NuScale Power [SMR] — Criteria & Two-Horizon Valuation

Phase Space Research · 2026-07-29 · Framework: Criteria (references/criteria.md, 2026-07-29), superseding the gate framework v1.5.1 · Archetype: INFLECTION

This document supersedes the Gate block and the Gate 4 expected-return arithmetic in NuScale_Trade_Construction_2026-07-27.md and NuScale_Valuation_Analysis_2026-07-27.md. The liquidity-runway analysis, the ENTRA1 fee mechanics and the mention-frequency work in those files stand unchanged.

There is no position verdict in this document. The previous manifest carried investment_decision: "Watchlist"; that field is retired.

0. What moved, and why

1. Data hygiene — two corrections, both material

1.1 Share count — the automated figure was wrong by 54%

Shares used: 365,481,156 — 346,105,785 Class A + 19,375,371 Class B, cover page of the Q1 2026 Form 10-Q, as of 2026-04-30.

The 2026-07-28 universe scan returned 236,754,948. That figure is us-gaap:CommonStockSharesOutstanding with a period end of 2022-05-03 — a cover-page figure from the 2022 10-K, four years stale, surfaced because NuScale's current cover-page count is tagged dimensionally per share class and therefore does not aggregate into dei:EntityCommonStockSharesOutstanding at all. Querying that concept directly returns HTTP 404.

This is the same trap already documented on TEM in coverage_scan.py: "dei cover-page count tagged DIMENSIONALLY per share class, so SEC aggregation surfaces nothing — any dual-class registrant can have a correct tag no companyfacts consumer ever sees." The EntityPublicFloat sanity floor did not catch it, because a 54% share-count error is not a 10x error.

NuScale is an Up-C. Class B units are exchangeable 1:1 into Class A, so total economic shares is the correct enterprise-value base. Using 236.75m would have understated market capitalisation by $1.06bn and made the company look far cheaper than it is.

The verified count ties to the existing memo: 365,481,156 × $8.29 = $3,029.8m, exactly the market capitalisation on the memo's cover line.

1.2 TTM revenue — the automated figure was 62% too high

The universe scan returned $30.236m. coverage_scan.ttm_revenue() sums "the last four quarterly periods" by end-date without checking they are consecutive; because US registrants do not tag Q4 separately, it summed 2025Q1 + 2025Q2 + 2025Q3 + 2026Q1 — a 15-month gap-riddled window. The existing memo quoted $22.7m, also different; $18.669m is the figure computed here from the filed statements.

The direction of travel is the finding, not the level. Revenue declined FY2024 $37.045m → FY2025 $31.479m, and Q1'26 was $0.565m against Q1'25 $13.375m — a 96% year-on-year fall. The Q1 disaggregation shows Power Plant and NPM related services collapsing from $12.969m to $0.487m.

1.3 Net cash

$1,008.2m, with zero borrowings. At 2026-03-31: cash $341.129m + short-term investments $549.000m + non-current investments $118.634m = $1,008.763m, excluding $5.100m of restricted cash. Total liabilities are $38.430m. This confirms the existing memo's $1,008.2m.

Net cash is $2.76/share — 33.6% of the market capitalisation.

2. The Criteria

2.1 Quality Criteria — INFLECTION standard

All three limbs are evaluable and all three fail. This is a FAIL, not an INDETERMINATE — no input is missing. That distinction is the whole point of the D1 correction: INDETERMINATE is for absent data, and NuScale's data is present and clear.

criteria.md: "Cheap cannot rescue a failure here. This is what stops value traps." The 33.6% of market capitalisation sitting in cash is the cheapness argument, and under the stated rule it does not rescue the Quality Criteria.

3. Valuation Criteria — the two mandatory outputs

The existing memo declined to build a DCF at all, on the reasonable ground that "a discounted cash flow model requires a cash flow to discount" and "every line of a five- or ten-year DCF on this company would be an assumption." The reverse DCF does not have that problem, because it does not require the analyst to supply the path — it solves for it. That is precisely why the framework mandates it here.

3.1 The implied-path test

What the price requires: a 124.2% revenue CAGR for five years.

Demonstrated: 15.1%. Margin = 15.1 − 124.2 = −109.1pp.

That would take revenue from $18.7m to roughly $1.06bn by 2031 — a 57-fold increase — against a company that delivered $565,000 of revenue in its most recent quarter and whose first commercial module delivery is 2031 at the earliest.

Demonstrated CAGR window: FY2022 $11.804m → TTM 2026-03-31 $18.669m, 3.25 years = 15.1%. This is already generous to the name: it measures from a low base and it ends on a declining trend. Measured FY2024 → TTM the demonstrated rate is negative.

