Self-Help

What the record establishes

Oscar has never repurchased a share as a public company — Item 5 of the FY2025 10-K answers the question with one word, "None" [1]. The share count has risen every year since listing, and rose 33.1 million shares in the fourth quarter of 2025 alone on convertible-note conversions [2]. The February 2026 credit agreement restricts repurchases outright [3]. Across twelve earnings calls there is no repurchase authorisation, no executed repurchase, and no analyst question on the subject.

The balance sheet against the problem's duration

The face of the balance sheet reads comfortably. At December 31, 2025 Oscar held $2,774.2 million of cash and equivalents, $1,216.5 million of short-term investments and $1,471.0 million of long-term investments — about $5.46 billion — against $430.1 million of carrying-value long-term debt [4]. On a net-debt basis that is roughly $5.0 billion of net cash.

Two liabilities on the same page take most of it back. Benefits payable stood at $1,455.4 million and the risk adjustment transfer payable at $2,587.7 million — $4,043.1 million of claims and federal-program obligations sitting inside the regulated insurance subsidiaries [5]. Management sized the third-quarter 2025 risk adjustment payment for the 2024 policy year at approximately $1.6 billion in a single transfer [6].

Sources: FY2025 10-K, Consolidated Balance Sheets [7]; Management Discussion and Analysis, Liquidity and Capital Resources [8]. Total cash and investments is the sum of cash, short-term and long-term investments as reported.

Where the money sits decides what can be done with it. Of the $5.5 billion, $414.2 million was held at the parent and entities outside the Health Insurance Subsidiaries, of which $14.7 million was restricted; the other $5.1 billion sat inside the insurance subsidiaries [9]. Combined statutory capital and surplus was estimated at approximately $1.0 billion, and the subsidiaries' excess over the minimum risk-based capital requirement fell from $734 million at the end of 2024 to approximately $315 million at the end of 2025 [10].

The parent-only statements in Schedule I make the direction of travel explicit. In FY2025 the parent generated $0.3 million of operating cash, put $160.9 million into subsidiaries, and funded itself with $410.0 million of new convertible notes [11]. Across 2023–2025 the parent received $210.0 million in capital distributions and loan repayments from the Health Insurance Subsidiaries and sent $469.6 million back down [12]. Distributions up to the parent fell from $133.0 million in 2024 to $25.0 million in 2025, while contributions down to the subsidiaries ran $146.6 million and $120.8 million [13][14].

The maturity schedule

Source: FY2025 10-K, Note 9 Debt [15][16][17]. The 2027–2030 rows are holder put dates on the same $35.0 million of 2031 Notes, not additive obligations.

The wall is small and late. Total principal outstanding at December 31, 2025 was $445.0 million: $410.0 million of 2030 Notes at 2.25% maturing September 1, 2030, and $35.0 million of 2031 Notes at 7.25% maturing December 31, 2031 [18]. Cash interest paid in FY2025 was $12.8 million [19]. Refinancing risk over the next four years is close to nil: nothing is contractually due before the June 2027 put on $35.0 million, and both instruments convert into stock rather than cash at the company's election — the 2031 Notes at approximately $8.32 [20] and the 2030 Notes at approximately $24.82 [21], both below the July 27, 2026 close of $28.34 (fit_features.capitulation_gauge).

The covenants that bind

The constraint is not the maturity schedule; it is the February 6, 2026 credit agreement. The $475.0 million three-year secured revolver prices at Term SOFR plus 4.50%, expires February 6, 2029, and carries a 0.50% commitment fee on undrawn amounts [22]. Substantially all of the company's assets are pledged as collateral, and the negative covenants restrict Oscar's ability to "pay dividends or make other distributions on equity interests, or redeem, repurchase or retire equity interests" [23].

The financial covenants run in two phases. Through the fourth quarter of 2026: specified levels of direct policy premiums each quarter, a minimum consolidated adjusted EBITDA each quarter, and minimum liquidity plus undrawn commitments of at least $200.0 million, of which at least $100.0 million must be unrestricted cash at the Company and the guarantors. From the first quarter of 2027: maximum total net leverage of 3.50:1.00 and minimum fixed-charge coverage of 3.00:1.00 [24].

That last package sets the practical ceiling on any repurchase. Parent unrestricted cash and investments were $414.2 million less $14.7 million restricted, or $399.5 million; the covenant reserves $100.0 million of it. The envelope before the leverage and coverage tests engage in 2027 is on the order of $300 million — about 3.5% of a market capitalisation of roughly $8.5 billion at 301.1 million shares outstanding and the July 27, 2026 close of $28.34 [25]. A repurchase programme large enough to add ten points to earnings per share would need roughly $850 million a year.

