Snowflake (SNOW.US) Zen Horizon Report

Note: this report's rating is "Watch," which under site rules does not automatically trigger the review/supplement re-check process.

II. Company Profile

Business model. Snowflake is a cloud-native data platform whose core is the complete decoupling of data "storage" from "compute": customers load their data into its managed cloud warehouse and pay on a consumption basis (usage-based) for the compute (credits) and storage they actually consume, rather than the fixed per-seat subscription of traditional software. The upside of this model is that the more customers use, the more they pay, and revenue grows naturally with the volume of a customer's business data; the cost is that revenue is tied directly to swings in customer usage and lacks the smoothness of subscriptions (a point fully exposed when growth halved in 2022-23, see Section X on risks).

Product portfolio. The foundation is its cloud data warehouse and lakehouse capabilities; on top of that it has built data sharing and the Marketplace (a data exchange with 3,000+ data/application listings and 700+ providers), Clean Rooms (privacy-preserving cross-party data collaboration), and Snowpark (letting developers do data engineering and machine learning directly within the platform using Python/Java); the AI layer is Cortex (managed LLMs and AI functions) plus newly launched 2026 agentic capabilities such as Cortex Code. The company's strategic positioning has moved up from "cloud data warehouse" to "AI Data Cloud," with the pitch of "bringing AI to where the data is" rather than moving the data to the AI.

Cross-cloud neutrality. Snowflake is deployed across all three major public clouds (AWS, Azure, Google Cloud), and customers can share and migrate data across clouds and regions. This is its core differentiation against the hyperscalers' self-operated data warehouses (AWS Redshift, Google BigQuery, Azure Synapse/Fabric, all tied to a single cloud); but it also means it runs on its competitors' infrastructure, with gross margin and pricing leverage constrained by the upstream cloud providers (a double-edged sword, see Section V on the moat and Section X on risks).

Customer structure. At the end of FY2027 Q1 it had roughly 13,900 customers, of which 779 were large accounts paying >$1 million annually (+29% YoY, +46 net adds in the quarter vs. 26 in the prior-year quarter), and 813 of the Forbes Global 2000 are customers; no single customer accounts for ≥10% of revenue, so customer-concentration risk is low. The number of large accounts and NRR (below) are the core metrics of the health of its "land and expand" motion.

Headquarters and management. Headquarters are in Menlo Park, California (135 Constitution Drive, listed as the legal principal office on the cover of the FY2026 10-K). Note that Bozeman, Montana, was the headquarters in the Slootman era and is now just one of many offices (the historical reason the 2024 data-breach MDL litigation sits in Montana, see Section X). The current CEO is Sridhar Ramaswamy (took over in 2024-02, formerly head of Google's advertising business, who joined when Snowflake acquired Neeva, the AI search company he founded, in 2023); his predecessor was Frank Slootman (2019-2024, who led the company through its IPO and to its valuation peak). Board insiders (14 directors and officers) hold roughly 4.8% in aggregate, so management's personal skin in the game is on the weak side (see Section X on governance); the company has no dual-class share structure (Class A has been consolidated into a single class of common stock), and the governance structure is clean.

III. Vertical Analysis: From Hypergrowth to Stalled Growth, and Then a First Reacceleration

History. Snowflake was founded in 2012 by three engineers from Oracle's database background, became commercially available (GA) in 2014, and Frank Slootman took over as CEO in 2019 and drove its IPO. The 2020-09-16 IPO was the largest software-company IPO in history: priced at $120, it opened at roughly $245 on the first day and closed at $253.93 (doubling on day one), with Berkshire Hathaway (Warren Buffett's company) and Salesforce each subscribing for roughly $250 million through a directed placement, making it a star stock that "even the Oracle of Omaha bought." It then made acquisitions to strengthen its AI/data-engineering capabilities: Streamlit (a data-application framework) in 2022, Neeva (AI search, which brought in current CEO Ramaswamy) in 2023, and Natoma (an enterprise-grade MCP platform) in 2026.

⚠️ Correction note: the market once circulated that Snowflake "acquired Reka AI" in 2024; in fact the 2024-05 talks broke down and no deal closed, and it should not be counted in the acquisition list.

Growth curve (product revenue YoY), the main thread for understanding the rise and fall of Snowflake's valuation:

Five straight years of slowing from +120% all the way to +29% is the textbook trajectory of a hypergrowth stock stalling out; the FY2027 guidance of +31% is the first increase in growth since the IPO. The quarterly view is clearer: product revenue growth went from +26% in FY2026 Q1 to +30% in Q4 to +34% in FY2027 Q1, three straight quarters of acceleration.

