Yield

Bottom line

On the framework's basis — free cash flow minus stock-based compensation minus the five-year average of acquisition spend — Oscar's FY2025 adjusted FCF is $970.8 million, a 13.1% yield on the market capitalization of record. Almost all of it is float: the risk-adjustment payable owed to CMS rose $1,029.4 million in the same year, more than the entire adjusted figure. Stripped of working-capital movement, FY2025 adjusted FCF is negative $488.2 million.

The deterministic feature file could not compute this tab's core numbers

fit_features.adjusted_fcf, adjusted_fcf_yield, yield_baseline, fcf_stability, balance_sheet_class and float_retirement_years all return not_computable for this run. The stated reasons are "no annual free cash flow or operating cash flow plus capex" and "debt or cash missing for FY 2025" — the structured cash-flow feed carries operating, investing and financing totals but no capex, stock-based compensation or acquisitions line.

Every figure below is therefore built from the filed cash-flow and balance-sheet statements themselves, page-anchored, using the framework's definition unchanged. Two further defects in the feature file are recorded rather than silently corrected:

• revenue_trajectory reads Oscar's Other revenues line ($28.6 million in FY2025), not total revenue ($11,701.4 million) [1]. The revenue series in that feature is unusable.
• market_cap.native of $7,436.1 million uses 262.388 million shares — the FY2025 weighted average basic count [2], not shares outstanding. At 31 March 2026 Oscar had 263.552 million Class A plus 35.591 million Class B shares outstanding, 299.143 million in total [3], which at the same $28.34 close is $8,477.7 million — 14.0% higher. Every yield below is shown on the feature file's basis first, with the filed-share-count reading alongside.

The adjustment, line by line

Oscar has made no acquisitions in any of the seven years on file: no acquisition line appears in the FY2021, FY2022 or FY2025 cash-flow statements [4][5][6]. The five-year average acquisition deduction is therefore zero, and the whole of the adjustment is stock-based compensation.

All figures $ millions. Adjusted FCF = reported FCF − stock-based compensation − 5-yr average acquisition spend; derived from the filed cash-flow statements. Sources: FY2025 10-K, Consolidated Statements of Cash Flows [7]; FY2022 10-K [8]; FY2021 10-K [9].

The stock-based compensation deduction removes $87.7 million in FY2025 and $159.7 million in FY2023 [10] — 8.3% and 53.6% of reported FCF respectively. In FY2023, when reported FCF was already negative, the stock-based compensation add-back was the difference between a $297.7 million cash outflow and a $457.4 million one. Across the full seven years the adjustment removes $625.2 million, and the size of the deduction has been falling in absolute terms since FY2023 as the share-based expense base has come down.

The reported cash is float

The FY2025 adjusted figure of $970.8 million is larger than any operating result Oscar has ever posted. The company lost $443.2 million that year [11]. The reconciliation is working capital: changes in assets and liabilities contributed $1,459.0 million to the $1,094.9 million of operating cash flow, of which the increase in the risk-adjustment transfer payable alone was $1,029.4 million [12] — 106% of the adjusted FCF the yield is computed on.

That payable is money owed to CMS. It stood at $2,587.7 million at 31 December 2025, and the final market risk-score report that fixes it arrives from CMS in June of the following year [13]. Management says as much in its own liquidity discussion: the timing of risk-adjustment transfers "can be significant" and can swing operating cash flow in any given period [14].

Working capital = the sum of the changes-in-assets-and-liabilities lines in the operating section; ex-working-capital = adjusted FCF less that sum. Derived from filed cash-flow statements: FY2025 10-K [15]; FY2022 10-K [16].

On this decomposition Oscar has generated positive adjusted cash before working capital exactly once in six years — $1.6 million in FY2024. The pattern continued into 2026: first-quarter operating cash flow of $2,619.0 million included $1,993.1 million from the increase in payables to CMS [17], and payables to CMS reached $4,723.2 million by 31 March [18].

The counter-fact worth holding against this: float that grows with a growing book is genuinely usable cash while the book grows, and Oscar's membership rose 56% year-on-year to approximately 3.2 million at 31 March 2026 [19]. Insurance float is a real funding source. What it is not is cash a company can repurchase stock with — it is a regulated liability with a dated settlement.

