Phase Space AI

Valuation Analysis

Edwards Lifesciences [EW]

Edwards Lifesciences [EW] — Valuation, Street Bridge & Factor Scorecard

Tasks 2–3 · investment-memo v1.6.0 · analysis dated 2026-07-28, migrated onto the Criteria framework 2026-07-29

Spot $83.675 · Archetype COMPOUNDER · Implied path: 21.1% required vs 10.8% demonstrated → margin −10.3pp · 12-month target $96 base (+14.9%)

This document no longer states a position verdict. Under v1.6.0 the memo scores every Criteria, blocks on none, and emits an analysis. Whether that analysis justifies a position is a question about a particular book.

---

0. Valuation under v1.6.0 — two horizons, both mandatory

Reporting only one of these is a defect: they answer different questions over different horizons, and a 5-year instrument provably cannot be calibrated to a 12-month one (tested 2026-07-29 against a professional's targets; best achievable correlation +0.08).

Inputs are taken from reports/scan/EW_analysis.json (as_of 2026-07-28) rather than recomputed, per the update spec. Where the scan and the older body of this document disagree, the scan wins and the correction is stated below rather than silently applied.

0.1 The implied-path test (reverse DCF) — the Valuation Criteria · FAIL

assets/reverse_dcf.py, solving for the parameter the price requires rather than asserting a path and reporting an expected return.

Sensitivity — run over the exit multiple, never over scenario probabilities. Running the range across probability weights while point-estimating the multiple is the exact defect this replaces.

The margin turns positive nowhere in a defensible range. Breakeven requires a 35.4x exit multiple — i.e. no compression at all from today's 35.2x, five years out, on a business already at $6bn of revenue.

DISCLOSED DEFECT IN THE PRE-COMPUTED INPUT (2026-07-29). The update spec describes a "derived terminal margin" in the scan file. There is none — EW_analysis.json has no terminal_margin key. The published required_cagr_pct inverts exactly to a flat 20.0% terminal EBIT margin applied to every name in the universe, which is the hardcoded 0.20 that coverage_scan.py's own comments record as a defect fixed after these scan files were written. This is disclosed rather than presented as a derived figure.

EW's own operating margin is 20.8% (GAAP FY2025). Substituting it: required CAGR 20.1%, margin −9.3pp. The scan figure (−10.3pp) is the headline because the spec instructs use of the pre-computed inputs; the corrected figure is −9.3pp and the direction is modestly favourable.

0.2 The 12-month target · $96 base, +14.9% to spot

Built the way the horizon demands — from near-term estimates and named product-cycle events, with the multiple anchored on EW's own trading history and percentile, never a peer median projected years forward.

Step 1 — near-term revenue. Trailing revenue twelve months from now (the denominator the multiple will be applied to at July 2027) is a half-and-half blend of the FY2026 and FY2027 consensus:

Alpha Vantage EARNINGS_ESTIMATES returned the rate-limit message for EW; the shared 25/day quota was exhausted. The FY2027 revenue figure is therefore derived and is labelled as such everywhere it appears. A quota gap leaves the Consensus Criteria blank and blocks nothing.

Step 2 — named product-cycle events inside the window (each dated in the Catalyst Criteria calendar): CMS final NCD decision memo 2026-09-13; PROGRESS moderate-AS readout at TCT (2026 Q4, company-indicated); FY2027 guidance, ~Feb-2027. No event is asserted here that the Catalyst Criteria calendar does not carry with a date.

Step 3 — the multiple, on EW's own history. Price-to-sales on a constant current share count (575.8m), against an internally-consistent EDGAR-derived TTM revenue series, daily, 2023-07-03 → 2026-07-28 (n = 770).

EW's multiple sits near the bottom of its own three-year range. That is the anchor: mean reversion within its own history, not convergence to a peer.

The base case contains no multiple opinion at all — it is revenue growth at an unchanged multiple. The band is generated entirely by the multiple, which is where the judgement lives.