NRR (net revenue retention) curve, a measure of expansion among existing customers: a FY2022 peak of 178%, then 131% in FY2024, 126% in FY2025, 125% in FY2026, and back up to 126% in FY2027 Q1. It has held around 125% for the last five quarters and ticked up in Q1, stabilizing for the first time after a three-year decline. The reacceleration in growth plus the stabilization in NRR form the two data pillars of the "reacceleration" story.

Stock price and valuation history. After the IPO the stock touched an intraday all-time high of $429 in 2020-12 and a closing high of $401.89 in 2021-11 (when EV/Rev briefly spiked to a bubble peak of about 77x); the 2022 double hit of rate hikes plus slowing growth dragged it down to the $110 area; it hit a historical low of $107.13 on 2024-09-06 (EV/Rev compressed to about 12x). It then recovered on the AI narrative and a fundamentals repair, to roughly $280 in 2025-10, a retest of the 52-week low of $118.30 on 2026-04-10, then a single-day +36% post-earnings jump on 2026-05-28 that touched a 52-week high of $284.99 intraday, closing at $238.26 on 2026-06-05. The current price is about -41% from the closing ATH of $401.89 and about -44% from the intraday ATH of $429. The valuation multiple (EV/Rev) went from about 82x at the IPO and about 77x at the bubble peak, to about 12x at the end of 2024, and now to about 16x, in the lower one-third of the range since the IPO (about 5.5 years) (⚠️ because it only went public in 2020, there is no "10-year" historical window; the median of about 22x is on a since-IPO basis).

Two landmark shifts in ownership are worth noting: Berkshire (Buffett) fully exited in 2024 Q2 (about 6 million shares, roughly $800 million); Salesforce, which subscribed to the IPO that same year, has also exited. The departure of these marquee anchor shareholders overlaps with the period of stalled growth.

IV. Financial Review

FY2027 Q1 (ended 2026-04-30, reported 2026-05-27), the proof of reacceleration:

• Product revenue $1.3343 billion (+34% YoY), total revenue $1.391 billion (+33%), with product accounting for 96% (the rest being professional services, which run at a gross loss)

• Net revenue retention NRR 126% (stabilizing and recovering); remaining performance obligations RPO $9.21 billion (+38%), with contracted-but-unrecognized revenue growing well ahead of the current revenue growth rate, a strong leading indicator of future growth

• Non-GAAP operating margin 11.9%, non-GAAP EPS $0.39 (above the consensus of $0.32); GAAP net loss of -$295.6 million

• Adjusted free cash flow $265.5 million (margin 19.1%, with the quarterly figure below the full year because of cash-flow seasonality)

• Stock-based compensation (SBC) $433.7 million (about 31% of revenue); weighted average share count 345.4 million shares (+3.8% YoY)

Full-year FY2026 (ended 2026-01-31), the full quality picture:

• Total revenue $4.6839 billion (+29.2%), product revenue $4.4723 billion (+29%)

• GAAP operating loss of -$1.4352 billion, GAAP net loss of -$1.3316 billion (full-year loss per share of -$3.95; ⚠️ note the TTM trailing loss per share is about -$3.51; the two cover different windows and coexist without contradiction)

• Stock-based compensation SBC $1.5995 billion (34.1% of revenue), which is the key to understanding "huge GAAP loss vs. non-GAAP profit": the scale of SBC even exceeds the GAAP operating loss, so excluding SBC the business is already profitable at the operating level. ⚠️ Correction: some secondary sources (macrotrends) restate "FY2026 SBC of $3.96 billion / 88.7% of revenue," which is badly wrong; the first-party SEC figure is $1.5995 billion / 34.1%.

• Non-GAAP operating profit $489.7 million (margin 10.5%)

• Adjusted free cash flow $1.1927 billion (margin 25.5%), strong and sustainable cash generation

Balance sheet (solid): cash and investments of roughly $4.39 billion in total; the only interest-bearing debt is two tranches of 0%-coupon convertible notes totaling $2.28 billion ($1.15 billion due 2027-10, $1.15 billion due 2029-10, with almost no cash interest cost); net cash of +$2.11 billion. On buybacks, cumulative authorization is $4.5 billion, executed down to roughly $800 million remaining; FY2027 Q1 repurchased $300 million (about 1.7 million shares @ $178.95).