The company can comfortably outlast the problem — the maturity schedule proves that much. What it cannot do is fund repurchases while it is doing so. Management framed the same capital position around growth, not return: on the Q4 2025 call the CFO gave the rule of thumb that "for every $1 billion of premiums, we are required to hold approximately $50 million of capital" [26], against 2026 revenue guidance of $18.7–19.0 billion [27]. On that arithmetic the 2026 book alone absorbs roughly $900 million of statutory capital.

The repurchase record — executed, not authorised

There is nothing to grade. Item 5 of the FY2025 10-K reports issuer purchases of equity securities as "None" [28]. The Consolidated Statements of Cash Flows carry no repurchase line in 2023, 2024 or 2025 [29]. Treasury stock has been frozen at 315 thousand shares across both balance-sheet dates [30]. The only repurchase in the structured record is $3.0 million in FY2019, two years before the IPO (fit_features.share_count_trend.buyback_cash_per_year). No repurchase authorisation appears anywhere in the FY2021–FY2025 10-Ks; the only mention of buybacks in the archive is a FY2022 risk factor explaining how the Inflation Reduction Act's 1% excise tax would apply if the company were ever to repurchase stock [31].

Source: reported weighted-average basic share counts. FY2023–FY2025 tie to the FY2025 Consolidated Statements of Operations [32]; FY2019–FY2022 are fit_features.share_count_trend.per_year. The feature file's FY2024 entry is the diluted count of 265.9 million, so the chart plots the 240.4 million basic figure the 10-K reports. Repurchase cash was zero in every year from 2020 onward.

The feature file records a five-year share-count CAGR of 55.1% and marks the trend rising: true. Some of that is the 2021 IPO recapitalisation, which is not a governance signal. The post-IPO record is the relevant one, and it points the same way: weighted-average basic shares went from 212.5 million in FY2022 to 262.4 million in FY2025, and from 216.9 million in the first quarter of 2023 [33] to 298.2 million in the first quarter of 2026 [34] — 37.5% more shares in three years.

The fourth quarter of 2025 supplied the largest single increment. Holders converted $270.0 million of 2031 Notes into approximately 32.4 million Class A shares, and Oscar issued a further 0.7 million shares as part of a $17.8 million inducement payment to Dragoneer [35][36]. That is 33.1 million shares — 12.6% of the FY2025 average count — issued to retire $270.0 million of debt at an effective $8.32 conversion price, in a quarter when the stock traded in the twenties. The economics of that exchange belong to the noteholders.

Alongside it, cash-flow items connected to equity in FY2025: $34.4 million spent on capped calls that cap dilution from the 2030 Notes at $37.46 per share, $4.4 million of cash inducement, $4.0 million of tax on net share settlement, against $55.0 million received from option exercises [37][38]. Net, share-related activity was a $12.1 million inflow. Stock-based compensation ran $87.7 million through the cash-flow statement in FY2025, $109.8 million in FY2024 and $159.7 million in FY2023, with a further $12.8 million capitalised into software in 2025 [39][40]. Another 13.9 million shares remain available for future issuance across the 2021 and 2022 plans [41].

The framework treats a persistently rising share count driven by stock-based compensation and dilutive issuance as disqualifying rather than as a matter of degree. Oscar's count rises on both, and no repurchase has ever offset either.

What the cash flow will actually support

The deterministic feature file cannot compute adjusted free cash flow, the adjusted yield, or float-retirement years for Oscar: fit_features.not_computable records "no annual free cash flow or operating cash flow plus capex" for adjusted FCF, and float-retirement years fails for want of a positive adjusted FCF figure. The reason is a gap in the structured feed, not in the filings — the FY2025 cash-flow statement reports every component. The figures below are therefore derived here from the filed statement, and are labelled as such rather than substituted for the feature.

Source: derived from the FY2025 10-K Consolidated Statements of Cash Flows [42]. Adjusted FCF = operating cash flow − capital expenditure − stock-based compensation; no acquisition line is disclosed in FY2023–FY2025 investing activities, so the five-year acquisition adjustment is zero as reported. The second series removes the year's movement in the risk adjustment transfer payable, a federal-programme liability that is later paid out in cash.

