Tempus AI, Inc. [TEM] — Valuation Analysis (Task 3)

Valuation date 2026-07-27 · Spot $42.91 (Alpaca daily close; parent-supplied last trade $43.11) Model TEM_Model.xlsx — live formulas, verified by evaluating the workbook (see §1.1) Framework investment-memo v1.4.2, unmodified

1. The model and its tie-out

1.1 Verification — and what it caught

The skill's v1.4.2 rule is explicit: a zero balance check verifies internal consistency, not input accuracy; every actual-year line must be tied back to the filed statement. Excel automation is unavailable in this environment, so the workbook was verified by evaluating its formula graph with a calculation engine (formulas) and reading back computed cells — not by trusting the formula strings openpyxl wrote.

This caught a real defect. The first build carried a line labelled "Adjusted EBITDA" computed as operating income + SBC + D&A. For FY2025 that returns −$25,801k. The filed figure is −$7,385k. The model was internally consistent and wrong by $18.4m, because TEM's definition also adds back employer payroll tax on SBC, acquisition expenses, loss on debt extinguishment, franchise and other taxes, fair-value changes and equity-method losses, while subtracting amortisation of the SB Tempus technology licence. The line has been relabelled "simplified", a company-add-back row inserted, and an explicit tie-out check added.

Tie-out, FY2025A (model value vs. filed 10-K):

Computed WACC reads back at 13.63%; equity weight 86.2%; cost of equity 14.69%.

1.2 Cost of capital

This is a high discount rate. It is also the honest one for a business with 65.7% trailing realised volatility, a 2.50 regressed beta and negative operating cash flow. A sensitivity across 11.5–15.5% is given below precisely because this input carries the answer.

1.3 Base-case operating assumptions and the required base-rate check

Fading thereafter to 5.0% growth and a 16.0% terminal EBIT margin by FY2035.

Base-rate check (required). The empirical finding (Chan, Karceski & Lakonishok 2003) is that growth persistence beyond chance is close to nonexistent, and sustained 20%+ growth for 5+ years at multi-billion revenue scale is rare. This base case assumes ≥19% growth for three consecutive years past $1.6bn of revenue — a top-quartile, arguably top-decile, outcome for the reference class. It is adopted anyway, for two stated reasons: the FY2026 figure is essentially management's own guide, which consensus has independently matched; and the FY2027 figure is below Street consensus. The prior is being overridden, and this sentence is where that is named.

Note on GAAP breakeven. The base case does not reach GAAP operating breakeven until FY2029. That is not a bearish flourish — it is what the company's own FY2026 guidance implies: ~$65m of Adjusted EBITDA less ~$210m of annualised SBC less ~$105m of D&A ≈ −$250m of GAAP operating income, essentially flat on FY2025's −$253m despite 25% more revenue. The model reproduces −$249.6m.

2. DCF

2.1 Base case output (read back from the evaluated workbook)

2.2 Sensitivity ($/share)

(The $13.11 in this grid is the standalone Python replication; the workbook returns $12.80 on marginally different NWC seeding. The difference is immaterial to any conclusion.)

No cell in the plausible grid comes within 55% of the market price. That is the finding, and it needs confronting rather than presenting as a verdict.

2.3 Reverse DCF — what $42.91 actually requires

Each of these also assumes the bull revenue path — 26% fading to 7%, reaching $6.15bn by FY2035 (a 3.8x from FY2026, ~16% CAGR over nine years).

So the market price embeds: a ~16% nine-year revenue CAGR, a 20–24% terminal EBIT margin against −15.5% today, and an ~11% cost of capital on a stock with 66% realised volatility and a 2.5 beta. That is demanding but it is not absurd, and it is emphatically not "priced for perfection". Equally, it is not cheap on DCF at any defensible discount rate.