FY2027 guidance (already raised): product revenue of $5.84 billion (+31%, raised from the original $5.66 billion / +27%); non-GAAP operating margin of 13.5% (raised from 12.5%); non-GAAP product gross margin of 75%; adjusted FCF margin of 23%. The margin guidance was raised in tandem, indicating the reacceleration is not being bought with cash burn in exchange for growth.

Financial verdict: software-grade gross margin (product 75%), an FCF margin in the 25% range, net cash plus zero-coupon debt; quality is top-tier among unprofitable growth software. The structural flaws are in GAAP and the share count: persistent large GAAP losses, SBC at 34% of revenue (declining year over year but still high), with non-GAAP profitability achieved in large part by "saving" stock-based compensation, the weighted share count still up +3.8% YoY, and the roughly $300 million of buybacks each quarter used mainly to offset dilution rather than to genuinely return capital to shareholders.

V. Moat: Composite 3/5 — The Core Stronghold Is Real, the Fastest-Growing Frontier Is Lost

Snowflake's moat in its core data-warehouse stronghold is real and already monetized, but the composite score is only 3/5, because the moat has failed to extend into the fastest-growing adjacent AI/ML market and is being eroded by multiple forces, with the trend narrowing.

The three pillars supporting the moat:

• Switching costs (strongest, about 4/5). Once an enterprise core data warehouse is in place, the cost of migration is extremely high: it requires redoing data governance, permission systems, ETL/data pipelines, and BI report integration, and carries business-continuity risk. This switching cost is monetized in the financials: RPO of $9.21 billion, NRR of 126%, 779 large accounts >$1 million, and a 71% GAAP product gross margin (a reflection of pricing power). This is where Snowflake's moat is genuinely stronger than prior names in this observation tier (such as the contract-manufacturer-flavored Fabrinet or the cycle-eroded Harmonic Drive): its switching costs not only exist but are continuously converted into revenue and profit.

• Data network effects (nascent, about 3/5). The Marketplace (3,000+ listings, 700+ data providers), data sharing, and Clean Rooms create "data gravity": the more parties put data in, the more valuable the platform is to other parties. This is real but has not yet formed a dominant flywheel, and it is being leaked by open formats (see below).

• Cross-cloud neutrality (differentiation, about 3/5). Three-cloud deployment is a structural selling point versus single-cloud rivals, but it is also a weakness: running on rivals' infrastructure leaves pricing power and gross margin constrained by the upstream.

The triple erosion that pulls the composite down to 3:

• The fastest-growing AI/ML frontier is lost. This is the most important deduction. Snowflake's AI revenue run-rate is about $100 million, while core rival Databricks's AI-product run-rate is about $1.4 billion; ⚠️ the two are not fully comparable (Databricks's definition of "AI products" is broader, and Snowflake's $100 million refers only to Cortex/AI workloads), but the order-of-magnitude gap of about 14x and its direction are not in doubt. The moat covers the legacy core data-warehouse market that is being disrupted, yet it has failed to extend into the fastest-growing adjacent AI/ML market, so the franchise is narrower than it appears on the surface.

• Self-promotion of Apache Iceberg/Polaris to de-proprietize. To address customers' demand that "data not be locked in," Snowflake itself embraces and promotes the open table format Iceberg and the open-source catalog Polaris; at the industry level this systematically lowers data lock-in and switching costs, which amounts to deliberately "letting water out" of its own data gravity (a double-edged sword: it retains customers who want openness but weakens long-term lock-in).

• Microsoft Fabric squeezes from the low end via bundling + the hyperscalers surround it on three sides. Microsoft Fabric (ARR >$2 billion, +60%, reaching 70% of the Fortune 500, 31,000 customers) enters the mid-market with a low price bundled with Office/Azure; AWS Redshift and Google BigQuery lock in through their respective cloud ecosystems. Snowflake faces Databricks grabbing the AI high ground above and Fabric grabbing the BI mid-market below.

Moat conclusion: the switching costs of the core stronghold would warrant 3.5-4 on their own, but the loss of the AI frontier + Iceberg self-de-locking + the Fabric squeeze pull the composite back to 3/5: real, but narrowing. This is the moat-level basis for a rating of "Watch" rather than "Hold": its moat quality is better than prior 3/5 narrow-moat names (switching costs more fully monetized), yet weaker than wide-moat, oligopoly/contract-locked names (4/5).

