Phase Space AI

Valuation and Trade

Guardant Health [GH]

Guardant Health [GH] — Valuation, Criteria Scorecard, Trade Construction & Catalysts

Phase Space Research | original 2026-07-27 · methodology update 2026-07-29 | framework v1.7.0 (Criteria ruleset, references/criteria.md 2026-07-29) | Price at update $143.48 (2026-07-28 close) | EV $19.31bn

This memo no longer issues a position verdict. It scores Criteria and produces two valuation outputs. Sections written under the retired "Gate / E[R] vs cash hurdle" ruleset are marked SUPERSEDED where their conclusions no longer stand. Their evidence is retained deliberately.

---

0. Methodology update — 2026-07-29. What changed, and why the answer moved

0.1 The old conclusion, and the defect that produced it

The 2026-07-27 memo concluded "WATCHLIST — no position; probability-weighted value $88.95 against a $145.47 spot; scenario-weighted E[R] −38.9% against a +4.7% cash hurdle, a −43.6pp shortfall." Two defects produced that number:

1. The exit multiple came from a comparator set with no growth dispersion. Base 6.0x sales was anchored on the Abbott/Exact precedent (~7x) and a "listed diagnostics/tools peer median ~8x" — a set spanning 1.0–7.5% revenue growth (ILMN 1%, QGEN 3%, BRKR 2%, DGX 7.5%, LH 7.0%) used to value a company compounding near 30%. A growth/multiple slope cannot be estimated from a sample with no growth dispersion, and it was then extrapolated far beyond its support. Base was also set below both stated anchors, a distant-year haircut stacked on a multiple that had already mean-reverted.
2. The sensitivity ran over scenario probabilities while the exit multiple got a point estimate. The multiple determined the answer and was the one parameter never varied.

0.2 Inputs — pre-computed, and the three corrections made to them

From the pre-computed scan at reports/scan/GH_analysis.json (as-of 2026-07-28) rather than recomputed here. Three fields are wrong and are corrected, with the correction stated rather than silently applied:

Resulting EV: $19,314m = 17.9x TTM revenue.

0.3 The implied-path test (reverse DCF) — the Valuation Criteria

assets/reverse_dcf.py, solving for revenue CAGR. Held fixed: terminal EBIT margin 20.0%, exit multiple 46.8x EBIT, WACC 10.0%, horizon 5 years, EV $19,314m, revenue base $1,080.2m.

What the price requires: a 25.2% five-year revenue CAGR.

What Guardant has demonstrated: 29.6% over three years (32.0% over two).

MARGIN = demonstrated − required = +4.4pp on the 3-year basis (+6.8pp on the 2-year). Valuation Criteria: PASS — but thinly, and the thinness is the finding.

The exit multiple and its compression, as a number. 46.8x EBIT at a 20% terminal margin is 9.36x exit-year sales, against 17.88x trading today — the path embeds a −47.6% multiple compression. Basis GROWTH_MATCHED, n=10: BSX (16.5%, 20.0x), DXCM (17.0%, 30.9x), ISRG (17.4%, 40.2x), PEN (18.3%, 65.7x), PANW (18.8%, 185.4x), NOW (22.4%, 57.4x), AVGO (24.4%, 64.1x), FSLR (25.8%, 13.7x), ANET (27.1%, 53.4x), GMED (42.2%, 22.0x). Disclosed weakness: growth-matched but not sector-matched, and the dispersion inside the set is enormous (13.7x to 185.4x), which is why §0.4 matters more here than for any of the other three names.

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

Break-even exit multiple: 39.4x EBIT / 7.88x exit sales at the demonstrated 29.6% CAGR. GH's Valuation Criteria verdict flips to FAIL if the true exit multiple is more than ~16% below the growth-matched anchor. That is a genuinely marginal PASS and it is labelled as one. Contrast NTRA, which passes at 0.6× the anchor, and TXG, which fails at 1.5×.

Second sensitivity. Holding 29.6% growth and 46.8x, the break-even terminal EBIT margin is 15.4% — against the flat 20% assumed and against Guardant's actual −44.5% FY2025 operating margin. This is the least robust of the four names in this batch on both parameters, and that is the honest reading.