The decomposition that matters: the gap between $12.80 and $42.91 is almost entirely discount rate and terminal margin — not near-term numbers. House and Street FY2026 revenue differ by 1.1%. This is a disagreement about the multiple/required return, and §5 proves it rather than asserting it.

3. Comparable companies

*EXAS was acquired by Abbott and delisted 23 March 2026; its figures are the final trading metrics. It is included as a transaction comp, not a live trading comp.

Source: stockanalysis.com as of 28 July 2026; TEM gross margin from the filed Q1 2026 10-Q; returns computed from Alpaca daily bars.

3.1 Reading the comps honestly

The superficial read — TEM at 6.16x against peers at 14–18x is a 60% discount — does not survive contact with the EXAS datapoint. Abbott, a strategic acquirer with full diligence, paid 6.65x revenue for a $3.25bn-revenue, 69.7%-gross-margin diagnostics franchise in March 2026. TEM trades at 6.16x. Against the only completed control transaction in the set, TEM is roughly fairly valued, not deeply discounted. It is NTRA at 14.4x and GH at 18.4x that look like the outliers.

Where the discount to NTRA/GH is real, much of it is earned:

• NTRA generates positive free cash flow (+$104.5m) and holds net cash of $848m. TEM burns ~$73m a quarter and carries $589m of net debt.
• GH carries a large screening-optionality story (Shield) that TEM has no analogue for.
• TEM's FY2026 guided growth (+25%) is slower than either peer's trajectory, and its FY2025 headline 83% was only 33.5% organic.
• Neither peer has a 30:1 voting Class B, $171m of CEO selling, or 19% pending acquisition dilution.

Conclusion: the peer discount is substantially earned, and the residual unexplained discount is far smaller than the headline 60%. Any thesis resting on "it should trade like NTRA" is resting on the two most expensive names in a five-name set, one of which is FCF-positive and the other of which has a different product story.

3.2 Football field

The methods span $8 to $145 — a 17x range. This is the single most important valuation finding in the file: valuation carries almost no information on this name. Terminal value is 72% of the DCF; the comp set disagrees with itself by 3x; and the answer is set by whichever anchor is chosen. Any Task 5 conclusion that leans primarily on valuation is leaning on a number that cannot bear the weight.

3.3 Probability-weighted target

Built on FY2027E revenue of $2,091m (consensus $1,961m plus ~$130m of Personalis), post-deal net debt of $395m and post-deal shares of 214.7m — i.e. the deal is assumed to close in all three states.

The bull exit multiple of 7.0x is deliberately set below the NTRA/GH trading range and close to the EXAS transaction comp; the bear at 3.0x is where decelerating, cash-burning diagnostics assets have historically cleared.

Note for the calibration record: the house probability-weighted target of $43.93 sits 2.4% ABOVE spot ($42.91). This is materially different from the pattern logged in CALIBRATION_WATCH.md item B9, where 16 of 16 house targets sat below spot at a median −46.1%. It is consistent with B9's own exculpatory split — that "handed a name that has not run, the house prices it at fair value, not at a discount." TEM has not run; it has collapsed; and the house price is fair value. That is an out-of-sample confirmation of the exculpatory reading, and it is recorded as such.

4. Consensus and positioning (Alpha Vantage EARNINGS_ESTIMATES, one call, cached)

4.1 Estimates

4.2 The revision trend — and it is the wrong direction for a bear

Estimates have been revised upward monotonically over 90 days for FY2026 and over 60 days for FY2027. Revision counts on FY2027 over 30 days: 6 up, 5 down — genuinely mixed at the analyst level, but the average has moved up. Per Chan/Jegadeesh/Lakonishok 1996 this is a tailwind for a long and a headwind for a short, and it is happening while the stock falls 33%.

4.3 Consensus revenue equals guidance — the load-bearing observation

Company FY2026 guidance: $1.59bn–$1.60bn. Consensus FY2026 revenue: $1,592.9m, with a full 18-analyst range of $1,580.9–1,600.2m — a 1.2% spread, entirely inside the guided band.

