Damage Math

The Bottom Line

Oscar's 2025 problem cost $646 million of guided operating earnings in a single year, about 85% of it a step-up in risk-adjustment transfers from 14.5% to 18.5% of direct premiums. At the March 2026 trough the market took out $12.42 a share against $3.28 to $5.72 of plausible NPV damage — a gap of roughly $2.0 billion to $2.7 billion. The judges put the probability the impairment is temporary at 0.71. At $28.34 that gap has closed.

The Near-Term Hit

The trigger has a clean paper trail. On 4 February 2025, reporting FY2024, Oscar introduced a 2025 outlook of $11.2 billion to $11.3 billion of revenue, a medical loss ratio of 80.7% to 81.7%, and earnings from operations of $225 million to $275 million [1]. It reaffirmed that outlook across every metric on 7 May 2025 [2].

On 22 July 2025 the company pre-announced preliminary second-quarter results and revised the full-year outlook [3]. The revised guide: revenue of $12.0 billion to $12.2 billion — higher — and a loss from operations of $200 million to $300 million [4]. Revenue guidance went up $850 million at the midpoint while the earnings line moved $500 million the other way. The problem was margin, not volume.

The year finished worse than the revision. FY2025 total revenue was $11,701.4 million, the loss from operations $396.4 million, and diluted loss per share $1.69 [5], on a medical loss ratio of 87.4% [6]. The fourth quarter alone carried a 95.4% MLR and a $333.7 million operating loss [7].

Sources: FY2025 outlook [8]; July 2025 revision [9]; FY2025 actual loss from operations of $396.4 million [10]; FY2026 outlook of $250 million to $450 million of earnings from operations [11].

The arithmetic of the numerator: guided FY2025 operating earnings of +$250 million at the midpoint against a reported −$396.4 million is a $646.4 million swing in one fiscal year — roughly 5.5% of that year's revenue.

The longer-dated hit is measurable too. At its June 2024 Investor Day, management set out a 2027 target of at least 20% revenue CAGR, roughly 5% operating margin, and EPS of $2.25 or better [12]. Asked in August 2025 whether those targets survived, Bertolini said the forecast was not changing and "5% is still our target," while conceding a revision might follow the pricing season [13]. Consensus now carries FY2027 normalised EPS of $1.51 and a 2.69% operating margin, reaching $2.21 only in FY2028 — the target level arrives about a year late, on a share count 21% larger.

What the Price Did

Over the same window the equity fell further than the earnings line, and then more than recovered.

Sources: closes from the daily price series, as reported; share counts are Class A plus Class B outstanding at the nearest filed balance-sheet date — 30 Sep 2024 [14], 30 Jun 2025 [15], 31 Mar 2026 [16].

Three moves matter.

The trigger leg. From $20.45 on 1 July 2025 to $13.62 on 21 July 2025, the shares fell 33.4% in thirteen sessions, all of it before Oscar's own pre-announcement, on which the shares then rallied 7.9%. Turnover peaked at 82.7 million shares on 22 July 2025, against a median of 2.68 million a day in the 180 days before the drawdown began. The deterministic gauge records a volume spike of 15.5x on its own definition (peak 20-day average volume in the peak-to-trough leg over the pre-peak 180-day median), which the anatomy in Dislocation covers in full.

The full drawdown. Peak close $23.27 on 19 September 2024 to trough close $10.85 on 30 March 2026: −53.4% per share over 557 days. Market capitalisation fell less — $5,753 million to $3,246 million, −43.6% — because the share count rose 21% in between. That increase is not free-float creep: 33.2 million Class A shares were issued on conversion of convertible notes during 2025, retiring $270.4 million of carrying value into equity [17].

Side by side. Consensus FY2025 EPS finished at −$1.69 against a company guide that had implied positive operating earnings; the price fell 53.4%. That is the shape of the pattern the framework hunts. What follows is the arithmetic on whether the shape was real.

On enterprise value. For an insurer the usual EV bridge misleads. Oscar closed FY2025 with $2,774 million of cash, $1,216 million of short-term and $1,471 million of long-term investments against $1,455 million of benefits payable, a $2,588 million risk-adjustment transfer payable and $430 million of long-term debt [18]. Almost all of that cash sits inside regulated subsidiaries backing reserves. Parent-level cash was $279 million at 31 March 2026 [19] against $431 million of long-term debt [20] — net parent debt of about $152 million, under 2% of the current market capitalisation. Enterprise value and market capitalisation therefore move together here, and market capitalisation is used throughout.