VI. Industry Demand: The Long Runway Is Real, but TAM Definitions Diverge Enormously and AI Is a Double-Edged Sword

The track. Cloud data and AI are a high-certainty long runway: enterprise data volume keeps exploding, and moving analytics and AI workloads to the cloud is an irreversible trend. The problem is that TAM definitions diverge enormously:

• ⚠️ Company self-estimate (broad definition): TAM of roughly $170 billion (CY2024) growing to $355 billion (CY2029), a CAGR of about 15.8%; but this definition covers data engineering, analytics, AI, and application development across broad data-cloud spending, far exceeding the market Snowflake currently actually reaches.

• Third-party narrow definition (cloud data-warehouse software): only about $10-23 billion, an order of magnitude smaller than the company's number.

On the company's broad definition, the FY2027 guidance of $5.84 billion in product revenue corresponds to penetration of only about 3.4%, an ample long runway; but on the narrow definition, penetration is already not low and growth depends more on the TAM expanding on its own and on taking share. Third parties such as Morningstar consider the company's older $152 billion-to-$342 billion (CY2028) definition "broadly reasonable," but investors should be clear-eyed: if the valuation is priced on the broad definition's low-penetration narrative, it embeds a great deal of optimistic assumption about "swallowing the AI/application market too."

AI demand is a net tailwind, but with an efficiency self-cannibalization tail:

• Tailwind: 13,600+ customers already use AI features, and agentic products such as Cortex Code are scaling; AI raises the value of "running analytics/agents directly on data," which benefits consumption.

• Self-cannibalization: AI is also making queries/compute cheaper; the company's own optimizations such as AISQL can cut the cost of certain workloads to about 1/70 of the original, lowering consumption per unit of workload. Snowflake management has publicly acknowledged the need to use new workloads to "neutralize" the revenue decline from efficiency gains, a structural double-edged sword for the consumption-pricing model.

• Non-exclusive: the AI data layer is split among five strong players (AWS, Google, Microsoft, Snowflake, Databricks); Snowflake holds no exclusive position and lags Databricks on AI monetization progress (see Sections V and VII).

VII. Horizontal Analysis: Absolutely Expensive, but Not a Bubble in Relative Terms

Valuation multiples (2026-06-05):

Absolute dimension: expensive. A forward P/E of 112x and EV/FCF of 71x put it at roughly the 90th percentile in tech; after the 36% surge on 2026-05-28 it is only about -16% from its 52-week high, with pricing already quite full.

Relative dimension: not a bubble.

• Growth-adjusted EV/Rev (EV/Rev ÷ growth): Snowflake at about 0.51x is cheaper than Datadog (DDOG) at 0.87x, Palantir (PLTR) at 0.95x, and Cloudflare (NET) at 1.24x, and in line with MongoDB (MDB)/Confluent (CFLT); on a growth-adjusted basis it is not expensive.

• Versus its own history: EV/Rev of 15.9x is well below the since-IPO median of about 22x and the 2022 peak of 77.6x, in a zone of valuation compression.

• Versus sell-side consensus: 51 analysts at Strong Buy (35 strong buy / 9 buy / 6 hold / 0 sell / 1 strong sell), an average target price of $288.43, a median of $295, and a range of $110-500; the current price is at about a 21% discount to the average and about a 24% discount to the median, with 30+ firms raising targets after earnings.

⚠️ Key caution (red-team insight): the discount in "a 21% discount to consensus" is to a Strong Buy consensus that was just raised 30+ times and chased higher by momentum. This is momentum, not a cushion. Sell-side consensus is systematically optimistic and should not serve as independent evidence of "cheap."

On Databricks (private, the core rival): ARR of about $5.4 billion (+65%), an AI-product run-rate of about $1.4 billion, and a latest valuation of $134 billion (2025-12 Series L), with reports it will file for an IPO in H2-2026 (the Q3 timing is rumor-level). Its EV/ARR is about 24.8x, about 55% more expensive than Snowflake in absolute terms, but its growth is 2x Snowflake's; on a growth-adjusted basis Databricks is actually cheaper (about 0.38x vs. Snowflake's 0.51x), and on a forward basis they basically converge. A Databricks IPO will become the "overhang benchmark" for Snowflake: once a peer growing 2x as fast and an order of magnitude ahead in AI goes public, it will keep amplifying the market's discount on Snowflake's growth and AI lag.

