PROCEPT BioRobotics [PRCT] — Valuation

As of 2026-07-29 · spot $18.09 · 56,918,844 shares · net cash $193.977m · EV $835.687m · TTM revenue $322.024m · EV/Sales 2.595x

Two outputs, per references/valuation.md. Neither replaces the other.

1. Implied-path test — the Valuation Criteria

Instrument: assets/reverse_dcf.py. Terminal value is 100% of modelled EV in this construction, so the reverse DCF is mandatory as the primary long-horizon output and the forward DCF is supporting evidence only.

Solved for: 5-year revenue CAGR. Held fixed: terminal EBIT margin 13.4%, exit multiple 22.8x EBIT, WACC 11%, horizon 5 years, net cash $193.977m, shares 56.919m, base revenue $322.024m.

1.1 The required parameter

The market requires a 5-year revenue CAGR of 7.4% (6.5% at WACC 10%, which reproduces the screen exactly).

1.2 The margin — demonstrated − required

The screen reports +53.7pp against a "demonstrated" 60.1%. That 60.1% is the FY22→FY25 three-year CAGR and it is not what this business demonstrates today. The margin is therefore reported against every honest comparator, and the strategy should rank on the most conservative one:

The margin survives every one of them. Even measuring "demonstrated" as the growth rate of the only line that is genuinely recurring, and after management removed the 8–16% of channel fill that inflated it, the business grows more than twice as fast as the price requires.

1.3 The exit multiple and the implied compression

• Exit multiple used: 22.8x EBIT, basis GROWTH_MATCHED, peer n = 97 (from the Tier-1 record).
• At the 13.4% terminal margin this is 22.8 × 0.134 = 3.06x exit sales.
• Today's trading multiple is 2.595x sales.
• Implied compression = 2.595 − 3.06 = −0.46x, i.e. an implied EXPANSION of +17.7%.

This is stated because it is not conservative and the framework requires the number rather than the adjective. The exit multiple assumes PRCT trades richer on sales in five years than it does today.

I could not re-derive the 97-name comparator set — the constituent list is not carried in the screen record. Per criteria.md that would ordinarily make the anchor UNIDENTIFIED. It is not left there, because §1.5 below re-runs the whole test on a sales basis anchored on PRCT's own current multiple, which requires no peer set and no terminal-margin assumption at all.

1.4 Sensitivity — over the exit multiple, never over scenario probabilities

Required 5-year revenue CAGR, WACC 11%, net cash $194.0m:

Where the answer flips. Against the recurring line's 16.2%, the test fails only in the top-left corner: a 12x exit on an 8% terminal margin requires 35.4%, and a 16x on 8% requires 27.9%. Everything at or right of a 20x exit passes on every terminal margin tested.

WACC (13.4% margin, 22.8x): 9% → 5.5% · 10% → 6.5% · 11% → 7.4% · 12% → 8.4% · 13% → 9.4%. The answer is insensitive to WACC across any defensible range.

Net cash — the parameter the instrument treats too kindly. The reverse DCF credits today's $194.0m in full, but PRCT burns it: FY25 operating burn $49.0m, and Q1'26 alone $38.1m.

Even at zero net cash the required CAGR is 12.0% — still below the recurring line's 16.2%.

1.5 Assumption-light cross-check — sales basis, self-anchored

This removes both contested inputs (the terminal margin and the peer-derived EBIT multiple) at once:

The cleanest statement in this memo: at an unchanged 2.6x sales multiple and no margin assumption whatsoever, today's price requires 11.0% revenue growth for five years. The business grew 20.2% last quarter, its recurring line grew 16.2%, and it guides 27–33%.

1.6 What the price requires if you insist on the growth

Inverting — what the price implies about profitability if growth is taken as given:

At the company's own guidance the price requires a 5.2% terminal EBIT margin — a fifth of Intuitive Surgical's. This is the real question in the name, and it is a margin question, not a growth question. PRCT has never earned an operating profit; FY26 guidance implies −22.5%; the FY22→Q1'26 record shows operating margin improving from −129% to −33.9%, so the direction is right, but Q1'26 delivered +0.5pp of YoY expansion, not the +11.2pp the FY26 guide requires.

1.7 Verdict

Valuation Criteria: PASS

Required 5-year CAGR 7.4% at the base parameters, 11.0% on the self-anchored sales cross-check, and 12.0% even crediting zero net cash. The most conservative honest measure of what the business demonstrates — its recurring handpiece line, post-destock, at +16.2% — clears all three.

This is a PASS, not a PASS WITH ARGUMENT: the required path sits below what the business has already shown, so no narrative is being asked to do work.

The named residual risk is the terminal margin, not the growth. At an 8% terminal margin and a 12x exit the required CAGR is 35.4% and the test fails. That corner is the bear case and it is priced nowhere in this analysis.

2. The 12-month target

2.1 The multiple anchor — and why the long history is unusable

PRCT's own EV/Sales history, built point-in-time (TTM revenue as it was known at each date, shares from the contemporaneous cover page, net cash from the last-filed balance sheet; scripts/prct_multiple_history.py, n = 1,107 sessions, 2022-02-28 → 2026-07-28):

Current 2.60x sits at the 0.5th percentile of the entire history.