Sanity band vs the external reference. Street average target $100.96 (+20.7%) sits inside this band, between base and the reversion case. That is a check on the output, never a calibration target for the model.

CORRECTION — the old target was $92.00 (+10.0%), built from probability-weighted scenarios on a forward P/E. The new base is $96 (+14.9%). The change is not a change of view about Edwards; it is a change of instrument. The old construction weighted three scenarios whose multiples (18x / 24x / 28x) were asserted, and the sensitivity was run over the probabilities. The new construction runs the range over the multiple, and anchors it on EW's own observed distribution rather than on a judgement.

0.3 Data hygiene and corrections to this document

Nothing below this line has been deleted. §§1–7 are the v1.4.2 body, retained because the evidence in them is unchanged and independently useful. Where they state a conclusion this framework no longer draws, a superseding note marks it.

---

1. The model, and how it was verified

EW_Financial_Model.xlsx — six tabs (Assumptions, Income_Statement, Balance_Sheet, Cash_Flow, DCF, Scenarios), live formulas throughout, FY2022A–FY2032E. Verified by opening it in Excel via AppleScript and reading computed cells back, then re-verified after every fix.

1.1 Tie-out to the filed FY2025 statements — stated explicitly

A zero balance check verifies internal consistency, not input accuracy. Every actual-year line is therefore tied back to the filed figure, and the tie is stated:

All residuals are sub-$0.10m rounding on six-decimal ratio inputs. Balance check = 0.0000 in every one of the eleven modelled years.

1.2 Three real defects found in verification and fixed

Recorded because a model that passes silently is a model nobody checked.

1. Stock-based compensation leaked the balance sheet by exactly its own value every forecast year. SBC was added back in operating cash flow but never credited to equity. Assets grew by NI + SBC − buybacks while liabilities and equity grew by NI − buybacks. Fixed by adding SBC to the equity roll.
2. Non-controlling interest broke the FY2023 and FY2024 actual-year balance by $69.4m and $64.5m. StockholdersEquity excludes NCI; Assets does not. Fixed by using StockholdersEquityIncludingPortionAttributableToNoncontrollingInterest.
3. The interest-income line was circular. Interest was driven off same-year cash, which is driven off free cash flow, which is driven off net income, which is driven off interest. Excel returned blanks. Fixed by driving interest off prior-year cash.

A fourth issue was a data problem, not a formula problem: the FY2022 and FY2023 product-group tables predate the Critical Care discontinued-operations restatement and sum to $63.4m and $66.7m more than the filed continuing-operations revenue. Rather than quietly using the wrong total, the model carries an explicit "Reconciling items" line so revenue ties to the filed figure exactly in all four actual years. FY2021 was dropped entirely — no restated continuing-operations figure for it exists, and a reconstructed number would have been unsourced.

1.3 The forecast

FY2025 operating margin of 20.8% is depressed by the $325.4m intellectual-property/litigation charge. On an adjusted basis FY2025 operating margin was 28.7%.

Guidance calibration — the model must reproduce what the company has told the market. FY2026E adjusted EPS of $2.99 sits inside the company's guide of $2.95–3.05 and within $0.01 of the $3.00–3.01 consensus. The FY2025 adjusted figure of $2.56 reproduces the $2.564 implied by the company's own "growing 17% at midpoint." The normalised tax rate is the calibrating input and is disclosed as such on the Assumptions tab rather than buried.

FY2026E revenue of $6,816.2m is +0.8% above the $6.76bn consensus and inside the company's $6.6–6.9bn guide. It implies H2-2026 revenue growth of ~+9.9% against H1's +15.1% — i.e. the model already embeds most, but not all, of the ACURATE-exit deceleration consensus assumes.

1.4 Base-rate check on every scenario's growth assumption

Chan, Karceski & Lakonishok (2003): growth persistence beyond chance is close to nonexistent, and the modal error in financial models is extrapolating the recent rate.