On the hyperscalers: Microsoft (MSFT) at about 22.5x forward, Google (GOOGL) at about 28.7x, and Amazon (AMZN) at about 31.6x — all P/Es on GAAP profits; Snowflake is at 112x forward and GAAP-loss-making. On an earnings basis Snowflake is 3-5x more expensive, and it faces bundling pressure from all three.

Horizontal conclusion: Snowflake's valuation is "absolutely expensive but not a bubble in relative terms." It is not a Harmonic-Drive-style "extreme overshoot" (where the price exceeds the entire sell-side average), but rather a "full pricing of a high-quality growth stock"; yet full enough that there is no margin of safety, and the relative advantages (growth-adjusted cheapness, low historical percentile) rest on the premise of "the reacceleration is sustainable + share is held," which is precisely what Databricks/Fabric challenge.

VIII. Current Fundamentals

Price anchor (close on 2026-06-05): $238.26 (prior close $244.18, -2.42% on the day); market cap $82.58 billion; shares outstanding 346.6 million; 52-week range $118.30-284.99. Valuation snapshot: forward P/E 112.71x, P/S 16.4x (trailing), EV/Rev 15.9x (LTM).

Current status: the company just delivered one of its strongest sequential results since the IPO (product revenue +34%, the largest single-quarter dollar increment in its history, NRR recovering, RPO +38%) and raised its full-year revenue and margin guidance, triggering the violent single-day +36% re-rating on 2026-05-28 (see Section XI on catalysts). Fundamentals are at a positive inflection of "reacceleration + margin expansion," but the stock has already fully reflected this: the current price is close to double the 52-week low of $118 and only about -16% from the 52-week high. This is a moment where "fundamentals are good but the good news is largely priced in."

IX. Valuation

Three scenarios (ranges in the Section I table):

• Base (170-290): product revenue growth sustains around 30%, FCF margin delivers 23%, and the valuation holds a forward EV/Rev of 8.7-15x. This corresponds to the space between the upper bound of the sell-side average of $288 / median of $295 and the lower bound of DCF fair value. At $238 the current price sits in the lower middle of this range.

• Bear (110-170): Databricks keeps taking share on the AI side + a second halving of consumption-driven growth (a replay of 2022-23) + multiple compression to a forward EV/Rev of 6-8x, back to the 2024 historical low of $107 / 52-week low of $118 range.

• Bull (290-430): AI/Cortex monetization scales, the reacceleration continues to 35%+, Databricks competition does not materially take share, the valuation returns to a forward EV/Rev of 18-20x, and it challenges the closing ATH of $401.89 / intraday ATH of $429.

DCF cross-check: under neutral growth and discount assumptions, an independent DCF gives a fair value of about $168; the current price of $238 is about 42% above it. This aligns closely with "ideal buy ≤ $170" and converges with the ~$175 entry line the red team derives from a forward EV/Rev of 8-9x.

Ideal buy price ≤ $170 (forward EV/Rev of about 8-9x, about 29% below the current price). The logic for locating this pullback magnitude: it lies between a Harmonic-Drive-style "extreme overshoot" (requiring -50% or more) and a Cheniere/Siemens-style "reasonable/fully priced valuation" (about -16%), reflecting Snowflake's intermediate state of "expensive but not extreme, high quality but unsettled competition." A pullback to $170 would leave a cushion for the triple risks of competition (Databricks/Fabric), consumption-driven volatility, and multiple compression.

X. Risks

① Valuation risk [High]. Forward P/E 112x, P/S 16x, EV/FCF 71x, roughly the 90th percentile in tech; after the 36% surge on 2026-05-28 it is only about -16% from the 52-week high, with full pricing and minimal tolerance for error. Any growth miss or a pullback in macro risk appetite could trigger multiple compression (the DCF puts the current price about 42% overvalued).

② Competition risk [High]. Databricks is growing about 2x as fast, its ARR of $5.4 billion already exceeds Snowflake's full-year revenue, its AI run-rate leads by an order of magnitude, and its impending IPO becomes an overhang benchmark; Microsoft Fabric squeezes from the mid-market with bundled low pricing (60-80% cheaper in some scenarios) and has reached 70% of the Fortune 500; AWS/Google lock in through their respective cloud ecosystems. Snowflake is pressured at both ends at once.

③ Consumption-driven revenue volatility risk [Medium-High]. Consumption pricing lacks subscription smoothing, with a precedent already in 2022-23: NRR halved from 178% to 131% and growth was cut from triple digits in half. In a macro slowdown or when customers optimize usage, revenue has high downside elasticity.