The full-history anchor is declared UNIDENTIFIED.

valuation.md requires that a multiple anchor be rejected rather than used where the history "spans a regime change". This one spans two. A 16.5x median was set by a company growing 60–80% whose recurring revenue was running 8–16% ahead of consumption. The company that exists today grows 20%, has eliminated that practice, has reset guidance, and is defending a securities class action. Mean-reverting to the 16.47x median would be mean-reverting to a business that no longer exists — it implies $96.59 on today's TTM revenue, a +434% move. It is not used.

2.2 The post-reset regime — used, with its weakness stated

Since the 2026-02-25 Q4 print (n = 106 sessions):

Its weakness, stated rather than buried: corr(EV/Sales, time) = −0.56 within the window. This is a monotonically compressing series, not a series oscillating around a mean. A median over a downtrend is not a mean-reversion anchor, and it is not treated as one. The window is used as an observed range, not as a central tendency. Base case sits at 2.90x — between today's 2.60x and the 30-session median of 3.09x, below both the post-reset median (3.79x) and everything the name traded at before February.

2.3 Revenue at the target date (2027-07-31)

TTM revenue at the target date ≈ 0.45 × FY26 + 0.55 × FY27.

2.4 Net cash and shares at the target date — estimates, labelled as such

FY26 guided adjusted-EBITDA loss $17–30m; FY25's adjusted-EBITDA loss of $50.2m converted to $49.0m of actual operating burn, so the two track closely. Adding working capital (Q1'26 alone: AR +$12.9m, inventory +$6.8m) and ~$12–15m/yr capex, and running five quarters from 2026-03-31:

These are estimates, not disclosures. Share growth assumes ~2.5%/yr net issuance, consistent with 54.7m → 56.9m over the last five quarters. No equity raise is assumed in the bear case here — the raise is modelled separately in the Downside Criteria.

2.5 The targets

12-month target: $25, +38% to spot

Decomposition: +24.5pp comes from revenue growth alone at today's unchanged 2.60x multiple; the remaining ~13pp comes from partial reversion to 2.90x, which is still below the 30-session median of 3.09x, below the post-reset median of 3.79x, and at the 1.17th percentile of PRCT's full trading history (2.35x → 0.09th · 2.60x → 0.45th · 3.09x → 1.36th · 3.79x → 4.88th · 4.54x → 10.66th). Most of the target is earned by the numerator.

2.6 Named product-cycle events inside the 12 months

Each appears in PRCT_Catalyst_Calendar.md with its date and its source:

1. Q2'26 results — Tuesday 4 August 2026, after close (company-announced 2026-07-14). The single largest estimate-moving event: it is the first clean quarter with the handpiece:procedure ratio at ~100%, and it settles whether the FY26 guide survives the Q1 seasonal-share arithmetic.
2. CPT 52597 (Category I) in its first full year, live since 1 Jan 2026, with APC 5377 opening an ASC payment pathway. Unmodelled by management and unmentioned in ten consecutive earnings releases — a source of upside surprise rather than of expected revenue.
3. FirstAssist AI 2nd generation, FDA-cleared April 2026, now rolling into the HYDROS base.
4. FY2027 guidance, expected with Q4'26 results (~late February 2027, by three-year filing pattern).

No WATER IV readout date is asserted; none is disclosed.

2.7 Sanity band against external targets

The base target coincides with Truist's and sits 25% above UBS's. This is a check on the output, not a calibration target — the number was built from PRCT's own multiple range and its own guided revenue before either external target was consulted. The divergence from UBS is entirely the multiple: $20 corresponds to ~2.55x on the same FY27 revenue, i.e. UBS assumes no re-rating at all, which is this memo's "no re-rating" case at $22.5.

3. Forward DCF — deliberately not produced

Terminal value is 100% of modelled EV on this name; there is no interim free cash flow to discount, because PRCT does not generate any and does not guide to any inside the forecast window. A forward DCF here would be a reverse DCF with the answer pre-filled and the sensitivity run on the parameter that cannot change it — the exact failure valuation.md was written to stop. The implied-path test above is the long-horizon output.

4. Reproducibility

Two defects found and fixed while building the multiple history, recorded because the framework's premise is that defects are the primary output: 1. SEC companyfacts retains only the latest filing that restated an old fiscal year, so a Q4 derived from a 10-K inherits that filing's date. Using it as the "publicly known from" date pushed every TTM window containing a Q4 years forward and produced EV/Sales medians of 50–120x. Fixed by falling back to the actual reporting lag whenever the recorded filing date is >100 days after period end. 2. A four-quarter TTM window was validated by requiring ~365 days between the first and last period end — but four consecutive quarter-ends span ~274 days, so the check silently admitted windows with a missing quarter and rejected every valid one. Both errors were confidently wrong, not loud — the same failure mode the brief warns about in the scanner.