On the reported basis FY2025 adjusted free cash flow is $1,094.9 million of operating cash flow less $36.4 million of capital expenditure less $87.7 million of stock-based compensation, or $970.8 million [43]. Against the feature file's market capitalisation of $7,436.1 million that is a 13.1% adjusted yield and 7.7 years of float retirement; against the roughly $8,533.8 million implied by the 301.1 million shares actually outstanding on April 10, 2026, 11.4% and 8.8 years [44].

The reported basis flatters the business. Of that $1,094.9 million of operating cash flow, $1,029.4 million was the increase in the risk adjustment transfer payable — the balance grew from $1,558.3 million to $2,587.7 million over the year [45][46]. That is money owed into the federal risk adjustment programme and settled in cash the following year, as the $1.6 billion Q3 2025 transfer for the 2024 policy year demonstrates [47]. Strip that single item and FY2025 adjusted free cash flow is negative $58.5 million, against $339.1 million in FY2024 and $3.1 million in FY2023. Measured against total revenue of $11,701.4 million in FY2025, $9,177.6 million in FY2024 and $5,862.9 million in FY2023, adjusted cash generation on that basis runs −0.5%, 3.7% and 0.1% respectively [48][49]. The yield question in full is the business of Yield; what matters here is that neither basis produces cash the parent can reach.

The absurdity check

fit_features.float_retirement_years is not_computable, for the reason quoted above. Stated as arithmetic from the filed statement instead: $8,533.8 million of market value divided by $970.8 million of reported adjusted free cash flow is 8.8 years; divided by the $58.5 million negative figure that remains once the risk adjustment payable build is removed, the calculation has no positive solution. The framework's reference point for a price that cannot survive is roughly three years.

The levered exception

It does not apply, and two of its three legs fail independently. Oscar's adjusted yield is 11.4% on the flattering basis, not the ~25% the exception requires. fit_features.balance_sheet_class is unknown — the deterministic pass could not resolve it — but the filed figures show net cash, not leverage. And the exception's second leg, a demonstrated multi-year reduction in share count, fails outright: the count has risen in every year on record.

Management's intent, from the record

Across the twelve earnings calls in the corpus — Q2 2023 through Q1 2026 — the words "repurchase" and "buyback" do not appear once, in prepared remarks or in the question-and-answer sessions. No analyst has asked, and there is no statement of repurchase intent in the record to weigh.

What the calls do contain is a consistent framing of the same capital as growth funding and loss absorption. In November 2024, reporting $575 million of excess capital as of September 30, the CFO said Oscar "continue[s] to believe our excess capital positions us well to fund future growth and allow us additional opportunities to optimize our capital position over time" [50]. That is the closest the archive comes to a capital-return hint, and it was not developed on any subsequent call.

When Josh Raskin of Nephron Research pressed on uses of cash in August 2025 — the only direct analyst question on the subject in the set — the answer was entirely about absorbing losses: "We think that the bulk of the remaining losses that we're forecasting for this year are going to be absorbed by that excess capital position… I do think that parent cash will decline in the back half of the year, largely due to us making some additional capital contribution to the insurance subsidiaries" [51]. Excess capital had fallen roughly $300 million in that quarter alone.

By the Q4 2025 call the framing had moved to raising capital rather than returning it: "we have taken opportunistic steps to strengthen our capital position and optimize our capital structure… during the third quarter, we increased our capital in preparation for 2026 growth, completing a $410 million convertible notes offering due 2030, generating $360 million of net proceeds" [52]. Parent cash and investments then fell from $414 million at December 31, 2025 to $279 million at March 31, 2026, even as subsidiary excess capital recovered to $809 million on a strong first quarter [53].

Insider buying alongside

One purchase, and it came from the company rather than the market. On April 3, 2026 — four days after the $10.85 trough — Oscar entered a stock purchase agreement with Mark Bertolini and sold him 1,000,000 Class A shares at $11.92, the prior trading day's close, for $11.9 million, in a private placement under Section 4(a)(2) [54]. The Form 4 record carries it as a purchase settled on April 6 [55]. It is the chief executive putting personal cash in near the low, and it is the only insider buying in the file since August and September 2021, when Joshua Kushner and the Thrive vehicles bought approximately 5.0 million Class A shares for roughly $79 million in the $12.69–$18.35 range and Mario Schlosser bought 57,300 shares at $17.53 [56]. The Form 4 feed codes the April transaction P, its open-market purchase code; the 10-Q's description of a stock purchase agreement with the company is the more exact one. Nothing else was bought on either side of the March 30, 2026 close of $10.85 (fit_features.capitulation_gauge).