• Bull (25%) — 10.9% revenue CAGR for seven years from a $6.07bn base. Edwards' own FY2022–25 continuing-ops CAGR was 10.8%. The bull case therefore assumes the last three years repeat for seven. Sustained double-digit growth for 5+ years at multi-billion revenue scale sits in roughly the top decile of the reference class. This is named as the above-base-rate assumption it is; it may still be right, but it is not the modal outcome.
• Base (50%) — 8.0% CAGR. Below the FY2022–25 run-rate and below management's own stated "approximately 10% total company sales growth, on average, over the longer term." Reference-class-consistent for a $6bn franchise with a dominant share of a maturing core market plus one genuinely growing second franchise.
• Bear (25%) — 5.5% CAGR. TAVR reverts toward the +3.8–6% printed in 24Q4–25Q1 once the ACURATE exit fully laps; the NCD delivers access but not incremental procedures; TMTT decelerates faster than modelled as SAPIEN M3's second-line label bites.

---

2. Valuation

2.1 DCF

WACC 7.9% (risk-free 4.69%, measured beta 0.56 from 251 days of daily returns against SPY via Alpaca, ERP 5.5% — beta is measured, not sourced from a website), terminal growth 3.0%, mid-year convention.

Terminal value at 76.9% of EV is high but normal for a profitable mature compounder — materially healthier than the 98.8% recorded on GH or the negative enterprise value the DCF returned on TWST. The DCF is a real input here, not a decoration.

Sensitivity (WACC × terminal growth):

Today's price of $83.68 is reached at roughly a 7.4% WACC with 3.0% terminal growth, or a 7.9% WACC with ~3.6% terminal growth. Neither is unreasonable — which is the honest conclusion. The DCF says EW is fairly priced to modestly expensive, not that it is cheap or dear.

2.2 Comparable companies

Operating margins from EDGAR XBRL latest fiscal year; forward P/E and NTM growth from stockanalysis.com and aggregators, 2026-07-28. Gross margin not separately tagged for ABT/MDT/ZBH.

Regressing forward P/E on NTM growth across the seven names puts EW essentially on the fitted line. EW has the highest gross margin in the set by a wide margin (78.0% vs a 64–69% cluster) and mid-pack growth. The comps say the same thing as the DCF: fairly valued. Neither method produces a valuation-driven thesis in either direction, which is the correct starting point and is why Task 5 does not reach for one.

2.3 Scenarios and the 12-month target

Horizon: 12 months to end-July 2027. Valued on a forward P/E applied to FY2028E adjusted EPS — which is the forward year as at July 2027, mirroring how the stock is priced today.

EW trades at 26.4x forward today. The base case therefore assumes a modest de-rate to 24x, not a re-rating — holding the multiple flat at 26.4x would give a base target of $102.7 (+22.7%). The bear assumes a de-rate to 18x, roughly where Medtronic trades.

Long-horizon cross-check to FY2032E — and it disagrees with the 12-month view, which is disclosed rather than suppressed:

The six-year base case annualises to +4.5% while the 12-month view returns +11.6%. These are not inconsistent — near-term adjusted EPS growth of 13–15% is faster than the six-year average, so a one-year holder captures the fast part.

SUPERSEDED (2026-07-29). The original text compared the six-year annualised return to a 4.7% cash hurdle and used the shortfall to justify a position verdict. Both the hurdle and the verdict are retired. The surviving, and better, statement of the same observation is §0.1: over five years today's price requires a 21.1% revenue CAGR against 10.8% demonstrated — a −10.3pp margin — while over twelve months the multiple sits at the 18th percentile of its own three-year range. The horizons genuinely disagree, and v1.6.0 reports both rather than resolving one into the other.