One numeric caveat. The deterministic feature file records a market capitalisation of $7,436 million at the 27 July 2026 close, using 262,388 thousand shares. That figure is the FY2025 weighted-average basic share count, not shares outstanding; the filed balance sheet shows 261,851 thousand Class A plus 35,838 thousand Class B outstanding at 31 December 2025, rising to 299,143 thousand at 31 March 2026 [21]. On the filed count the current market capitalisation is $8,478 million. Every valuation figure below uses the filed count; the discrepancy is recorded as a data gap rather than silently reconciled.

Risk Adjustment Step-Up

One line did the damage. Oscar's premium revenue is direct and assumed policy premiums less risk-adjustment transfers — the federal payment a plan makes when its members score healthier than the market.

Sources: 2023-2025 derived from direct policy premiums, assumed premiums and risk adjustment transfers as filed [22]; 2026 is the company's guided assumption of approximately 20% [23].

The filed table: direct plus assumed premiums of $6,647.7 million in 2023, $10,511.7 million in 2024 and $14,077.9 million in 2025, with risk-adjustment transfers of $950.7 million, $1,526.4 million and $2,596.8 million [24]. That is 14.30%, 14.52% and 18.45%. Management states the 2025 figure as approximately 18.5%, "a 390 basis point increase year-over-year" [25].

The decomposition. Hold the 2025 premium base at $14,077.9 million and apply the 2024 transfer rate of 14.52%: transfers would have been $2,044.3 million rather than $2,596.8 million — $552.5 million less expense. Against a total shortfall of $646.4 million versus the February guide, the risk-adjustment rate step accounts for 85%, leaving about $93.8 million for utilisation, mix and everything else.

The step arrived in three instalments and each one was a re-estimate, not a claim: a $316 million increase to the 2025 risk-adjustment payable in Q2 [26], and a further $275 million accrual increase in Q4 after an updated report showed Oscar's book skewing healthier than the broader market [27]. Oscar's own 10-K states the general case: actual risk-adjustment transfers "have in the past materially differed, and could materially differ in the future, from our assumptions" [28].

Repricing Mechanism

ACA individual coverage is a one-year contract repriced annually with state approval, so the corrective path is procedural rather than commercial. Oscar refiled 2026 rates in states covering approximately 99% of membership at a weighted average increase of about 28%, and priced market contraction at the high end of a 20% to 30% range following the enhanced-subsidy sunset [29] [30].

The whole pool repriced, not Oscar alone. Centene bridged $2.4 billion of 2025 pre-tax earnings to Marketplace morbidity shifts against its prior forecast [31]. Molina repriced its Marketplace book at 2026 rate increases averaging 30%, ranging 15% to 45%, and cut its county footprint by 20% [32].

Two things about that mechanism deserve to be stated precisely, because they cut in opposite directions.

The risk-adjustment level does not revert. Management guides approximately 20% of direct premiums for 2026, above 2025's 18.5% and well above 2024's 14.5% [33]. What corrects is the pricing for it: the 82.4% to 83.4% MLR guide already embeds a 20% transfer rate and still produces $250 million to $450 million of operating earnings [34].

And the first data point after repricing is strong. Q1 2026 revenue was $4.6 billion with an MLR of 70.5% and utilisation largely in line with expectations [35]; earnings from operations were $704 million, net income $679 million or $2.07 a diluted share, membership 3.2 million, and full-year guidance was reaffirmed at every metric [36].

That quarter needs its qualifiers. The 10-Q states that deductibles and out-of-pocket maxima "shift more costs to us in the second half of the year" [37]. Risk adjustment was accrued at 24% to 24.5% of premium in the quarter against the 20% full-year assumption — Blackley attributes the level explicitly to seasonally low Q1 claims rather than to a cushion — and prior-period development was a net $68 million favourable, comprising $150 million of favourable claims run-out against $85 million of adverse state development, with offsetting favourable states not recognised [38].

Sources: 1Q26 reported [39]; consensus quarterly path from the driver-level estimate file, 1Q26 vintage 5 May 2026 and 2Q26-4Q26 vintage 14 July 2026.

Consensus carries the seasonality explicitly: $141 million in Q2, −$80 million in Q3 and −$416 million in Q4, summing with the reported $704 million to $350 million for the year — the exact midpoint of guidance. The Q1 print beat the pre-release consensus of $435 million by $269 million, and the sell side has so far let the beat sit in the guide rather than raise the year.

