirs. We are claiming that the risk/reward is unusually shaped: a 17.6% bear case that requires only a de-rating, against a 37.4% bull case, on a business whose cash generation never broke even when its accounting earnings did. That is a position worth taking in a small size, not a conviction holding. The verdict is Buy — Tactical precisely because the qualifier carries information: buy it, keep it small, and let the falsifiers below decide whether it becomes more.

Pre-registered KILL criteria — what would take this to Avoid:

• Operating margin below 6% in any quarter. The recovery to 8.0% and 7.1% is the whole thesis; a relapse falsifies it directly.
• An adverse ruling or settlement in the Department of Justice False Claims Act matter, or the court rejecting the Special Master's March 2025 recommendation.
• A material retrospective adjustment from a risk adjustment data validation audit, which the 10-Q names as a live possibility.
• A goodwill impairment against the $110.499B balance — 117% of shareholders' equity sits in goodwill and intangibles.
• The Q3 print (2026-10-27) missing a $4.08 consensus that is already 36% below the quarter just delivered and equal to the year-ago figure. A miss against that bar would be severe.
• The medical care ratio improvement being disclosed as predominantly reserve development rather than pricing, which would make the FY2027 consensus of $22.415 unreachable.

Pre-registered FLIP conditions — what would take this to Buy — Core:

• A court order adopting the Special Master's recommendation and disposing of the False Claims Act matter. That single event removes the largest unquantifiable in the file and would justify a materially larger position.
• Two consecutive quarters with operating margin above 8% and the improvement attributed in the filing to pricing and affordability rather than reserve development.
• The buyback returning toward $9 billion a year, which at current prices retires roughly 2.4% of the shares annually and converts a recovery story into a compounding one.
• A reclaim of the 50-day average at $411.31 followed by a break above the 52-week high of $436.35 with the fundamental leg intact.
• Genuine named-speaker knowledge-base conviction appearing. At present the entire lane is channel-attributed and 70% of it is one episode; that alone caps the position size regardless of the arithmetic.

Where UNH fits in the Synthos Framework Portfolio. The healthcare / defensive-value sleeve, at 0% today with a 1.5% initial position appropriate to the legal uncertainty, and a 3% target on either of the first two flip conditions. On batch overlap: UnitedHealth and Merck are the two healthcare names here and they are not the same exposure — Merck's risk is patent expiry and pipeline, UnitedHealth's is reimbursement policy and litigation. Neither correlates meaningfully with the eight technology and financial names in this batch, and at a 0.633 beta UNH is one of three genuine diversifiers alongside Coca-Cola and Chevron. Logged as a tracked Synthos call (Buy — Tactical) as of 2026-08-04 at $407.71.

Single biggest risk: the Department of Justice False Claims Act action. Filed under seal by a whistleblower in 2011, unsealed and pursued by the Department from February 2017, alleging "the Company made improper risk adjustment submissions and violated the False Claims Act." In March 2025 a court-appointed Special Master recommended summary judgment in the company's favour on all remaining claims; in April 2025 the Department moved to reject that report. The 10-Q states the company "cannot reasonably estimate the outcome which may result from this matter given its procedural status." False Claims Act damages are trebled. No figure is disclosed, none can be estimated, and no discount for it is visible in a stock trading at 18.2x forward earnings with 83% of analysts positive. This is the definition of an unpriced binary, and it is the reason the position is Tactical and small rather than Core.

Most fragile assumption in the price: that the medical-care-ratio improvement is pricing rather than reserve development. Consensus needs adjusted EPS to compound 15.7% a year from FY2026 to FY2029 on revenue growing 2.9-5.8%, which is entirely a margin assumption. The Q1 2026 10-Q lists "increased favorable reserve development" FIRST among the drivers of the improvement. Reserve releases are the return of over-reserving from the catastrophic quarters, and they end. If half the improvement is development, FY2027 adjusted EPS is nearer $20 than $22.42, and at 18x that is $360 — 12% below today's price. That is the specific way an investor loses money here while every headline number continues to improve.