0.5 Caveat on the terminal margin

The scan does not persist terminal_margin; back-solving from its published required_cagr_pct returns exactly 0.2000 for NTRA, GH and TWST alike — a flat 20% for every name, not a per-name derived figure. A 20% terminal EBIT margin for a company running −44.5% today is an assumption, not a measurement, and for GH the break-even (15.4%) sits uncomfortably close to it.

0.6 The 12-month target

Built per references/valuation.md: near-term consensus revenue, multiple anchored on Guardant's own trading history and percentile.

• Revenue base: TTM $1,080.2m grown at the Street's implied FY2026 growth (+33.4%: consensus FY2026E $1,310m, 25 analysts, cached 2026-07-27, vs FY2025 actual $982.0m) → NTM $1,441m. Disclosed gap: a direct NTM consensus blend could not be built — FY2027E revenue is not on file and the Alpha Vantage quota (25/day, shared) was exhausted on 2026-07-29. Per the Consensus Criteria a quota gap leaves the field blank and blocks nothing.
• Multiple anchor — Guardant's own EV/Sales history, 2022-07-01 to 2026-07-28 (1,021 trading days, EV rebuilt daily from then-current shares and then-reported TTM revenue, net of the convertible debt): today 17.72x = the 98th percentile of the four-year range (min 3.70x, median 7.85x, upper quartile 11.86x, max 21.02x). Against the trailing twelve months only, 17.72x is the 92nd percentile (1y median 12.47x).

GH is the one name of the four where today's multiple sits at the 98th percentile of its own four-year range and the 92nd of its own last year. There is nowhere in its own history that supports today's multiple as normal. That is the central 12-month fact.

Named 12-month product-cycle events feeding it (dated in §4): Q2-2026 earnings 30 July 2026; the first UnitedHealth-inclusive quarter reporting late October 2026; SHIELD Lung (NCT05117840) primary completion 31 Dec 2026, n=12,000 — registry-dated and in no consensus revenue line.

Sanity band, and a material divergence that must be explained. Street average target $171.08 (+17.6%). The external healthcare specialist's book (snapshot 2026-07-28) carries GH long at a +39% expected return — his highest-return long in the overlap set — on theses of "Oncology upside; SHIELD upside; Infinity AI partnership optionality." Our base target is −7.1% and his is +39%: a 46-point divergence, the widest in this batch. The explanation is identifiable rather than a calibration failure: his target capitalises SHIELD screening revenue that is not in the FY2026 consensus revenue base this target is built on, and it does so at a multiple GH has never traded at. Both can be true — the divergence is a statement about whether SHIELD Lung reads out, which is a dated binary on 31 Dec 2026, not about the arithmetic. We report the gap; we do not tune the multiple to close it.

0.7 Criteria scorecard (Gates retired 2026-07-29)

Archetype: INFLECTION.

0.8 What moved, in one line

Old: worth $88.95 against a $145.47 spot; E[R] −38.9%; WATCHLIST. New: the price requires a 25.2% revenue CAGR against 29.6% demonstrated — a +4.4pp margin, a marginal PASS — with a 12-month base target of $133 (−7.1%). Why it moved: the exit multiple is now drawn from comparators that bracket Guardant's growth (46.8x EBIT ≈ 9.4x sales versus 6.0x sales pulled from 1–7.5% growers), and the question changed from asserting a path and a multiple to solving for what the price requires. Offsetting the anchoring fix, the balance-sheet correction moved against the name: Guardant is in a $289m net-debt position, not the $355m net cash the scan reported. This is the least robust of the four: it fails at 0.8× the anchor multiple and at a 15.4% terminal margin.

---

1. Valuation (SUPERSEDED as to conclusions — retained as evidence)

1.1 DCF — reported, and deliberately down-weighted

Why the forecast runs ten years and not five. GH's unlevered FCF does not turn positive until FY2030E. A five-year DCF terminating in the first marginally-profitable year capitalises a near-zero cash flow and produces a negative equity value — an artifact of the window, not a valuation. Extending to FY2035E gives a normalised terminal year. (The five-year version was built first and produced −$4.22/share; that number is a modelling artifact and is recorded here only to explain why the window changed.)

This DCF is a low-weight input and the memo treats it as one. Terminal value is 98.8% of enterprise value. Per the SMR precedent, for a pre-profit name a DCF is often the wrong tool; the scenario framework below and the liquidity analysis carry the weight.

Sensitivity — every cell is far below spot:

(The grid fully re-discounts every forecast-year cash flow and the terminal value at each row's WACC. The $23.75 headline uses the base-case column at the model's own WACC; grid cells differ slightly because the grid re-solves rather than scaling.)

