Phase Space AI

Valuation

U.S. Physical Therapy [USPH]

U.S. Physical Therapy [USPH] — Valuation

Spot $79.675 (2026-07-30). Two outputs: a 12-month target and the 5-year implied-path test. They disagree in direction, and that disagreement is the finding.

---

1. COMPANY STATE

STATE A — mature and structurally stable. Evidence in USPH_Research.md §6. Consequence: terminal margin built from USPH's own economics plus an explicit forward bridge; industry data as a sanity band only. evidence_grade: B.

---

2. Verified inputs — and the correction that decides the memo

Basis consistency, stated once and applied throughout: consolidated EBIT is valued against an NCI-inclusive EV; per-share figures use attributable income against USPH-only shares. The effective tax rate applied is 25.5% (tax ÷ consolidated pretax income), not the 33.4% USPH discloses — that rate is computed on income after NCI because the partnerships are pass-throughs, and applying it to consolidated EBIT would double-charge tax.

---

3. TERMINAL MARGIN — 10.5%

Constraint and sufficiency test

m_EBIT,T ≤ m_gross,T      →     10.5% ≤ 19.6%     ✓   (9.1pp of headroom)

Terminal EBIT consumes 53.6% of terminal gross profit, leaving 9.1pp of revenue for corporate office and all other central cost against 9.03% actually spent on the TTM. Between Celestica (3.6pp of headroom, admissible) and UFPT (13.5pp). EV/EBIT is an admissible instrument for USPH — its 19.0% gross margin is well above the <15% trigger the low-margin probe in this project recommended.

The bridge

Arithmetic: 19.6 − 8.7 − 0.4 = 10.5.

Position vs the record: 10.5% is +0.5pp above the TTM actual of 10.00% and −0.6pp below the FY2025 actual of 11.10%. It sits inside USPH's own five-year observed range (8.6%–14.3%) and below its most recent full year. This is a conservative, well-supported terminal margin and it is not where the name fails.

steady_state_check.py — what it flagged

python3 steady_state_check.py --ticker USPH --terminal-margin 0.105 --exit-multiple 11.0
  (asset sha256: 6fbbb0f9…)

This is the most severe steady-state flag of the three names in this run, and it is correct.

Two things it surfaces that no other instrument in the process did:

1. ROIC of 6.0%. Against a WACC of roughly 9.5% for a $1.2bn, 42%-vol, leveraged clinic operator, USPH is earning below its cost of capital on consolidated invested capital. Invested capital is $963.7m (equity $469.0m + NCI $313.9m + debt $204.8m − cash $28.4m) of which goodwill $715.9m and intangibles $179.8m are 92.9%. Tangible invested capital is $68.0m. Serial acquisition of partial clinic interests at full prices is what a 6% ROIC looks like.
2. A terminal reinvestment rate of 50.3% implied by g/ROIC at g = 3%. Half of every terminal dollar of NOPAT has to be reinvested just to hold 3% growth. That is the arithmetic consequence of a 6% ROIC and it is what makes the warranted multiple 5.6x.

Note it did NOT flag the terminal margin. At +0.2pp versus the latest annual and 3.8pp below the window peak, the margin passes cleanly. Both UFPT and USPH failed on the exit multiple, not the margin — which is the same lesson the NOW memo recorded: the exit multiple was the unexamined parameter.

---

4. Exit multiple — 9.0x, derived and deliberately generous

EV_T/EBIT_T = (1−t)(1−g/ROIC)/(WACC−g) at t = 25.5%, g = 3.0%:

Base adopted: 9.0x — near the top of that range, and 1.6x the warranted multiple at the measured ROIC. Chosen generously on purpose: if the name fails at a generous exit multiple, the failure is not an artefact of the multiple. It implies 58% compression from today's corrected 21.4x.

Per criteria.md the base may not sit below every stated anchor without argument — it does not.