2.4 Probability-weighted output and the cash hurdle — RETIRED IN FULL

SUPERSEDED (2026-07-29). Everything this section computed has been retired by framework v1.6.0, and it is deleted rather than annotated because retaining it invites the reader to use it.

• The cash hurdle is retired. It is replaced by slot competition inside the strategy, not by an arithmetic comparison inside the memo.
• The E[R]-versus-hurdle test is retired. The Valuation Criteria is now the implied-path test: §0.1, required 21.1% vs demonstrated 10.8%, margin −10.3pp.
• The probability-weighted range is retired as a sensitivity. Ranging over bear/base/bull weights while point-estimating the exit multiple runs the analysis on the parameter that cannot change the answer. The sensitivity now runs over the exit multiple — §0.1.
• The "flip point at a 37.5% bear weight" statistic is retired. It measured the robustness of an arithmetic that no longer exists.

The old figures, for the record only: probability-weighted 12-month target $92.00; net 12-month E[R] +9.6%; cash hurdle 4.7%; margin +4.94pp; range +0.4% to +19.7%.

---

3. Street bridge — required, and the house is not far from consensus

3.1 Consensus

Alpha Vantage EARNINGS_ESTIMATES — the intended structured source, including analyst counts and the revision history — returned the rate-limit message; the shared 25/day cap was exhausted by concurrent work. Everything above is web-sourced per references/consensus-bridge.md, and the derived FY2027 revenue figure is labelled as derived every time it is used.

3.2 The house-versus-Street decomposition

The house is inside normal model dispersion on every fundamental line. The largest gap — +4.2% on FY2027 adjusted EPS — is worth dissecting rather than claiming, and the dissection is unflattering:

The FY2027 EPS gap is substantially a tax assumption, not an operating insight. The model steps the normalised tax rate from 23.0% in FY2026 (elevated by the enacted R&D-credit law change; the company's reported Q2-2026 effective rate was 53.6% and H1 36.4%) down to 21.5% from FY2027. That single step is worth roughly +1.9% of EPS. Holding the rate flat at 23.0% puts house FY2027 adjusted EPS at ~$3.38 versus Street's $3.30 — a +2.4% gap, comfortably inside a 28-analyst dispersion. I could not determine from the Q2 release whether the tax elevation is a discrete catch-up or a permanent rate change, and the 10-Q detail was not pulled. This is a modelling choice dressed up as a variant if it is not named, so it is named.

3.3 Numbers or multiple? — the required decomposition

Applying each price point to the same FY2028E adjusted EPS base of $3.89, so the multiple is isolated as the only variable:

The gap between the house and the Street is therefore almost entirely about the multiple, not about the numbers — a 26.0x versus 23.7x disagreement on an identical earnings base. That is a 2.3-turn argument, and 2.3 turns is not a thesis. Stated as the falsifiable claim it is: twelve months from now, either Edwards' forward multiple held at or above 26x, or it compressed toward 24x. Nothing about FY2027 or FY2028 earnings needs to be different for either to happen.

3.4 The honest risk to the house view

If medtech's multiple re-rating is structural rather than a reversion candidate — and there is a real case that it is, given rate expectations and the sector's demographic tailwind — then anchoring on 24x will systematically miss for as long as that regime holds. EW has traded at 40–50x forward as recently as 2021. The house target is below consensus for a reason that is a multiple opinion, and a multiple opinion is exactly what this framework says cannot carry a directional call.

Sell-side targets also skew structurally bullish industry-wide, and several of these were likely refreshed immediately after a guidance raise — extrapolation of recent momentum rather than independent multiple stress-testing. Both effects are real; neither is sufficient to explain an 8.9% gap on its own, and neither is being used to do so.

---

4. Factor & Anomaly Scorecard

All inputs computed from sources already wired: EDGAR XBRL (CIK 0001099800) and Alpaca daily bars.