Provenance & disclosures

• Traceability: 10 tagged knowledge-base claims name UnitedHealth by entity (raw entity hits 10, text matches 0, discarded 0). Two attribution findings, both severe: (1) seven of the ten claims — 70% — come from a SINGLE business_breakdowns episode dated 2025-07-12, chunked into seven distilled claims, so the lane is four voice-days deep rather than ten claims deep; (2) NOT ONE claim in the entire lane carries a named speaker — all ten are channel-attributed only (business_breakdowns 7, all_in 1, real_vision 1, jordi_visser 1), which is the weakest sourcing the 4-lane attribution policy admits and applies to 100% of this lane, the only name in this batch where that is true. No management speaker_role appears. Stance 7 bullish / 3 bearish / 0 neutral. Breadth 4, net conviction positive-low. Two claims are graded against outcomes in the body: a 2023-03-22 claim that regulated healthcare adopting artificial intelligence would see "margins go insane" is FALSIFIED (operating margin fell 450 basis points over the following two years and the 10-Q now lists the same technology as a risk factor); and a 2026-07-19 claim that "UNH is 'a dead company'", published three days after a 29.1% earnings beat, is recorded and explicitly disagreed with. All quotes verbatim from stored claim text.
• Data as-of: income statement, balance sheet and cash flow for fiscal 2025 ended 2025-12-31, from the 10-K filed 2026-03-02 · most recent filing-verified financials 2026-03-31, from the 10-Q filed 2026-05-05 · June-2026 quarter figures from the vendor payload, cross-checked against the earn_cal entry for 2026-07-16 (adjusted EPS $6.38 against a $4.94 estimate; revenue $112.032B) · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873605 = 2026-08-04T20:00:05Z ($407.71, −1.84%; 50-DMA $411.31; 200-DMA $345.20; RSI 43.4; MACD +0.69) · knowledge-base claims 2026-08-04. UnitedHealth's fiscal year is the calendar year. All figures come from the Synthos vendor data file for UNH or from the SEC filings in the UNH archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents and its gaps — unusually restrictive on this name. 10-K filed 2026-03-02 (fiscal 2025); 10-Q filed 2025-10-28 (September 2025 quarter); 10-Q filed 2026-05-05 (March 2026 quarter); 8-Ks filed 2026-05-11, 2026-06-05 and 2026-07-16. The manifest records "tables": false for UnitedHealth and inspection confirms it — no financial-statement tables were preserved in ANY document, so the 10-K's results summary, segment grid and medical-care-ratio series exist only as the narrative around them. All three 8-Ks are cover pages with no financial content, including the 2026-07-16 filing that carried the second-quarter release. There is NO 10-Q for the June 2026 quarter. Consequently: the June-quarter income statement comes from the vendor payload cross-checked against the earnings calendar; the most recent filing-verified balance sheet is 2025-12-31; and no segment figure for FY2025 is available from any source in this file.
• Where the filings contradicted or corrected the vendor (detailed in Section 7): the FY2025 seg_prod block rejected outright for omitting Optum Insight and Optum Rx and understating revenue by $95.3B (21.3%), with the FY2024 block flagged as overshooting by 1.1% and FY2023 missing entirely; seg_geo recorded as an empty array with no substitute available because the filings are prose-only; enterpriseValueTTM WITHHELD under the data contract's 15% rule — the implied net debt of $41.860B carries a 20.1% unexplained residual against the filed $50.268B, and enterprise value is separately inappropriate for an insurer whose $54.251B investment portfolio backs reserves; netDebt corrected from $54.024B to $50.268B for the omission of $3.756B of short-term investments; the quarterly gross-profit and gross-margin series rejected as internally impossible (cost of revenue at 67.3% of revenue in the June quarter against 77.3% in March and 82.1% a year earlier, while operating