Phase Space AI

Cloudflare Trade Construction 2026-07-26

Cloudflare [NET]

Cloudflare, Inc. [NET] — Trade Construction & Risk Management

Investing Hub Research | July 26, 2026 (revised; refreshed same day under skill v1.2.0) | Task 5 of 9

Updated 2026-07-29 — Criteria + two-horizon valuation. The numbered Gates are retired and replaced by named Criteria with types (BINDING / MEASURED) returning PASS / FAIL / INDETERMINATE. This memo no longer issues a position verdict — no Long, Short, Watchlist or Avoid. The expected-return-versus-cash-hurdle test is replaced by a reverse-DCF implied-path test plus a 12-month target, with sensitivity run over the exit multiple rather than over scenario probabilities. Momentum is entry timing only and vetoes nothing. Sections below that predate this update are retained as the historical record and are annotated where superseded.

Reading the retired "Gate N" numbering. Any Gate N below is historical. The mapping is: Gate 1 → Quality Criteria (BINDING) · Gate 1b → Short Mechanism Criteria (MEASURED) · Gate 2 / 2A / 2B / Path B → dissolved, absorbed by the Valuation Criteria · Gate 2C → Peer Spread Criteria (MEASURED) · Gate 3 → Catalyst Criteria (MEASURED) · Gate 4 / 4a → Valuation Criteria (BINDING) · Gate 5 → Liquidity Criteria (BINDING) · Gate 6 → Momentum Criteria (MEASURED, entry timing only) · (new) Downside Criteria (MEASURED). A Criteria returns PASS / FAIL / INDETERMINATE; a missing input is INDETERMINATE, never FAIL. Historical calibration items keep the old numbering on purpose — the record is the record.

v1.2.0 refresh note: the skill's methodology upgrade added a Factor & Anomaly Scorecard (Task 3, Section 2.F), a sixth hard gate (momentum/tape), a quantitative accruals/F-score corroboration standard for Gate 1, a PEAD check, and a defined-risk-spread default for options vehicles. This document was refreshed against all of them. (That paragraph originally concluded the Watchlist decision is unchanged and strengthened. The verdict is withdrawn as of 2026-07-29; the anomaly evidence itself is retained and is scored under the Criteria in §2.)

Prerequisite: Builds directly on Task 3's valuation (weighted target $105.27, current price $262.22, -59.9% implied downside) and the Street/positioning context added to Task 3, Sections 2.D-2.E.

v1.4.0 pricing note (2026-07-27): Sections written on 2026-07-25/26 use the 2026-07-24 close of $262.22 and a 352.6M diluted share count. The v1.4.0 sections appended below use the 2026-07-27 close of $265.61 and the company's guided ~375M FY2026 weighted-average share count. Both bases are stated wherever a figure depends on them; they are not blended.

Revision note: an earlier version of this section recommended an outright Short. On review, that recommendation was withdrawn — it was built on a valuation gap alone, without a fundamental thesis clearing the bar a real short requires. This version applies the gated framework below and reaches a different conclusion. The original recommendation and this revision are both preserved in the git history and the recommendation ledger (trade_recommendations.jsonl) rather than silently replaced — see the entry marked invalidated.

---

0. The default is no position

A stock can be expensive without being a good short. The correct causal chain for a fundamental short is: business deterioration → earnings below consensus → estimate revisions → multiple compression → negative expected return. Jumping from "the multiple is above historical/peer norms" straight to Short skips every economically important step — and Section 2.E already establishes that NET's own numbers are in line with or better than what the Street expects, which is the opposite of what a short thesis requires.

1. Four Conclusions, Kept Separate

1. Fundamental conclusion (Task 2/3): Revenue growth is accelerating (34% YoY in Q1 2026, guidance raised), and the house Base case shows EBITDA margin turning positive in 2027E. Nothing in the fundamental picture is deteriorating.
2. Expectations conclusion (Task 3, Section 2.D): House NTM revenue is within ~1.3-1.9% of Street consensus (essentially aligned); house non-GAAP EPS is actually 31-37% above Street consensus. There is no negative variant here — if anything, the house view is more bullish than the Street's on margins.
3. Valuation conclusion (Task 3, Sections 1-4, 2.E): Every method (DCF exit-multiple, trading comps) implies a fair value well below the current price, and the gap decomposes almost entirely into a multiple disagreement (Street's targets hold today's ~41x multiple flat or expand it; the house target implies compression to ~16x, within peer historical range) — not a fundamentals disagreement.
4. Portfolio conclusion (this section): given #1-#3, is a directional position actually warranted? See the gate analysis below.