Consensus is not an independent forecast; it is the guide, transcribed. House FY2026 of $1,610m is +1.1% against it. There is no near-term estimate variant available in either direction, for anyone, until guidance itself moves.

4.4 Street rating and target

16 analysts (stockanalysis.com, 28 Jul 2026): 7 Strong Buy, 1 Buy, 7 Hold, 1 Strong Sell. Consensus: Buy. Average target $66.57 (+55.1%), range $35.00–$100.00. A second aggregator (S&P Global via MarketScreener, 18 analysts) gives $66.06; a third reports $67.20 "down from $72.40". The count and average vary by source and that variance is disclosed rather than hidden behind one false-precise number. Recent actions: HC Wainwright $95; TD Cowen upgrade to Buy at $65; Needham reiterated Buy $75 after the Personalis announcement. Targets have drifted down (~$72 → ~$66-67) over recent months while ratings have not changed.

4.5 The required bridge: numbers, or multiple?

Per references/consensus-bridge.md, decomposed on a shared revenue base so the multiple is isolated as the variable.

The decomposition is unambiguous. House FY2026 revenue is +1.1% vs consensus; house FY2027 standalone is −0.7% vs consensus. On fundamentals house and Street are the same. The entire $42.91-to-$66.57 gap is the exit multiple: 4.59x versus 7.02x — a required 53% multiple expansion. Notably, the house bull case multiple (7.00x) and the Street average (7.02x) are the same number. The Street's base case is the house's bull case.

The real risk to the house view, stated plainly and not as a defence. If the correct anchor for a 63.8%-gross-margin, 25%-growing multimodal data platform is the NTRA/GH trading range (14–18x) rather than the EXAS transaction comp (6.65x), then a 4.59x anchor is systematically too low and this analysis will miss for as long as that regime holds. CALIBRATION_WATCH.md item B15 records that the exit-multiple anchor set for high-growth life-science names is unidentified — the clean anchors all grow 1–7.5% while the disputed names grow 15–39%, and a first-principles fade DCF implies the anchors themselves trade at ~2.5x their fundamental multiple. That criticism applies directly here and is not resolved by this memo. The 4.75x base multiple should be read as one defensible choice among several, not as a measurement.

4.6 Options market read

Realised volatility: 65.7% (1yr), 72.4% (3m). Implied is 77–93%. Options carry an 11–27 volatility-point premium to realised — the variance risk premium is unusually wide here, and it is widest in the short-dated puts (92.5% on the 14-Aug $43P), the most overpriced corner of the skew. Buying premium on this name is systematically negative expected value, which governs vehicle selection in Task 5.

Implied move through 14 August (bracketing the 30 July print): ATM straddle mid = $3.265 + $3.365 = $6.63 on $42.91 = ±15.5%.

5. Factor & Anomaly Scorecard (required)

Direction assessed for a contemplated long, since that is where the valuation discount, the Street target and the drawdown all point.

F-score detail: ROA>0 ✗ (−10.8%) · CFO>0 ✗ (−$218m) · ΔROA>0 ✓ (−76.2%→−10.8%) · CFO/TA>ROA ✓ · Δleverage down ✗ (debt $0.30bn→$1.23bn) · Δcurrent ratio up ✓ (2.29→3.13) · no new equity ✗ (shares issued for Ambry/Deep 6/Paige + $750m converts) · Δgross margin up ✓ (54.9%→62.7%) · Δasset turnover up ✗ (0.93x→0.79x). = 4/9.

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

They split, but the weight of evidence is negative for a long, and the two signals that point up are the two weakest.

Against: momentum (all three measures, all extreme), asset growth (+145.6%, an extreme reading on a well-documented negative predictor), F-score 4/9, short interest at a record. Four independent, decades-tested anomalies are pointing the same way.

For: estimate revisions (up), accruals (clean), the Q1 beat, gross profitability (solid but falling).