Growth-matched comparator set — USPH's own, shared with neither RMD nor UFPT

Outpatient healthcare services and clinic operators — a completely different operating model from RMD's branded devices or UFPT's contract manufacturing. This is the point of running three separate sets.

n = 5, but NONE is underwritten in this project and none has a verified multiple in this corpus. Per valuation.md rule 6, reference-class validity requires n ≥ 5 mature profitable firms with a matching operating model and dispersion in the dimension being matched — I can assert the operating-model match but I cannot assert verified growth or multiples for any member without underwriting them, which this run's time box does not permit.

The exit multiple is therefore declared UNIDENTIFIED from the comparator set and taken from the warranted-multiple identity instead. Declared, not defaulted to a peer median — which is exactly the failure mode the diagnostics-anchor defect (a 1.0–7.5% growth set used to value 15–39% growers) established.

One thing the set does establish qualitatively: Surgery Partners runs the same partial-interest, NCI-heavy structure, and the NCI treatment in §2 is the standard question for that whole class. Any comparison of USPH to a wholly-owned operator on EV/EBIT is invalid unless both EVs include NCI.

---

5. THE IMPLIED-PATH TEST — the Valuation Criteria

python3 reverse_dcf.py --spot 79.675 --shares 15.218789 --net-cash -490.298 \
  --revenue 795.488 --years 5 --wacc 0.095 --terminal-margin 0.105 \
  --exit-multiple 9.0 --fcf-margin 0.0841 --hist-cagr 0.122
  (asset sha256: a46b1c2c…;  net-cash = −(net debt $176.402m + NCI $313.896m))

Verdict: FAIL.

What the NCI omission alone is worth — run both ways

5.3 percentage points of required CAGR, from one omitted balance-sheet line. For scale, the interim-FCF defect that this project has documented at length was worth 5.1pp on STX and 5.3pp on WDC; the ORCL case moved required CAGR by 5.96pp. The NCI omission on USPH is the same order of magnitude as the largest previously catalogued arithmetic defect in the corpus, and it comes from a single line in temporary equity.

The screen reported +8.0pp because it also used a 20.5x exit multiple and a terminal margin of 11.1% (the FY2025 annual, not the TTM 10.0%). Corrected on all three inputs the margin is −8.9pp — a swing of 16.9pp.

The --fcf-margin sign

USPH's TTM consolidated FCF margin is +8.41% — positive, so omission would overstate required CAGR by roughly 1.8pp at this level (the tool's calibration: +2.13pp at a 10% margin). The flag was supplied. Terminal-only would have printed ~22.9%. FAIL either way, but the correct number is used.

A subtlety specific to USPH that is worth recording for the corpus: the $66.9m of TTM free cash flow is consolidated — roughly 34% of it belongs to the NCI partners and leaves the company as distributions, which run through financing, not operating. On an attributable basis USPH's FCF margin is nearer 5.5%. Because the reverse DCF here is run on a consolidated EV against consolidated revenue and margin, the consolidated FCF margin is the internally consistent input — but a memo that mixed an attributable FCF margin into a consolidated EV would understate required CAGR again, in the same direction as the original error.

Sensitivity over the exit multiple

The flip point against the trailing CAGR is roughly 14.5x — which is below the 15.6x the screen believed USPH traded at, and that is precisely why the screen passed it. Correct the EV and the name fails at every exit multiple below its own current trading multiple. Against same-clinic growth there is no flip point in the range at all.

---

6. TWELVE-MONTH TARGET — $97, +21.9%

And it points the opposite way to §5. That is not an inconsistency; it is the two-horizon design doing its job.

USPH's own EV/Sales history

Daily series 2020-07-27 → 2026-07-29 (n = 1,506), point-in-time TTM revenue (EDGAR-corrected, because AV's quarterly series is broken on this name) lagged 45 days, share count, net debt and NCI held at current verified values.

USPH trades at the 6.9th percentile of its own six-year history — as cheap, on its own record, as RMD. The 52-week range is $57.76–$91.69; spot sits 13.1% below the high and 37.9% above the low.