2. THE CRITERIA (2026-07-29 — the numbered "Gates" are retired)

Per ~/.claude/skills/investment-memo/references/criteria.md, every test is a named Criteria carrying a type, returning PASS / FAIL / INDETERMINATE. BINDING = a long-only absolute-return book treats failure as disqualifying. MEASURED = always scored and stored, informs timing or sizing, never blocks.

A missing input is INDETERMINATE, never FAIL (calibration item D1). The memo scores every Criteria and blocks on none of them.

Data note. The pre-computed scan record for NET is status: INDETERMINATE, note no share count, and therefore carries no EV, no growth figure, no exit multiple and no valuation result. The share count is recoverable and has been recovered — WeightedAverageNumberOfDilutedSharesOutstanding = 352,625,000 (period 2026-03-31, filed 2026-05-08). A second correction applies: the scan's "TTM revenue" omits Q4 on every name; NET's corrected TTM is $2,329m against the scan's $2,193m (+6.2%). Both in Valuation §0.1.

2.1 What changed versus the previous version of this section

3. ANALYSIS SUMMARY — no position verdict

This memo does not output "Long", "Short", "Watchlist" or "Avoid". Both the original §3 and the v1.4.0 addendum's §E issued INVESTMENT DECISION: WATCHLIST. Those verdicts have been removed, along with the conversion triggers framed as decisions. The evidence is retained above and the pre-registered Q2-2026 tests in §D survive as falsifiable predictions, which is what they always were.

NET — Cloudflare, Inc.                                framework: Criteria + two-horizon valuation (2026-07-29)
Spot $264.20 (2026-07-28)                             Archetype: INFLECTION

DATA         scan record is INDETERMINATE ("no share count") and carries NO EV, growth, exit multiple or
             valuation. Share count recovered: 352,625,000 (diluted, 2026-03-31, filed 2026-05-08).
             scan TTM revenue $2,193m -> $2,329m (+6.2%): the scan's "TTM" omits Q4 on every name.

BINDING      Quality ............ FAIL on 1 of 3 INFLECTION legs
                                  gross margin LEVEL 74.5% ......... PASS  (filed GrossProfit tag)
                                  revenue growth 30.5% / TTM +31.6% . PASS  (accelerated 27->28->31->34->34%)
                                  op margin CHANGE -0.3pp .......... FAIL  (needs ~+5pp; FY2024 did +5.0pp,
                                                                    FY2025 gave 0.3pp back)
                                  F-score 4/9 and gross profitability DEMOTED TO CONTEXT on INFLECTION;
                                  the F-score failures are 2025-convertible artifacts. Accruals -15.1%, clean.
             Valuation .......... FAIL by 16.3pp (sales basis) | INDETERMINATE (EBIT basis)
                                  required 46.9% vs demonstrated 30.5%  ->  margin -16.3pp
                                  exit 9.34x EV/SALES, GROWTH_MATCHED (n=26, peer growth 15.4-42.2%)
                                  implied compression: -30.3x / -76.4% from today's 39.6x - the largest of
                                  the five names updated today
                                  sign flips at ~17x EV/Sales. At today's 39.6x held flat: required 10.0%
                                  (+20.5pp). THE ENTIRE DISPUTE IS WHETHER ~40x SALES SURVIVES FIVE YEARS.
                                  EBIT basis INDETERMINATE: op margin -9.6% makes the terminal margin
                                  UNDERIVABLE; it is not asserted. (The EBIT anchor itself resolves
                                  GROWTH_MATCHED at 49.4x, n=12 - it is the margin that is missing.)
             Liquidity .......... PASS with caveat  Sep-2026 chain: 12 strikes with OI, median 617,
                                  total 12,231. Median quoted spread 10.9% of mid - 3-4x the mega-caps.
                                  Defined-risk spread fillable; naked premium is not.