Selling in the opposite direction is on the record. Between June 25 and June 30, 2026, Bertolini disposed of 2,445,306 shares at $28.35–$30.08 for approximately $70.7 million, and Schlosser sold approximately $30 million across June and July 2026 [57]. That left Bertolini's reported holding at 7,751,570 shares [58]. The fair counter-fact: every one of those dispositions is flagged as executed under a Rule 10b5-1 plan, and on November 10, 2025 Bertolini entered the company's standard sell-to-cover instruction, which provides for sales of as many shares as are needed to cover tax withholding on the vesting or settlement of his restricted stock units [59]. Those are settlements, not a view. What stands without qualification is the company's own side of the ledger: through a decline to $10.85 and a recovery to $28.34 it repurchased nothing, and the 1,000,000 shares it moved at the low it sold rather than bought.

Dividend safety

Immaterial to the case. Oscar "has never declared or paid any cash dividends" and does not anticipate paying any in the foreseeable future; state insurance holding company law and the 2026 credit agreement both restrict distributions [60][61].

Management credibility

The sample below is the five most material forward commitments in the transcript archive from two to four years back, each checked against what the company subsequently reported.

Sources: Q2 2023 call [62]; Q2 2024 call [63]; Q3 2024 call [64]; Q4 2024 call [65]; Q2 2025 call [66]; FY2025 results as reported [67].

The 2023 and 2024 commitments were met on the company's own subsequent reporting: full-year 2024 adjusted EBITDA of $199 million and net income of $25 million, both stated on the Q4 2024 call [68].

Revenue and membership guidance has been reliable; margin guidance has not. The FY2025 miss is the largest instance — an operating profit of $225–275 million guided in February 2025 became a $396.4 million operating loss, an 87.4% medical loss ratio against 80.7–81.7% guided [69]. Management attributed the gap to market morbidity and the resulting risk adjustment accrual, which is the genuinely hard estimate in this business and is the subject of Damage Math.

The instance that bears on character rather than forecasting is the long-term target. At the June 2024 Investor Day the company committed to "at least 20% revenue CAGR and a 5% operating margin by 2027" [70]. It was carried in prepared remarks on the three calls that followed, through February 4, 2025 [71], then appeared only under questioning, as late as August 6, 2025, when Bertolini said "we're not changing our longer-term forecast at this moment, but 5% is still our target" [72]. It has not been stated on any call since: not the November 2025 call, where it goes unmentioned, and not the Q4 2025 call of February 10, 2026 or the Q1 2026 call of May 6, 2026, neither of which restates it, withdraws it, or explains its absence; management now points to a September 16 investor day [73]. The +Oscar/Campaign Builder platform business, one of four named strategic pillars through 2023, disappeared from prepared remarks after Q2 2024 on the same pattern.

Set against that, the ownership test cuts the other way. Bertolini beneficially owned 11,925,092 Class A shares at April 10, 2026 — 8,599,999 held directly, 391,760 in near-term exercisable options, and 2,933,333 through the Anahata Foundation of which he is co-trustee [74][75]. His FY2025 pay was $1,149,308 in total, of which $619,178 was salary and nil was stock or option awards [76]; the CEO pay ratio was 10.4:1 [77]. Executives and directors as a group hold 21.8% of Class A on an as-converted basis and 77.2% of the voting power [78]. Even after the June 2026 settlements, the chief executive's directly held stake is worth roughly $220 million at $28.34 — many multiples of his cash compensation.

The read the evidence supports: this is not the promotional-CEO pattern the framework excludes. That pattern requires big claims, repeated misses and an absent economic stake, and the third leg is plainly missing here. What the record does contain is one substantiated instance of a multi-year target carried for six calls and then dropped without acknowledgement, in a company that has also quietly retired a named strategic pillar. The strongest fact against a benign reading is that the September 2026 investor day, not a call, is where the 2027 numbers were sent — a target restated or formally withdrawn there would settle it, and a third consecutive call without mention would not.

What would change this read

A board authorisation of a repurchase programme with cash actually deployed, disclosed in Item 5 rather than announced; a covenant amendment or refinancing that lifts the restricted-payments limitation; sustained parent-level free cash flow independent of the risk adjustment payable cycle; or an inflection in the share count from conversions and stock-based compensation into net retirement. None of these is present in the record through the first quarter of 2026. The timing of any of them is the business of Clock; the durability of the underlying cash generation, of Durability.