The yield, three ways

Adjusted FCF from filed cash-flow statements [20], divided by the market capitalization of record ($7,436.1 million at the 27 July 2026 close of $28.34).

Current. $970.8 million ÷ $7,436.1 million = 13.06%. On the 299.143 million shares actually outstanding at 31 March 2026 [21], the same numerator over $8,477.7 million gives 11.45%.

Three-year average. FY2023 through FY2025 adjusted FCF of −$457.4 million, $840.5 million and $970.8 million averages $451.3 million. Against $7,436.1 million that is 6.07%; against $8,477.7 million, 5.32%. The five-year average (FY2021–FY2025) is $259.8 million, or 3.49%.

The company's own baseline. Computing adjusted FCF against each fiscal year-end's own market capitalization gives a series with no baseline to speak of:

Same-year shares from the filed statements of operations, times the last close on or before each fiscal year-end. Oscar listed in March 2021, so FY2019 and FY2020 have no market price. Derived from company filings and the daily price series.

The sign flips in three of five years and the range spans 68 percentage points. There is no stable ~3.5–4% baseline here, so the fortress signature the framework looks for — a formerly steady yield that suddenly jumps toward the bar — is absent by construction. The direction is the opposite of a jump: the 13.06% current reading sits below the five-year median of 23.5%, because the share price rose 161% from the 30 March 2026 close of $10.85 to $28.34. At the March low the same FY2025 numerator would have computed to 34.1%. Roughly 21 percentage points of adjusted yield have been priced away in four months.

Which bar applies

The balance-sheet class selects the reference line, and fit_features.balance_sheet_class returns unknown. Computing it from the filed balance sheet at 31 December 2025:

• Long-term debt $430.1 million; cash and cash equivalents $2,774.2 million [22].
• Net debt = $430.1m − $2,774.2m = −$2,344.1 million. Adding the $1,216.5 million of short-term and $1,471.0 million of long-term investments takes it to −$5,031.5 million.
• EBITDA = loss from operations of −$396.4 million plus depreciation and amortization of $28.9 million = −$367.5 million [23].

The classification rule keys on net debt first: net debt at or below zero is fortress, regardless of EBITDA. Oscar classifies fortress, which selects the ~8–9% reference line (8.5% at the spec's midpoint).

That classification deserves a plain caveat rather than a footnote. Of the approximately $8.1 billion of cash and investments Oscar held at 31 March 2026, $279 million sat at the parent company; the insurance subsidiaries held roughly $1.7 billion of capital and surplus, of which $809 million was described as excess [24]. The rest is regulated statutory capital and money owed to CMS and to providers. The net cash that produces the fortress classification is policyholder and regulator money, not shareholder money. A reader who takes the harsher view and applies the 10% default line is doing something defensible, so both positions are shown.

Yields on the market capitalization of record ($7,436.1 million). Reference lines are the framework's balance-sheet-scaled bars: 8.5% fortress, 10% default. Derived from filed cash-flow statements [25] and the mid-cycle workings below.

Stated as arithmetic: 13.06% on FY2025 adjusted FCF against the 8.5% fortress line clears it by 456 basis points; against the 10% default line, by 306 basis points. On the filed share count those margins narrow to 295 and 145 basis points. The three-year average of 6.07% sits 243 basis points short of the fortress line and 393 short of the default line — and on the filed share count, 318 and 468 short.

Normalized mid-cycle yield

Oscar is cyclical in the specific way the framework's healthcare-insurance pattern describes: the swing factor is the annual pricing-versus-cost-trend cycle, not volumes. FY2025 was the trough of that cycle — medical loss ratio 87.4% against 81.7% the prior year [26], a $396.4 million operating loss driven by market morbidity that pushed the risk-adjustment accrual up [27][28]. FY2026 is the repriced recovery: first-quarter MLR 70.5% [29], operating earnings of $704.1 million against $297.1 million a year earlier [30]. Neither year is mid-cycle, and neither is annualizable — management guides full-year 2026 operating earnings to $250–450 million on $18.7–19.0 billion of revenue [31], which puts the remaining nine months at negative $254 to negative $454 million, because MLR is lowest in the first quarter and highest in the fourth [32].

The assumptions, stated so they can be replaced.