1.2 Scenario framework — this carries the weight

Valued on FY2030E revenue at an anchored exit multiple, discounted back 4.5 years at 9.14%.

SUPERSEDED (2026-07-29). Probability-weighted value $88.95 · Scenario-weighted expected return −38.9% · versus a 4.7% cash hurdle, a shortfall of −43.6 percentage points. This output is retired. The cash hurdle no longer exists in the framework, and the exit multiples below (3.5x / 6.0x / 9.0x) were anchored on a 1.0–7.5% growth peer set applied to a ~30% grower. The Shield volume and ASP build is real work and is kept; its valuation conclusion is not. Replacement: §0.3.

Where the exit multiples come from — they are anchored, not picked. Base 6.0x rests on two real anchors: (1) Abbott paid ~7x sales in cash for Exact Sciences, closing 23 March 2026 — a strategic acquirer's price for the incumbent US CRC screening franchise; (2) the listed diagnostics/tools peer median is ~8x today. A 6.0x multiple applied to a year five years out sits below both, which is the right direction of travel.

The conclusion is robust to that choice, which is what makes it usable. Holding base FY2030E revenue at $3.2bn: $61 at 4.5x · $86 at 6.0x · $101 at 7.0x (the Abbott precedent) · $115 at 8.0x (today's peer median). Even at today's peer median applied five years forward, the stock is 21% overvalued. Today's $145.47 requires roughly 10x EV/Sales on FY2030E revenue — above the price a strategic just paid in cash, and above the current peer median applied to a distant year.

1.3 Comparables

GH carries the highest multiple in the set at roughly the same growth rate as Natera, and roughly 3x the price a strategic acquirer just paid in cash for the category incumbent.

1.4 Consensus bridge (required)

Decomposing the disagreement — is it numbers or multiple? It is entirely the multiple, and this is provable. On revenue the house and the Street are 2.8% apart, which is inside forecast noise. The entire $82 gap between the Street's $171.08 and the house's $88.95 comes from the exit multiple applied to a broadly agreed revenue path. The Street is implicitly capitalising FY2030E revenue at ~11–12x; the house uses 6.0x, anchored on a cash takeover at ~7x and a peer median at ~8x.

The real risk if the Street is right: scarcity value. There is exactly one FDA-approved blood-based primary CRC screen, it is in ACS and NCCN guidelines, and a $21bn strategic just demonstrated that these franchises get bought. A monopoly asset in a reimbursement inflection can hold a multiple that looks indefensible on a spreadsheet for a very long time — and can be acquired at a premium to it. That is a genuine, non-trivial argument, and it is why this is a Watchlist and not a Short.

---

2. Factor & Anomaly Scorecard

Universe percentiles computed from a reconstructed liquid US universe (2,929–4,061 names depending on the liquidity filter; monthly Alpaca SIP bars).

Synthesis. The anomaly factors and the fundamental view split, and they split in the direction that kills the short. Momentum, trend, 52-week-high proximity, accruals and revisions are all supportive of ownership; only asset growth is negative, and that is mechanically inflated by a financing event. A short here would be fighting the most robust documented anomaly in the record while the accruals signal — the one that actually predicts where shorts work — says earnings quality is clean. That combination is close to the textbook profile of a bad short.

2.1 Screen provenance — correcting the record

The brief stated that GH "did not clear the top-8% momentum shortlists in either direction" on the 2026-07-27 screen. That is not what happened, and the distinction matters.

Reconstructing the universe from monthly Alpaca SIP bars: GH's 12-1 momentum of +188% (monthly-bar basis) ranks in the top 6.66% of 2,929 liquid names — comfortably inside the top 8% long shortlist. The result is robust: top 5.91% at an ADV>$1m filter, 6.46% at >$2m, 6.66% at >$5m, 6.78% at >$20m.

GH did not appear in the screen output because Stage C was capped at the top 75 names per side (stage_c_cap: 75), which on this tape required 12-1 momentum above +405%. The top-8% threshold was +160%. GH cleared the stated threshold and was truncated by the cap.

This inverts the framing the batch was set up with. These are not neglected names from a broken sector that the screen passed over; GH, TWST and TXG are all top-decile momentum names in a sector that has ripped, and they were invisible only because an unusually violent tape pushed the scoring cap to +405%. That is a finding about the screen's configuration, and it is logged for the screener skill.