MEASURED     Momentum ........... 3rd quintile (12-1 +22.3%, 49.6th pctile, RSI-14 42.0, 93.9% of 52w high,
                                  ABOVE the 200-day) - TIMING ONLY
             Catalyst ........... Q2-2026 earnings ~6 Aug 2026 - the first observation of whether the
                                  wrangler inflection (5.8M -> 78.8M downloads, 10.3x YoY) converts to
                                  consumption revenue. Pre-registered tests: growth >34% AND cRPO >34%.
             Downside ........... multiple compression, NOT business failure. Core business $105.27 plus
                                  AI optionality +$25 to +$95 (midpoint ~$40) = ~$145 (-45%), TAIL-DOMINATED
                                  and labelled an option, not a base case. Named business risks
                                  (restructuring execution, hyperscaler pricing) are NOT evidenced as
                                  occurring. No going-concern case; +$932m net cash, CFO +$603m.
                                  Vol 60.4% - highest of the five; a SIZING input, not the risk measured.
             Consensus .......... Street $251.87 (34 analysts) - 4.7% BELOW spot, the only one of the five.
                                  House revenue -1.3%/-1.9% vs Street; house non-GAAP EPS 31-37% ABOVE.
             Peer Spread ........ INDETERMINATE - a peer comp set exists but no peer's own multiple history
                                  and percentile was assembled
             Short Mechanism .... FAIL on BOTH legs (growth accelerating; margin runway barely begun)

12-MONTH TARGET   $369  (+39.8%)   own four-year median 42.21x EV/Sales on NTM revenue $3,063m
                  no-re-rating $347 (+31.2%) at today's 39.61x; downside $312 (+18.2%) at own p25
                  NET sits at the 43rd PERCENTILE of its own four-year range - it is NOT expensive against
                  its own recent history. This is a growth target, not a re-rating target.
                  Street $251.87 (-4.7%) implies 28.69x - the Street prices MORE compression than this does.
                  NTM revenue cross-check: consensus 2026E $2,812.9m / 2027E $3,590.6m brackets $3,063m.

THE TWO HORIZONS DISAGREE, AND THAT IS THE OUTPUT
                  +39.8% over 12 months against a five-year implied path failing by 16.3pp. Not a
                  contradiction: ~40x sales is unremarkable for NET WITHIN ITS OWN RECENT HISTORY and very
                  hard to defend AS A TERMINAL ASSUMPTION five years out. This name is the clearest
                  illustration in the batch of why reporting only one horizon was a defect.

DISCLOSED GAPS    NTM revenue is corrected-TTM YoY held flat, NOT consensus (Alpha Vantage quota exhausted),
                  though it is bracketed by the memo's own sourced consensus. Peer Spread INDETERMINATE.
                  EBIT-basis implied path INDETERMINATE by construction on a loss-making name.

---

4. What the Original Short Got Wrong, Concretely

The original recommendation computed "risking ~30% (to $340 invalidation) to make ~60% (to $105.27 target)" as a point-estimate risk/reward. Attempting the actually-required scenario-weighted E[R] exposes the problem:

Approximate E[R] for the short ≈ (0.20 × 60%) + (0.60 × 0%) + (0.20 × -30%) ≈ +6%, before borrow cost, slippage, and (for an options structure) premium paid and IV crush. Net of those costs, this is close to zero or negative — a materially different picture than the "risking 30% to make 60%" framing suggested, and consistent with Gate 4 not being properly cleared. The point-estimate framing made the trade look far more attractive than the honestly scenario-weighted version does — exactly the failure mode the gated framework exists to catch.

5. Position Sizing / Vehicle Selection

Not applicable — no position is being sized, and no classification is issued. This memo does not carry an entry, target, invalidation level, sizing, or vehicle recommendation, because there is no position to construct. Constructing one anyway (as the original version did) would repeat the exact error being corrected here: using valuation math to manufacture the appearance of an actionable trade where the fundamental/expectations/catalyst gates don't support one.

Standing vehicle note, retained for any strategy that acts on this analysis (v1.2.0): if it ever becomes a directional position, the options default is a defined-risk vertical spread, not naked long premium — option buyers systematically pay the variance risk premium (Coval & Shumway 2001; Bakshi & Kapadia 2003), and it is largest in exactly the short-dated OTM catalyst structures the original (withdrawn) recommendation used. A naked put/call purchase would require showing the house expected move materially exceeds the option-implied move (from the ATM straddle) for the catalyst window, with both numbers stated. This corrects the original ledger entry's structure (naked OTM puts into earnings) as well as its thesis.

