Phase Space AI

Valuation

Snowflake [SNOW]

Snowflake [SNOW] — Valuation

As of 2026-07-29. Framework v1.5.1. Tier-2 memo, cluster member (data & dev infrastructure). Cluster analysis: reports/clusters/Data_Dev_Infrastructure_Cluster_2026-07-29.md

This memo issues no position verdict. The book decides.

Spot $283.04. Two outputs are produced, as required: a 12-month target and the implied-path test. Neither replaces the other.

---

1. Corrected inputs used

Screen defect carried into this valuation: Net cash overstated by $905m (56%): omitted $1,432.5m of non-current AFS securities AND $2,281.9m of convertible notes.

---

2. Implied-path test — the Valuation Criteria

This is the primary long-horizon output. Terminal value is effectively 100% of EV for this name (GAAP operating margin is -26.11%, so there is no interim EBIT to discount), which is far above the 60% threshold at which the reverse DCF becomes mandatory as the primary instrument.

Run on assets/reverse_dcf.py, EV/Sales basis. Parameters held fixed and named: 5-year horizon; WACC 10.0%; net cash $2,105.6m; shares 346.601m; TTM revenue $5,032.8m; exit multiple as shown in each column. Solved for: the required revenue CAGR.

Demonstrated: 33.5% (latest-quarter YoY). Per the brief, the run-rate is used, not a trailing CAGR — trailing CAGRs are stale for this entire cohort.

The required parameter, and the margin

Base exit multiple: 8.0x EV/Sales. Required revenue CAGR: 30.9%. Demonstrated: 33.5%. Margin = +2.6pp.

Implied multiple compression: 19.07x → 8.0x = +11.07x, a 58% compression.

Result: PASS WITH ARGUMENT.

FAILS at a 6.0x exit (requires 38.6% vs 33.5% demonstrated). The verdict is exit-multiple-dependent and that is declared.

Why the exit multiple is anchored where it is

The five cluster names bracket each other on current growth (23.1%–33.5%), so they are each other's growth-matched comparators today. They are not growth-matched comparators for their own year-5 selves — a multiple reflecting 23–33% growth cannot be applied to a business growing 12–15% five years out. Extrapolating today's cluster median forward would reproduce exactly the defect the framework closed.

The cluster base of 8.0x EV/Sales is argued in the cluster document: it sits above GTLB's (4.37x) and TEAM's (4.29x) current multiples, so it is not a haircut stacked on names that have already fully de-rated — the NTRA double-counting failure — and below SNOW's, DDOG's and MDB's, which is required because those trade on 30%-growth expectations that will not persist to year 5. 8x sales on 80%+ gross margins and 20–25% FCF margins implies roughly 32–40x FCF, where mature mid-teens-growth infrastructure software has traded post-2023.

Sensitivity is run over the exit multiple, never over scenario probabilities — the NTRA failure was running the range on the parameter that could not change the answer. Here the exit multiple can and does change it.

Required terminal EBIT margin — the second solve

Solving the same instrument for the required terminal EBIT margin instead, with growth held at 33.5% and an EBIT-basis exit multiple:

48.3% at 15x / 36.2% at 20x / 29.0% at 25x / 24.1% at 30x

Current non-GAAP operating margin ≈ +6.1% (GAAP -26.11% plus SBC of 32.2% of revenue).

The price requires roughly 36% at a 20x exit, versus 6.1% earned today before SBC — a gap of 30.1pp.

Screen defect on this input: the Tier-1 record assigned a terminal margin of exactly 0.107 to four of the five cluster names, described as an "industry median of mature profitable peers." The same 10.7% was applied to a 66%-gross-margin consumption business and an 86%-gross-margin seat business. It is not a name-specific estimate, it produced "reverse DCF unsolvable in range" and valuation: INDETERMINATE, and the correct output was the table above.

---

3. 12-month target

Not a DCF. Not a peer median projected forward. Built per references/valuation.md: near-term revenue, named events inside 12 months, and the name's own multiple history with the percentile stated.

Consensus: INDETERMINATE. Alpha Vantage's 25/day quota is shared and exhausted. Per the brief this blocks nothing; the NTM revenue figure above is an explicit house estimate off the quarterly YoY trend, labelled as such, not a consensus number presented as one.

Regime-change treatment. The 2021 SaaS bubble and the 2022 de-rating are a regime change, so the full-history percentile is reported for completeness but not used as an anchor. The post-2024 series for this name is genuinely two-sided, with round trips in both directions, so median reversion is a legitimate 12-month anchor and the anchor is IDENTIFIED.

Sanity band: no external professional target was available for this name, so the gap cannot be reported. That is a gap in the check, not a defect in the output.

---

4. Reconciliation of the two outputs

These are two different instruments measuring two different things over two different horizons, and they are not calibrated to each other by design. The framework tested indexing a 5-year reverse DCF to 12-month targets directly and achieved a best correlation of +0.08, requiring absurd multiples. Reporting only one of the two is the defect valuation.md exists to prevent.

The two do, however, agree on the ordering across the cluster — implied-path margin and own-history percentile rank the five names identically (TEAM ≈ GTLB > MDB > SNOW > DDOG) despite sharing no inputs. That agreement is the strongest evidence produced in this cluster.