2.2 Correlation vs. the AI-infrastructure cluster

Trailing 252-day daily returns:

Mean correlation to the ten-name AI-infrastructure cluster: +0.071 (range −0.05 to +0.20). Versus SPY: +0.20. Versus XBI: +0.35.

GH is the most genuinely uncorrelated name in the pipeline — materially lower than TWST (+0.199) or TXG (+0.211). And, exactly as the ISRG memo concluded, that is a reason to prioritise it for conversion, never a reason to own something with a negative standalone expected return. A −38.9% expected return does not become acceptable because it is uncorrelated. Diversification scales a positive edge; it cannot manufacture one.

---

3. Trade Construction — the four conclusions, kept separate

1. Fundamental conclusion — POSITIVE. Shield volumes compounding 9k→44k per quarter; realised ASP held at $945 through the PAMA reset; UnitedHealth coverage live 1 July 2026; gross margin +380bp y/y; operating burn down three years running. This is a business executing.
2. Expectations conclusion — FULLY DISCOUNTED. 25 analysts at Strong Buy, average target $171.08 above spot, consensus revenue within 2.8% of the house build, stock +220% in twelve months. There is no gap.
3. Valuation conclusion — RESTATED 2026-07-29. (Was: "NEGATIVE — probability-weighted value $88.95 vs $145.47.") On the implied-path test the price requires a 25.2% five-year revenue CAGR against 29.6% demonstrated: a +4.4pp margin, Valuation Criteria PASS — but marginal, flipping to FAIL if the exit multiple is more than ~16% below the growth-matched anchor. 12-month base target $133 (−7.1%). See §0.3–§0.6.
4. Portfolio conclusion — REMOVED. (Was: "NO POSITION.") This memo issues no position verdict. It scores Criteria (§0.7) and produces two valuation outputs. Whether that justifies a position belongs to a book.

3.1 Criteria-by-Criteria (evidence retained; verdicts restated in §0.7)

"Gates" were retired on 2026-07-29. Below is the original gate block with names mapped forward. The reasoning column is unaltered — it is the record. Where a verdict changed, §0.7 is authoritative.

3.2 Analysis summary — no position verdict is issued

POSITION VERDICT: NONE. This memo does not output Long / Short / Watchlist / Avoid.
                  It outputs an analysis. Whether it justifies a position belongs to a book.

VALUATION CRITERIA (implied path, 5y):  price requires 25.2% revenue CAGR
                                        demonstrated 29.6% (3y)   MARGIN +4.4pp   PASS - MARGINAL
   exit multiple 46.8x EBIT = 9.36x exit sales, basis GROWTH_MATCHED (n=10)
   implied compression from 17.88x trading today: -47.6%
   FLIPS TO FAIL below a 39.4x EBIT exit multiple, i.e. ~16% below the anchor,
   and at a terminal EBIT margin below 15.4% against the 20% assumed
12-MONTH TARGET: $133 base (-7.1% to $143.48 spot); low $83 (-42.1%); high $190 (+32.7%)
                 multiple anchored on GH's OWN EV/Sales history (17.72x = 98th pctile of 4y,
                 92nd pctile of 1y); revenue base NTM $1,441m off Street FY2026E
QUALITY CRITERIA (BINDING): PASS on all three INFLECTION limbs - gross margin 64.5%,
                 operating margin +15.5pp YoY, growth 29.6%. Strongest of the four in this batch.
LIQUIDITY CRITERIA (BINDING): PASS on equity; INDETERMINATE on options (single listed expiration,
                 chains previously found thin, open interest not re-pulled on 2026-07-29)
MOMENTUM CRITERIA (MEASURED - ENTRY TIMING ONLY, NEVER SELECTION):
                 12-1 +247.2% (87.0th pctile), 6-1 +34.3% (72.2nd), RSI-14 36.6
DOWNSIDE CRITERIA (MEASURED): SHIELD Lung missing its 31-Dec-2026 endpoints, against $1,503.5m of
                 convertible notes and $437m of annual operating loss - the net-debt position is
                 what turns a de-rating into a financing question
CATALYST / PEER SPREAD / CONSENSUS / SHORT MECHANISM: see §0.7
TIME HORIZON:    Q2-2026 earnings (30 Jul 2026), then the SHIELD Lung readout (by 31 Dec 2026)
CONSENSUS:       Street Strong Buy, 25 analysts, average target $171.08 (+17.6%).
                 External reference book: long GH at a +39% expected return - a 46-point
                 divergence from our base target, explained in §0.6, not tuned away.