Implied compression, as a number. NuScale is loss-making, so the comparison is made on sales: current EV/Sales 106.7x against the exit multiple restated on sales (22.5x × 13.5% = 3.04x) — a −97.2% compression.

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

This is the section the parent brief anticipated, and the answer is the one it asked for plainly.

The price cannot be justified anywhere in the plausible parameter range, holding TTM revenue and the growth-matched terminal margin fixed.

The margin is negative at every exit multiple tested, and the flip point — the multiple at which today's price is justified by what NuScale has actually demonstrated — is 178.3x EV/EBIT. That is roughly eight times the growth-matched anchor of 22.5x and sits far outside any observed comparator. The reverse DCF does return a solution; the solution is simply that the required path lies outside the achievable range, and that is reported as the finding rather than smoothed into a number.

Valuation Criteria: FAIL, by the widest margin of the four names in this batch.

A note on what would change this, and it is not the multiple. Because 33.6% of the market capitalisation is cash and the revenue base is $18.7m, the implied-path test is extraordinarily sensitive to the revenue input and almost insensitive to the multiple. The entire valuation question for NuScale is whether the ENTRA1 / TVA programme converts — a binary, contractual event — not what multiple a converted business deserves. The reverse DCF's honest output here is that a multiple-based instrument is the wrong tool for this name, which is a stronger version of what the original memo said when it declined to build a DCF.

3.3 The ENTRA1 asymmetry is preserved

The existing memo's central analytical claim is unchanged by any of the above and is restated because the implied-path test does not capture it:

The event the Street is waiting for — a binding TVA power-purchase agreement — is simultaneously the event that crystallises a cash obligation larger than NuScale's entire balance sheet. The ENTRA1 fee at 77.9 modules × $16m is $1,246.8m, or 123.0% of total cash and investments, forcing a 12–25% equity issue. The good news and the funding hole arrive in the same 8-K.

Liquidity runway with no further milestones: 23.6 quarters = 5.9 years at a $43.0m quarterly underlying burn. NuScale is not going bankrupt, and no bear case here is built on insolvency.

3.4 The 12-month target

The own-multiple anchor is reported, and the target band it produces is dominated by something else.

Over 844 sessions, SMR's EV/Sales on an as-known annual revenue basis ranges min −26.0x (its enterprise value went negative when the equity traded below net cash) / p25 86.1x / median 132.6x / p75 227.4x / max 500.2x. It now trades at 63.3x — the 17th percentile of its own history. On the multiple alone, SMR is cheap against its own past.

But the target is not set by the multiple. It is set by an 8.1x disagreement about revenue.

12-month target: $5.15 – $22.17. The band is 4.3x wide and none of that width comes from the multiple — it comes entirely from whether NuScale earns $53.5m or $6.6m in FY2026.

The Street's own figure is the one under pressure. To reach $53.5m, NuScale must produce $52.9m across Q2–Q4 against a Q1 run-rate of $0.565m. That is a checkable, dated claim and it is the single most informative number in this document. NTM revenue is taken as the FY2026E Street figure of $53.5m alone, because no FY2027E consensus is documented in the file and none has been invented.

EV / NTM revenue today: 37.2x on the Street figure.

Note the direction. The previous house target was $6.75, below spot. The new band straddles spot and its upper bound is +170% — not because the view of the business improved, but because the target is now anchored on the name's own multiple history rather than on a 5-year DCF, exactly as item B16 requires. The Quality and Valuation Criteria still both FAIL. A 12-month target above spot and a failing implied-path test are not contradictory: they answer different questions over different horizons, which is why valuation.md requires both.

4. Downside Criteria — MEASURED, blocks nothing

Logged and scored; it rejects nothing. Inverse-volatility sizing is the active control — realised volatility is 100.0%.

5. Momentum Criteria — MEASURED, entry timing only

Momentum governs when, never whether. All veto and blocking language is deleted.

SMR sits in the bottom quintile of the cross-section, below the −39.8% bottom-quintile threshold that the universe scan computes as the entry-staging trigger. Under the current framework this is a statement about timing and nothing else. It does not reject the name — the Quality and Valuation Criteria already do that, on their own evidence, and they would do so identically if momentum were top-decile.

That separation is the point criteria.md makes about this Criteria being "the change most likely to be silently reversed under pressure." Here the temptation runs the other way: it would be easy to let a −79.9% tape stand in for the analysis. It does not.

6. What this document deliberately does not say

• No Long, Short, Watchlist or Avoid.
• No E[R] versus a 4.7% cash hurdle.
• No 5-year DCF-derived price target, and no forward DCF at all — the original memo's refusal to build one was correct and is upheld.
• No claim that the reverse DCF was unsolvable. It solved. The solution is simply outside the achievable range, and that is stated as the result.
• No insolvency bear case.

7. Provenance