F-score detail (FY2025 vs FY2024, continuing operations):

4.1 Synthesis — do the factors support, contradict, or split?

They support a long, moderately, and the strongest signals are the earnings-quality ones. Accruals at −3.81% with 1.49x cash conversion, gross profitability of 0.346, low asset growth and upward-revised guidance are, taken together, a clean profile. Nothing in the scorecard argues for a short at any level.

Two honest qualifications. First, momentum is neutral, not supportive — a stock sitting exactly on its 200-day with a 50-day above it is going sideways. Momentum Criteria is MEASURED: it informs entry timing and has no power to help or hinder ownership of this name. Cross-sectionally the 12-1 sits at the 44.7th percentile, the middle quintile. Second, the F-score of 5/9 is middling and four of the four failures are deterioration tests — ROA, current ratio, gross margin and asset turnover all moved the wrong way year-on-year. Three of those four are explained by the Critical Care divestiture reshaping the asset base and by the $325.4m litigation charge, but "explained" is not "absent," and the score is reported as computed rather than adjusted into looking better.

Single-name factor readings are noisy. These are evidence weights, not verdicts.

---

5. Transcript mention-frequency table (required exhibit)

Source: 19 quarters of SEC EDGAR 8-K Item 2.02 Exhibit 99.1 earnings releases, 2021Q4–2026Q2. Mentions per 10,000 words. History window stated: 19 quarters — this is not "all history." Every release is company-authored prepared text with no Q&A section, so the prepared-remarks share is 100% for every term.

Emerging: guidelines, NCD, centers, physician, competition, implantable/IHFM, heart failure, SAPIEN M3, proactive, pathway, regardless, EACTS, determination, ECLIPTIS. Decaying: structural heart, clinical trial, pivotal, enrollment, Critical Care, asymptomatic (substitution, not de-emphasis). Stable: TAVR, TMTT, Surgical, Japan, Europe, PASCAL.

Two false positives were caught and discarded rather than reported: the R&D spike to 27.5/10k in 2026Q2 is mostly the phrase "law change on R&D credit," a tax item; and a large block of apparently vanished balance-sheet vocabulary (percent, equity, payable, receivables, inventories) is a formatting artifact of Q1/Q3 releases omitting the financial statements, not a disclosure withdrawal.

---

6. Options market read

Alpaca options snapshots, 2026-07-28, 64 contracts across two expirations.

Implied sits above realised at the January expiration by the usual margin — the standard variance risk premium, which argues for spreads over naked long premium. The September expiration is closer to fair. Note this is the opposite of the GOOGL finding, where implied sat below realised and inverted the spread default.

Liquidity is the real constraint, not pricing. Bid-ask spreads are wide relative to premium: the 18-Sep-2026 $85 call quotes 1.97 / 3.54, a spread of 57% of mid. Open interest is concentrated in a few strikes ($100 calls, 4,874 contracts; $92.50 calls, 978). For a $48bn company the chain is surprisingly thin, and any options expression pays a meaningful crossing cost before the thesis has a chance to work.

---

7. Portfolio context — measured, not assumed

The brief flagged that a prior run asserted correlations that measurement contradicted. These are measured from Alpaca daily bars, 251 trading days to 2026-07-28.

The EW–ISRG correlation is +0.399 on the 251-day window and does not exceed +0.471 on any window measured. It is BELOW the book's 0.60 disclosure threshold on all three horizons. The med-device overlap with the existing ISRG watchlist entry is real but does not trigger the correlation rule. This was measured, not assumed — and the answer is the opposite of what "two med-device names" would suggest.

EW is also the lowest-volatility name in the entire coverage universe: 24.3% realised versus GOOGL's 31.4% (the prior low), ISRG's 35.9%, NTRA's 43.8%, GH's 60.7%, TXG's 68.6%, TWST's 68.2% and SNDK's 109.1%. Beta 0.56. On the sizing grid that is unambiguously the Low volatility tier on both criteria (beta < 0.8 and realised vol < 25%) — the first name in coverage to qualify.