Two-Scenario NPV

The question the framework asks is narrow: under conservative assumptions, how much of the net present value of future cash flows does this event plausibly destroy, and how does that compare with what the price destroyed. What follows is deliberately simple, so it can be recomputed line by line.

Assumptions, stated. Discount rate 12% on equity — a single-line regulated insurer with no meaningful parent leverage and a demonstrated 4-point swing in annual MLR. Owner earnings are earnings from operations less $25 million of interest, taxed at 21% in cash from 2026 (conservative: Oscar's FY2025 income tax expense was $5.6 million on a $437.3 million pre-tax loss, and the loss carryforwards shelter several years of the forecast). Cash flows are taken at each year-end and discounted from July 2026, so t equals 0.5, 1.5, 2.5 and 3.5 years. Terminal value from 2030 grows at 3%. Net parent debt of $152 million is deducted. Excess subsidiary capital of $809 million is not credited. The share count is 299,143 thousand. Figures below are rounded to the nearest $1 million from an unrounded series, so recomputation from the printed numbers can differ by a dollar or two of millions.

The forecast path. 2026 and 2027 operating income come from consensus at the 14 July 2026 vintage: $349.8 million and $530.5 million. The scenarios separate at 2028.

Source: temporary scenario; 2026-2028 operating income is the driver-level consensus mean at the 14 July 2026 vintage, 2029 grown 12%; owner earnings and present values derived as described above.

Scenario T, temporary. The pool reprices and Oscar holds the 3.48% operating margin consensus carries for 2028 ($772.7 million on $22,226.6 million of revenue). Terminal value is $664 million times 1.03 divided by 0.09, or $7,598 million, discounted to $5,110 million. Explicit-period present values of $242 million, $337 million, $445 million and $447 million bring the total to $6,581 million; less $152 million of net parent debt gives $6,429 million, or $21.50 a share.

Scenario P, permanent. The margin never improves past the 2.69% consensus carries for 2027 and stays there: 2028 operating income of $598 million and 2029 of $646 million. Terminal value $5,612 million, discounted to $3,775 million; total present value $5,025 million, less net debt, or $16.29 a share.

A cross-check on Scenario T without any model. Take the market's own pre-event valuation as the unimpaired NPV: $5,753 million at the 19 September 2024 peak. Push the entire stream out by one year at 12% and it falls to $5,137 million. Add the $270.4 million of debt converted to equity and divide by 299,143 thousand shares: $18.07 a share, or 22.3% below the pre-event $23.27. Two independent routes put the temporary case at $18 to $22 a share.

And the plan. Running the same model at the 5% operating margin management targeted for 2027, applied to the 2028 consensus revenue base, gives $8,846 million of equity value, or $29.57 a share.

Source: derived as set out above from driver-level consensus at the 14 July 2026 vintage and the June 2024 Investor Day margin target [40].

At the trial's 0.71 probability the impairment is temporary, the probability-weighted value is 0.71 × $21.50 plus 0.29 × $16.29 = $19.99 a share, or $5,979 million. On the more conservative cross-check construction of the temporary case it is 0.71 × $18.07 plus 0.29 × $16.29 = $17.55 a share.

The Damage Gap

Measured against the pre-event $23.27, the two scenarios destroy $1.77 to $6.98 of per-share value; probability-weighted, $3.28 (model basis) to $5.72 (cross-check basis). At the 30 March 2026 trough of $10.85, the price had destroyed $12.42.

The gap at the trough was therefore $6.70 to $9.14 a share — $2,004 million to $2,733 million on 299,143 thousand shares, or 38% to 46% of the probability-weighted value. That is the mispricing the framework looks for, and on this arithmetic it was there.

It is no longer there. At the 27 July 2026 close of $28.34 the price sits 42% to 61% above the probability-weighted value and 32% above the temporary scenario's $21.50. It sits just under the $29.57 that falls out of the model when management's original 5% operating margin is achieved on the 2028 revenue base. On these assumptions, the current price already carries the plan working.

The stated multiples are consistent with that reading: 25.8 times FY2026 normalised EPS of $1.098, 18.7 times FY2027 at $1.515, and 12.8 times FY2028 at $2.208, all at the 14 July 2026 consensus vintage.

What would move the answer: a lower discount rate. At 10% rather than 12% the temporary scenario is worth materially more, because roughly 78% of its value sits in the terminal value. Anyone underwriting a 5% margin as the steady state rather than 3.5% reaches a different conclusion — which is exactly the argument the trial adjudicated.