What is deliberately absent. No entry price, no sizing, no vehicle recommendation and no "conviction" tier. Those are strategy outputs. §3.3's priced structures are retained as liquidity evidence, not a recommendation.

This is the NET case, and it is named as such. The trade-construction reference warns that tech/growth names hit the "expensive but not a short" trap repeatedly, and that a rich multiple is more often earned in a genuine regime change. GH may well be one: FDA monopoly, guideline inclusion, a reimbursement step-change, and a strategic acquirer paying $21bn for the incumbent four months ago. I am skeptical of the multiple. Scored honestly, skepticism is not Gate 1 or Gate 2 evidence, and it is not dressed up here as if it were.

3.3 Vehicles priced, and why each was rejected

Alpaca options data pulled live (2026-07-27). GH's only listed expiration in the window is 16 October 2026, which does capture the 30 July print.

Variance risk premium check (required before any naked premium). ATM October IV is 56–67% against 60.7% realised — so premium is around fair on calls and rich on the put side, where 140-strike IV of 56.5% sits below call IV, i.e. the skew is unusually flat and offers no cheap downside. There is no structure here whose implied move is materially below a house expected move, because the house has no directional variant to express. The correct answer to "which vehicle" is: none.

IV crush would apply to any of the above: GH reports 30 July, three days out, and October IV will collapse immediately afterwards regardless of direction. Flagged, though moot given no position.

3.4 Portfolio-book context (portfolio_book.json contract executed)

Book is 100% cash; config: 5% single-name max, 4.7% cash hurdle, −2%-of-book position hard stop, 25% sector concentration cap, 0.60 pairwise-correlation disclosure threshold.

• Capital competition — SUPERSEDED. (Was: "FAILS. Net E[R] of −38.9% versus a risk-free 4.7%.") The cash hurdle has been retired from the framework; capital competition is slot competition inside the strategy and the memo does not run it. On the Criteria the memo does score, the Valuation Criteria passes at +4.4pp — marginally.
• Correlation: no constraint breached (mean +0.071 to the AI cluster; no existing positions).
• Sector concentration: a first life-science-tools position would open a new cluster — genuinely valuable for a book that is otherwise entirely one factor. Not a reason to buy at this price.
• Write-back: Watchlist entry with two-sided conversion triggers (below).

Conversion triggers - → LONG if price reaches ~$86 (probability-weighted value) without fundamental deterioration; or if SHIELD Lung (NCT05117840) reads out positively by 31 Dec 2026 and management guides FY2027 screening revenue above ~$600m, which would move the Base case toward the Bull path. - → SHORT only if all of: (i) Shield quarterly volume misses the 230–245k FY2026 guide, or realised ASP falls below ~$800 (fixing Gate 1); (ii) FY2027 consensus revenue revisions turn net negative for two consecutive months (fixing Gate 2); and (iii) a weekly close below the 200-day moving average (~$106) (fixing Gate 6). All three, not any one.

---

4. Catalyst Calendar

---

5. What would make me wrong

1. Scarcity value is a real, durable phenomenon in diagnostics. Monopoly FDA-approved franchises in guideline-included categories have historically sustained multiples that look indefensible for years.
2. Takeout. Abbott just paid $21bn for the incumbent. GH is the only FDA-approved blood-based primary CRC screen in existence. A strategic bid at a premium to $145 is a live possibility and would make the Bull case academic.
3. The exit multiple is doing all the work. At 10x FY2030E EV/Sales — which is roughly where the market is — the stock is fairly valued. My case that 6.0x is right rests on an M&A precedent and a peer median, both of which are anchors, not laws.
4. SHIELD Lung could be transformational and I have deliberately given it no value in the base case, because the registry gives a date but no readout.

---

Prepared by Phase Space Research, 2026-07-27. Options quotes, Greeks and IV: Alpaca (live, 2026-07-27). Momentum percentiles: reconstructed liquid universe, Alpaca SIP monthly bars. Consensus: public aggregators. Model: GH_Financial_Model.xlsx (verified by reading back computed cells in Excel; balance-sheet check is zero in all twelve columns).