6. Monitoring Plan (in place of a trade)

Track the Task 6 Catalyst Calendar as the mechanism for reassessment: - Q2 2026 earnings (8/6/26): watch specifically for evidence on Gate 1 (is the restructuring actually damaging anything, or delivering the promised leverage) and Gate 2 (does either house or Street estimate revision suggest a genuine variant is forming). - Pay Per Use effective date (9/15/26): watch for early adoption evidence that would bear on whether the "regime change" bull case (AI-monetization optionality) is real. - Continue monitoring short interest and Street target revisions (Task 3, Section 2.E) for signs the consensus view is shifting in either direction.

---

HISTORICAL RECORD — superseded 2026-07-29. Everything below was written under the retired numbered-Gate framework and its E[R]-versus-cash-hurdle test, and it states a position verdict. That verdict is withdrawn and the memo issues none. The evidence is retained because it is good evidence; the labels attached to it are not current. The live scorecard is §2 and the live valuation is §0 of the Valuation Analysis.

Output Format (feeds the front-of-report trade ticket, Task 7)

INVESTMENT DECISION: [REMOVED 2026-07-29 - the memo issues no position verdict]
GATES: 1(Mechanism): Fail (narrative AND quantitative — accruals/F-score show no deterioration signature)
       2(Variant vs. Consensus): Fail  3(Catalyst): Fail (wrong catalyst for this thesis)
       4(Expected Return): Weak (~+6% before costs, scenario-weighted)  5(Feasibility): Pass
       6(Momentum/Tape): Fail for a Short — against strong momentum on every measured signal
FACTOR SCORECARD (Task 3, 2.F): momentum/PEAD/revisions all against a short; earnings quality high (accruals
-15.1%); rising-but-low short interest the only bearish factor reading. ["Corroborates Watchlist" removed 2026-07-29.]
RATIONALE: Valuation is stretched (Task 3) but no evidenced fundamental deterioration, no negative variant vs.
Street, no catalyst that tests the actual disagreement (which is about multiple durability, not execution), and
a short would fight strong, accelerating momentum with an upward PEAD drift. A valuation gap alone does not
clear the bar for a Short position.
CONVERTS TO SHORT IF: evidenced growth/margin deterioration appears, house forecast falls genuinely below
Street, and a catalyst emerges that tests that variant directly.
CONVERTS TO LONG IF: restructuring efficiency gains and AI-monetization revenue convert to disclosed, measurable
results ahead of the house Base case.

---

HISTORICAL RECORD — superseded 2026-07-29. Everything below was written under the retired numbered-Gate framework and its E[R]-versus-cash-hurdle test, and it states a position verdict. That verdict is withdrawn and the memo issues none. The evidence is retained because it is good evidence; the labels attached to it are not current. The live scorecard is §2 and the live valuation is §0 of the Valuation Analysis.

ADDENDUM — v1.4.0 RE-RUN (2026-07-27): GATE 2 RE-TESTED UNDER THE AMENDED RULES

Decision: [REMOVED 2026-07-29 — the memo issues no position verdict. The addendum originally recorded "WATCHLIST — unchanged".] The reasoning underneath it changed materially, and in the direction of the bulls, which is why the evidence below is retained in full. Under v1.3.0 this memo recorded "Gate 2: FAIL" flatly, on the ground that house and Street revenue were within 2% and the house was above Street on EPS. That test only ever contemplated a near-term estimate variant. v1.4.0 adds Gate 2B — a duration/optionality variant against a four-part evidence bar — and NET is the first name in this coverage where 2B is genuinely arguable. It is tested honestly below and, unusually, it passes. The position still fails, at a later gate and for a better-articulated reason.

A. Gate 2, both branches

2A — Estimate variant: FAIL (unchanged). House 2026E revenue $2,775.0M vs. Street $2,812.9M (−1.3%); 2027E $3,523.3M vs. $3,590.6M (−1.9%). House non-GAAP EPS sits 31–37% above Street. There is no near-term numbers disagreement in either direction. This remains the correct verdict and the new evidence does not disturb it.

2B — Duration / optionality variant (for a LONG): PASS on all four legs.

So Gate 2 now PASSES for a long, where it previously failed outright. That is a real change and it is stated as one rather than buried.

B. What the price requires — the binding constraint (heading updated 2026-07-29; the original read "Why the position is still Watchlist")

Gate 4 (scenario-weighted expected return) fails, and the implied-penetration test is why.

Leg 4 asks whether consensus estimates embed the opportunity. They do not. But estimates are not the price. At $265.61 and the company's guided ~375M FY2026 shares, the $99.6B market capitalisation requires roughly $24.3B of revenue by 2033 to deliver a 10% return at an 8x exit — a 36% CAGR sustained for seven years from a $2.8B base. Adding the entire Base-path inference business to the consensus trajectory lands near $20B.