Consensus Record and Gaps

The vendor revision history has a real limitation, and it matters for the numerator. The estimate file's momentum series reaches back 30, 90 and 180 days from 27 July 2026 — that is, to 28 January 2026. It does not span the July 2025 guidance cut. The pre-event consensus path in this tab is therefore reconstructed from the company's own guidance record, which is primary and dated, not from a consensus tape.

What the vendor window does show is the recovery leg, and it is instructive.

Source: consensus values at the 180-day and current vintages of the estimate file; share price from the daily series ($14.87 to $28.34), as reported.

Over that six-month window FY2028 consensus EPS moved from $1.075 to $2.208, up 105%, and FY2028 revenue from $15,058 million to $22,227 million, up 48%. The share price rose 91%. Price and out-year estimate moved together. That is the opposite of the anchoring the framework hunts, where the price stays pinned to a cut near-term number while the out-years recover.

Three other gaps are worth stating plainly.

The deterministic feature file's revenue series is wrong for this purpose. It reports FY2025 revenue of $28.6 million and a 2023 decline of 67%, because it has picked up the "services and other" line rather than total revenue; the filed total revenue is $11,701.4 million, up 28% year over year [41]. The file's conclusion — no three-year high-single-digit revenue decline — happens to hold on the correct series, since total revenue rose in every year from 2021 to 2025 and 2026 is guided up 61% at the midpoint [42]. The structural-decline exclusion does not hit on this evidence.

The consensus free-cash-flow series is not usable as an earnings proxy here. It shows $596 million for FY2025, $3,435 million for FY2026 and $441 million for FY2027 — a 46% forward "yield" on the current market capitalisation in one year and 6% the next. That volatility is the risk-adjustment payable building and settling, not owner earnings; the payable alone stood at $2,588 million at year-end 2025 [43]. The adjusted-yield treatment sits in Yield; this tab uses operating earnings.

The adjusted free-cash-flow feature is not computable at all on this run, so no yield-based cross-check on the NPV work is available.

Temporary Versus Permanent

The temporary-or-permanent question was tried by two opposing briefs, each cited to the corpus, and ruled on by three judges reading blind and in different orders. Both cases are set out here at their strongest; the ruling is the report's, not this tab's.

Sources for the temporary case: FY2024 results and adjusted EBITDA of $199.2 million [44]; the 2022 net loss of $606.3 million [45]; the near-$750 million 2026 swing, paid membership and capital position [46] [47]; peer repricing [48] [49]; the member transition proceeding as expected and driven mostly by members who never made a payment [50].

Sources for the permanent case: 98% of revenue from ACA-regulated plans and 97% of direct policy premiums APTC-subsidised [51]; 93% of premiums earned directly from CMS [52]; risk-adjustment transfers materially differing from assumptions [53]; the quarterly payable increases of $316 million, $130 million and $275 million [54] [55] [56]; second-half cost seasonality [57]; the Street split and $24.20 target [58]; the non-renewal of the Cigna+Oscar arrangement and exit from the small-group market [59]; larger competitors pricing more competitively and obtaining better unit-cost economics [60].

The ruling. The judges put the probability the impairment is temporary at 0.71, with a mean of 0.697 and a spread of 0.06 across the three seats — 0.71, 0.66 and 0.72. Reading order moved the answer by 0.02 (temporary-first mean 0.71, permanent-first mean 0.69). The ruling is recorded as not contested.

Two honesty notes on that ruling. The judges logged nine citation-verification failures, eight of them against the temporary brief — mostly page-attribution errors where the quoted substance was verified one to four pages away, plus two exhibits resting on the vendor estimate files that were unverifiable under the judges' page-only access and were given no weight. The drawdown figures used by both briefs were identical and treated as uncontested. This tab cannot override the ruling and does not attempt to; the diagnosis probability the report carries is 0.71.

What flips it. The judges' conditions converge on a small set: FY2026 operating income below $250 million or an intra-year cut; the first claims-based 2026 Wakely report in Q2 showing market morbidity worse than pricing, with risk adjustment holding at or above 24% of direct premium rather than normalising to the guided 20%; risk transfer stepping above 20% again for 2027 or a second consecutive year of double-digit corrective rate filings; and effectuated membership below roughly 2.6 million by year-end 2026. Confirming the other way: FY2026 closing inside the 82.4% to 83.4% MLR guide with no Q4 true-up above $100 million. The timing of those tests sits in Clock.