The two supportive quantitative signals are precisely the ones the framework flags as least reliable here: accruals are near-meaningless for a company with negative operating cash flow in both terms of the ratio, and estimate revisions on a name where consensus is a transcription of guidance measure how management guides, not how the business is performing.

The long-side corroboration standard in references/trade-construction.md Gate 1 asks for strong profitability + low accruals + high F-score. TEM delivers solid-but-declining profitability, clean-but-uninformative accruals, and an F-score of 4/9. That is a partial corroboration at best, and the +145.6% asset growth is a direct, quantified contradiction that must be stated rather than ignored.

5.2 Would this name have survived the screen?

Almost certainly not. idea-screener filters on momentum/trend after liquidity. TEM is bottom-quintile on 12-1 momentum, 29% below its 200-day, and at a 52-week low. This is a user-directed ticker, not a screen advance, and per the SKILL's Task 0 note that is base-rate information: the funnel that produced this name is "someone named it", which carries no prior of positive expected return.

6. Mention-frequency table (required in Task 3)

Full generative treatment is in TEM_Research.md §0. Required summary table, per-10k words, EDGAR 8-K Ex-99.1 corpus, 8 quarters (2024Q2–2026Q1) — the company's entire public history:

Prepared-remarks share: not computable. The corpus is a written release with no Q&A section, so the prepared-vs-prompted split required by references/mention-frequency.md cannot be produced. Stated rather than fabricated.

7. The four conclusions, kept separate

Per references/trade-construction.md, these are produced as distinct labelled conclusions before any Long/Short/Watchlist decision.

1. Fundamental conclusion. Revenue growing 25–36% with a genuine, evidenced Diagnostics gross-margin expansion (45.3% → 61.3% in five quarters) and Adjusted EBITDA inflecting from −$104.7m to a guided +$65m. Simultaneously: GAAP operating loss is widening (−$68.7m → −$84.7m YoY) and will be roughly flat in FY2026 on 25% more revenue; SBC is compounding at 129% against 36% revenue growth; the balance sheet has gone from net cash to $589m net debt with ~$73m/quarter of operating cash burn; and growth is heavily acquisition-assisted (FY2025's 83% was 33.5% organic). The Data & Applications gross margin — the segment carrying the entire premium-multiple case — is compressing 350bp YoY.

2. Expectations conclusion. Consensus FY2026 revenue ($1,592.9m) is company guidance ($1.59–1.60bn), across an 18-analyst range spanning 1.2%. House is +1.1% on FY2026 and −0.7% on FY2027 standalone. EPS estimates are being revised up on both years. The Street rates the stock Buy with a $66.57 average target. There is no material near-term expectations gap in either direction.

3. Valuation conclusion. Methods span $8 to $145. DCF at a defensible 13.6% WACC returns $12.80 with 72% of value in the terminal; the market price requires an ~11% WACC, a 20–24% terminal EBIT margin and a 16% nine-year revenue CAGR. Comps read 6.16x against 14–18x for two peers but 6.65x for the only completed control transaction in the set. Probability-weighted target $43.93, +2.4% vs spot. Valuation on this name is not informative enough to carry a directional thesis in either direction.

4. Portfolio conclusion. Deferred to Task 5 (TEM_Trade_Construction.md).

Sources: SEC EDGAR CIK 1717115 (FY2025 10-K, Q1 2026 10-Q, 8-K Ex-99.1 ×8, 8-K 2026-07-20, Forms 4, XBRL company facts); Alpha Vantage EARNINGS_ESTIMATES (single call, cached data/av_estimates_TEM.json); Alpaca Markets daily bars and options snapshots; stockanalysis.com (28 Jul 2026) for share count, short interest, ownership and peer multiples; ClinicalTrials.gov API v2; company press releases via EDGAR. Model: TEM_Model.xlsx, formula graph evaluated and read back — see §1.1.