1. Window. FY2023–FY2026E, one complete underwriting cycle at scale. FY2019–FY2022 is excluded: Oscar was sub-scale, ran heavy quota-share reinsurance that made revenue non-comparable, and posted operating margins of −86.9% (FY2020), −29.6% (FY2021) and −14.9% (FY2022) that describe a start-up, not a cycle.
2. Revenue base: $18,850 million, the midpoint of the $18.7–19.0 billion guided for 2026 [33]. This is itself a share-gain year — membership up 56% partly on auto-assigned members picked up from a competitor that left the marketplace [34], achieved despite the sunset of the enhanced premium tax credits [35] — so it is arguably a high base for a shrinking individual market rather than a neutral one.
3. Mid-cycle operating margin: 3.5%, in a 2%–5% band. The anchors: Oscar's own 2027 target of ~5% set at the June 2024 Investor Day [36] and reaffirmed on the Q2 2025 call ("5% is still our target") [37]; the realized FY2023–FY2026E average of −1.2%; and consensus operating margins of 3.2% for 2027 and 4.1% for 2028. The 3.5% base sits below management's target and between the two market-derived anchors.
4. Below the operating line. Interest expense $21.5 million (the Q1 2026 run rate of $5.4 million annualized) [38]; a 5% effective cash tax rate, against a $3,294.4 million accumulated deficit [39] that shelters cash tax (FY2025 cash income tax paid was $17.5 million on $11.7 billion of revenue) [40]; depreciation and amortization $30 million; capex $50 million, between FY2025's $36.4 million and consensus 2028's $68.8 million.
5. Working capital: zero. At mid-cycle the CMS payable neither builds nor unwinds. This is the assumption that does the most work, and it is the reason the normalized figure is a fraction of the reported one.
6. Stock-based compensation is not deducted twice. Under the framework's definition, reported FCF adds SBC back inside operating cash flow and the adjustment then removes it; starting from GAAP net income, which already expenses it, the two cancel.

The arithmetic. 3.5% × $18,850m = $659.8m operating earnings; less $21.5m interest = $638.2m pre-tax; less 5% tax = $606.3m; plus $30m depreciation less $50m capex = $586.3 million of mid-cycle adjusted FCF. On $7,436.1 million that is 7.88%; on $8,477.7 million, 6.92%.

At the 2% margin floor the figure is $317.7 million and 4.27%; at management's own 5% target it is $854.9 million and 11.50% (10.08% on the filed share count).

So: the mid-cycle base case sits 62 basis points short of the 8.5% fortress line and 212 short of the 10% default line. Only management's full 5% target clears the default line, and on the filed share count even that lands at 10.08% — at the line, not through it.

The consensus check

fit_features.consensus_forward_yield computes forward yields of 46.2% for 2026, 5.9% for 2027 and 22.6% for 2028 on the market capitalization of record. Those numbers oscillate by 40 percentage points because the vendor's free-cash-flow line is modelling the same CMS payable timing, not an earnings path. Three specifics make the FCF consensus unusable as a yield anchor:

• The vendor reports no contributor count on the free-cash-flow, cash-from-operations or capex lines, while revenue carries nine estimates and EBITDA seven.
• The lines do not reconcile: 2026 cash from operations of $2,969.2 million less $38.0 million of capex is $2,931.2 million, against a free-cash-flow mean of $3,435.4 million. For 2027 the same subtraction gives $659.9 million against a free-cash-flow mean of $440.7 million.
• For FY2025, now actual, the consensus free-cash-flow mean was $596.3 million against reported FCF of $1,058.5 million — a 44% miss on a completed year.

The closest usable proxy is consensus GAAP net income converted on the same zero-working-capital basis used above (net income plus depreciation less capex; SBC is already expensed inside net income):

Consensus GAAP net income, capex and free cash flow are vendor consensus means as of 28 July 2026; depreciation held near the FY2025 run rate of $28.9 million. Yields on $7,436.1 million. Vendor FCF yields per fit_features.consensus_forward_yield.

On that proxy consensus does not clear the bar until 2028: 5.35% in 2026, 7.38% in 2027, and 9.94% in 2028 — the last still 6 basis points under the 10% default line and 144 basis points above the 8.5% fortress line. On the filed share count the 2028 figure is 8.72%, which clears the fortress line by 22 basis points and misses the default line by 128.