income moved the opposite way); earn_cal found to report the same revenue figure for two different quarters ($112.032B for both March and June); the FY2023 estimate row found 36% below actuals while the FY2024 row reconciles on an adjusted basis, establishing that the est block is struck on ADJUSTED earnings; sgaExpenseAvg recorded as exactly zero in all eight years; lastDividend of $8.95 flagged as exceeding the last filing-confirmed annual rate of $8.84 (10-K, June 2025) and implying an unverified mid-2026 increase; and quote.yearHigh/yearLow ($461.62/$239.50) rejected in favour of tech.hi52/lo52 ($436.35/$240.98), a 5.8% high-side discrepancy — the largest in this batch. tech.max_dd_from_peak of −34.79% against pct_from_hi of −6.56% is NOT a defect — the peak is a multi-year maximum and is the more informative figure. Where vendor and filing AGREED — worth recording: the implied share count is consistent with the diluted weighted average; the annual income statement matches the narrative figures the 10-K does supply; and the free-cash-flow figure cross-checks from two independent vendor fields.
• Basis note — established from the estimate block itself. The FY2024 est row (epsAvg $27.607) sits 78% above the FY2024 GAAP diluted EPS of $15.51 but reconciles to a well-known adjusted figure, while the FY2023 row is 36% below GAAP actuals. Consensus is therefore struck on ADJUSTED earnings. Trailing multiples are shown on both bases (26.2x GAAP on $15.55, 21.9x adjusted on $18.64); all forward multiples use consensus epsAvg and are adjusted-basis. The GAAP-to-adjusted gap is unusually wide on this name — $2.10 per share in the December 2025 quarter alone, where GAAP diluted EPS was $0.011 and adjusted was $2.11 — and no conclusion in this dive rests on the GAAP figure for that quarter.
• Estimate coverage: 16 analysts on FY2026 revenue and 18 on FY2026 EPS; 19 and 19 on FY2027; 19 and 13 on FY2028; 10 and 11 on FY2029. The FY2030 EPS row rests on five analysts and is excluded from every conclusion. est.ebitAvg and est.ebitdaAvg pass the arithmetic defect screens but are unused because the enterprise-value multiples they would feed are withheld. sgaExpenseAvg is zero in every year.
• Peer note: the vendor peer set is half appropriate. Cigna ($72.42B), CVS Health ($133.23B), Elevance ($81.98B) and Humana ($43.48B) are genuine comparables; Abbott, AstraZeneca, Merck, Novo Nordisk, Novartis and Thermo Fisher are life-sciences companies grouped by sector code and are not. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. UnitedHealth's $370.26B market capitalisation exceeds the combined $331.11B of the four genuine peers.
• Fair-value caveat: the $336 / $445 / $560 anchors are multiples of the FY2027 consensus ADJUSTED EPS of $22.415 — 15.0x, 19.9x and 25.0x. Stated arithmetic, not a discounted cash flow. Sensitivity: 18x gives $403 (essentially spot), 22x gives $493, and the street's $473.89 implies 21.1x — the entire disagreement between us and the street is one multiple point. The bear case requires no earnings failure, only a de-rating; the bull case REQUIRES multiple expansion and that is named as its weak leg. No provision or discount for the Department of Justice False Claims Act matter is reflected in any of the three cases, because the company states it "cannot reasonably estimate the outcome" and neither can we. False Claims Act damages are trebled. This is an unpriced binary and it is the reason the verdict is Tactical and small rather than Core.
• Timing: second-quarter 2026 results were released 2026-07-16, nineteen days before this dive, beating consensus adjusted EPS by 29.1% ($6.38 against $4.94). The next print is 2026-10-27, 84 days away, with consensus of $4.08 adjusted EPS on $111.311B of revenue — a figure equal to the year-ago quarter and 36% below the one just delivered. There is no other near-term company-specific catalyst; the legal calendar is the live variable.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.