The AI-inference opportunity is missing from consensus estimates and already paid for in the price. Those are different things, and conflating them is exactly how a well-evidenced duration variant becomes a bad trade.

Revised target ~$145 (up 38% from $105.27, entirely on the new evidence) against $265.61 leaves a −45% gap. The expected return for a long is materially negative on any honest weighting; the expected return for a short is positive on paper and inadmissible for the reasons below.

C. The other gates, re-checked

Net: a long now clears Gates 1, 2, 3, 5 and 6 and fails only on Gate 4 — price. That is a much more precise statement than "expensive," and it converts the name from an indefinite watchlist into one with an arithmetic entry condition.

D. Revised conversion triggers

→ Conditions that would materially improve this analysis (retained as falsifiable, dated predictions — not conversion triggers, because no verdict exists to convert): - P1 (price). The stock trades below ~$150 — the level at which the revised ~$145 target plus the tail-weighted optionality range makes the scenario-weighted expected return positive. This is derived from the §7 paths, not from a percentage discount to the current price. - A1 (adoption converts to revenue). At the Q2-2026 print (~6 August 2026), revenue growth re-accelerates above 34% and Current RPO growth breaks above the 33/30/34/34% band it has held since Q2-2025. That combination would establish that the wrangler inflection is pre-monetisation demand rather than free-tier and CI traffic, and would justify paying up. Both conditions, not either. - A2 (corroboration). wrangler monthly downloads hold above ~60M through Q3-2026 (i.e. May's 78.8M peak was not a spike), and the changelog's pricing/billing share keeps rising — metering follows monetisation intent.

→ Conditions that would establish a deterioration mechanism (none of which is currently evidenced): two consecutive quarters of decelerating revenue growth with Current RPO growth below 30%, and the tape breaks (weekly close below the 200-day). Nothing in the current evidence points that way, and the new corpora point firmly against it.

→ How to score this analysis after the fact: if Q2-2026 delivers both A1 conditions and the stock does not re-rate, the analysis was right and the entry was available; if Q2-2026 delivers both and the stock gaps up through $300, the opportunity was correctly identified and incorrectly sized to zero — and that outcome should be scored honestly in the ledger, because "I was right to wait" and "I missed it" are the same data point until the price is checked.

E. Revised output block

INVESTMENT DECISION: [REMOVED 2026-07-29 - the memo issues no position verdict]
   (this line read "WATCHLIST - long-biased, awaiting price or proof" under v1.4.0)
GATES (for a LONG): 1(Mechanism): PASS  2(Variant): PASS via 2B, all four legs  3(Catalyst): PASS —
       Q2-2026 print ~6 Aug 2026  4(Expected Return): FAIL — the price already embeds more than the base
       TAM path delivers  5(Feasibility): PASS  6(Momentum): PASS — a long is with the tape
GATES (for a SHORT): 1: FAIL (harder than before)  2A: FAIL  3: FAIL  6: FAIL
CURRENT: $265.61  |  REVISED TARGET: ~$145 (-45%)  [was $105.27]  |  CORE $105 + AI OPTIONALITY ~$40
IMPLIED PENETRATION: $99.6B market cap on $2.81B of guided FY2026 revenue = 35.4x forward revenue. To return
       10%/yr and exit at the peer-median 8x EV/revenue, Cloudflare must reach ~$24.3B of revenue by 2033 —
       8.6x the 2026 guide, a 36% CAGR for seven straight years. It is guiding the NEXT quarter to
       decelerate from 34% to ~30%.
SIZING: no position. On conversion at <$150: Medium conviction x High vol = 1% of book.
VEHICLE (on conversion): defined-risk call spread, not naked calls — see the variance-risk-premium rule.
THESIS IN ONE SENTENCE: Cloudflare's Workers toolchain is being invoked ten times more often than a year ago
       and its engineering output has nearly tripled with agent-related releases sextupling their share — an
       inflection that is real, independently measurable, and absent from consensus estimates — but it is
       already more than paid for in a price that requires seven consecutive years of 36% growth, and the
       company's own job board says it is hiring account executives, not inference engineers.
CONVERTS TO LONG IF: price <$150 (P1), OR Q2-2026 shows revenue growth >34% AND Current RPO growth >34% (A1).