The target

• NTM revenue $852m — Street consensus. AV EARNINGS_ESTIMATES: FY2026 $838.7m / EPS $2.93; FY2027 $895.7m / EPS $3.41. Revision direction trailing 30 days: 0 up, 0 down — no revisions at all, the weakest revision signal of the three names here.
• Multiple: 2.31x = the 25th percentile of the three-year window (current 13.8th). Not the median — a business earning 6% on capital with same-clinic growth of 2.5% does not warrant its own median.
• Target EV = 2.31 × $852m = $1,968m; − net debt $176.4m − NCI $313.9m = $1,478m; ÷ 15.219m shares.

Target: $97. +21.9% to spot.

Named events inside 12 months (all in USPH_Catalyst_Calendar.md): the 2026 MPFS increase of +1.75%, the first positive Medicare year in six and already visible in a +0.8% Q1 net rate per visit; and the CY2027 proposed MPFS rule, typically published mid-year.

Why the two horizons disagree, and which to act on

• The 12-month target says: the multiple is at the 6.9th percentile of its own history, Medicare turns positive in 2026, and a partial mean-reversion is worth +21.9%.
• The 5-year implied path says: at a 6.0% ROIC and 2.5% same-clinic growth, today's correctly measured EV requires a 21.1% revenue CAGR that nothing in the business supports.

Both are true, because the multiple de-rated for a reason and the reason has not gone away. The 12-month output is a mean-reversion trade on a cheap multiple; the 5-year output is the ownership test, and it is the ownership test that criteria.md designates BINDING. The Valuation Criteria verdict is FAIL, and a +21.9% 12-month target does not override it.

Recording this explicitly because it is the first name in this corpus where the two horizons point in opposite directions, and the framework's answer — the implied path is the ownership test, the 12-month target is what a book trades on — is worth having on the record with a concrete case attached.

No external professional target is on file for USPH. Recorded as unavailable, not as agreement.

---

7. Criteria scoring

The memo issues no position verdict.

---

8. Downside case — named cause

Cause: the 2027 or 2028 MPFS returns to cuts, and USPH cannot offset it a sixth time.

The mechanism is arithmetic and it is already visible in the disclosed metrics. USPH's margin expansion in FY2025 came entirely from holding adjusted salaries per visit to +0.5% against a +1.0% rate increase — a 0.5pp spread. Medicare is 32.8% of PT net patient revenue and rising. Physical therapist wage inflation has run well above 0.5% in every general labour series; USPH held it there by managing mix and productivity, and average daily visits per clinic decelerated from +5.9% to +1.9% in Q1 2026, which is the productivity lever running out.

If the CY2027 MPFS cuts 2% and salaries per visit rise 3%, the PT segment gross margin falls roughly 1.5pp — about $10m of gross profit, or 13% of TTM EBIT — on top of an operating margin that has already fallen to 6.29% in Q1 2026. At a 7x exit on that base the equity is worth roughly $45–50, a −40% permanent impairment.

Probability: 30% — the highest of the three names in this run, because it requires only the resumption of a pattern that held for five of the last six years. Not a going-concern case: net debt is $176.4m against $66.9m of consolidated annual free cash flow and a senior credit facility with $161.8m drawn at 2025-12-31.

Second, structural case — the one that is already happening. Redeemable NCI grew 20.5% in five quarters ($260.0m → $313.4m) while attributable net income did not. USPH has the right but not the obligation to purchase non-controlling interests, and the 10-K lists as a risk the "impact on the business and cash reserves resulting from retirement or resignation of key partners and resulting purchase of their non-controlling interest." Each such purchase converts equity value into cash outflow at a redemption value the company states equals fair value. A company whose minority claim compounds at 20% while its attributable earnings do not is transferring value to its partners, slowly, in plain sight. Logged to the ledger for Brier scoring.