The mean-reversion underwrite, since consensus sits below the bar in the near years. The path does not require reversion — it requires the repricing already in force to hold. The mechanism is specific and dated:

1. Repricing is done, not hoped for. Oscar planned the market on the assumption of no enhanced-subsidy extension and built its products around it [41], and its first-quarter reserves sit on morbidity assumptions consistent with that pricing [42]. First-quarter MLR came in 490 basis points better year-on-year and administrative expense ratio at 15.2%, the lowest in company history [43].
2. Operating leverage is the second lever. Revenue rises from $11.7 billion to a guided $18.7–19.0 billion [44] while the fixed administrative base grows far more slowly; that is the 2–4 points of the margin bridge management attributes to scale [45].
3. What consensus would have to concede. To reach 10% on the feature-file market cap by 2028, normalized FCF must reach $743.6 million — $4.6 million above the current consensus-implied proxy. To reach it on the filed share count, $847.8 million is needed, $108.8 million above consensus, which is an operating margin of about 4.6% on the 2028 revenue estimate against consensus's 4.1%. To clear the 8.5% fortress line on the filed share count, consensus need concede nothing at all.

Probability. I put roughly 55% on the adjusted forward FCF yield clearing the applicable 8.5% fortress line within three years on the market capitalization of record, and about 35% on clearing the 10% default line. The arithmetic behind those: consensus already implies 9.94% for 2028 on the feature-file share count and 8.72% on the filed count, so the fortress line needs no estimate revision on one basis and a modest one on the other; the 10% line needs an operating margin roughly half a point above consensus, which is inside management's own target but has never been delivered. The probability is not higher because three of the six years on file show the underwriting cycle turning against Oscar, and because a 2027 rate cycle that overshoots — the mirror of 2025 — would reset the base. The dominant risk to the estimate is the price itself, not the business: at $10.85 in March 2026 every one of these thresholds cleared comfortably; at $28.34 the margin for error is 3 percentage points of operating margin.

FCF-to-revenue conversion

Total revenue per the filed statements of operations: FY2025 10-K [46]; FY2023 10-K [47]; FY2021 10-K [48]. Adjusted FCF and its ex-working-capital variant derived from the filed cash-flow statements [49][50].

Read on the headline line, conversion is stable-to-slightly-down: 9.2% of revenue in FY2024, 8.3% in FY2025. Read ex-float — the line that matters for whether cash can be returned — the trend is improving and still below zero: −87.1% of revenue in FY2020, −29.5% in FY2021, −15.6% in FY2022, −4.5% in FY2023, 0.0% in FY2024, −4.2% in FY2025. Five years of steady improvement, one year of reversal on the cost-trend miss, and no year yet meaningfully above the line.

That trend is improving rather than deteriorating, which is the direction the framework's levered exception and buyback flywheel both require. What it has not yet done is establish that a positive ex-float number is the normal state rather than a single year's crossing. Whether any of this cash would reach shareholders is a separate question, taken up in Self-Help — no repurchase or capital-return discussion appears in the Q3 2025, Q4 2025 or Q1 2026 calls.

Limitations and what would change this read

• The five-year average acquisition deduction is zero because no acquisition line appears in any filed cash-flow statement in the window. Oscar's forward-looking statements do reference integrating strategic acquisitions [51], and the three Marketplace Subsidiaries were acquired [52], so the zero reflects the absence of a separately disclosed cash line rather than an absence of corporate development.
• The mid-cycle figure moves about $190 million of operating earnings — $179 million after tax — per percentage point of operating margin, which is 2.4 points of yield on the feature-file market cap. A reader who prefers a 2027–2029 window, or who assumes the individual market contracts and takes the revenue base to $16 billion, gets $491.5 million and 6.6% at the same 3.5% margin.
• Two things would change the read materially. If the CMS payable begins to unwind — the mirror of 2025 — reported FCF turns sharply negative while nothing about the business changes, and the headline yield inverts. If Oscar posts a full year of positive ex-float adjusted FCF at the guided revenue base, the normalization ceases to be an assumption and becomes an observation, and the base case moves toward the upper end of the 4.3%–11.5% band.
• The share-count basis moves every yield on this page: the 14.0% gap between the feature file's weighted-average count and the filed shares outstanding is worth 161 basis points on the FY2025 reading and 75 basis points on the three-year average.