④ Stock-based-compensation dilution risk [Medium]. SBC is 34% of revenue, the weighted share count is still up +3.8% YoY, and buybacks are used mainly to offset dilution; if the stock falls, the cost of buyback-based offsetting rises and dilution becomes more visible.

⑤ AI execution and self-cannibalization risk [Medium-High]. Cortex monetization is still early (run-rate of about $100 million vs. Databricks at about $1.4 billion), with heavy investment in building AI capabilities and uncertain returns; at the same time AI-driven efficiency lowers consumption per unit of workload (optimizations such as AISQL cut the cost of some workloads to about 1/70), and the company must keep using new workloads to "neutralize" the revenue erosion, a structural double-edged sword for the consumption model.

⑥ Security and reputational/legal risk [Medium]. A 2024 data-breach incident affected roughly 165 customers; ⚠️ the root cause was attackers using stolen customer login credentials (mostly accounts without multi-factor authentication enabled), not a breach of the Snowflake platform itself, and it should not be characterized as "Snowflake was hacked." The related class actions have been consolidated into MDL No. 3126 (federal court in Montana, consolidated complaint filed 2025-02) and remain ongoing; the AT&T-related case was settled in 2025-03, while Ticketmaster and others continue to litigate. The matter remains in the legal-proceedings/risk disclosures of the company's latest FY2026 10-K (verified against the most recent first-party SEC filing).

⑦ Governance and anchor-shareholder risk [Medium-Low]. Insiders hold only about 4.8%, so management's skin in the game is weak; both Berkshire (Buffett) and Salesforce, the two major IPO anchor shareholders, have fully exited; Vanguard at 5.1% is the only holder above 5%. The positive is no dual-class structure and no controlling shareholder, with transparent governance.

Pre-mortem (assume this investment loses money three years out; the most likely cause): the most likely is "paid too dear for the reacceleration" — buying at a 112x forward P/E after a 36% surge, after which Databricks goes public and keeps taking share with 2x growth + AI leadership, Microsoft Fabric eats away from the low end, Snowflake's growth slips again or AI monetization falls short, and the valuation compresses from about 16x EV/Rev back to 8-10x, so that even if revenue still grows the stock still halves. The secondary scenario is that consumption-driven revenue stalls a second time in a macro slowdown, replaying 2022-23.

XI. Catalyst Tracking

Realized catalyst: the single-day +36% surge on 2026-05-28 — four arrows at once:

• FY2027 Q1 earnings beat across the board (non-GAAP EPS $0.39 vs. expected $0.32, revenue of $1.39 billion vs. expected $1.32 billion, product revenue +34% the largest single-quarter dollar increment in its history);

• raised full-year guidance (product revenue $5.66 billion to $5.84 billion, operating margin 12.5% to 13.5%);

• reached a roughly $6 billion, five-year cloud commitment agreement with AWS — ⚠️ note this is a Snowflake purchase/cost commitment to AWS (i.e., it must spend at least this much on AWS over the next five years), not Snowflake revenue, and should not be misread as a revenue positive; its significance is in locking in upstream capacity and discounts to support long-term expansion;

• the acquisition of Natoma (an enterprise-grade MCP platform, strengthening agentic/tool-calling capabilities).

Future monitoring points (the validation items that determine whether the rating can be upgraded):

• whether product revenue growth can hold 30%+ (the sustainability of the reacceleration, rather than a single-quarter pulse);

• whether NRR can keep recovering to 127%+ (consumption-driven resilience);

• the pace at which AI/Cortex run-rate scales (the convergence of the order-of-magnitude gap with Databricks);

• Databricks IPO pricing and disclosures (the key external event for benchmark re-pricing);

• delivery of the non-GAAP operating margin toward 13.5%+, with the SBC ratio falling + net share increases narrowing;

• a valuation pullback to the ideal buy zone of ≤ $170.

XII. Where Horizontal and Vertical Meet

Vertically, Snowflake is at a real positive inflection: a first reacceleration after three years of stalled growth (product revenue +34%, NRR stabilizing at 126%, RPO +38%, guidance raised), with FCF and balance-sheet quality top-tier among unprofitable growth software (FCF margin 25.5%, net cash +$2.1 billion, zero-coupon debt, clean governance). This is where it is clearly better than prior "Watch" names in this tier: not a cyclically recovering contract manufacturer, but a high-quality platform with real monetized switching costs and a delivered reacceleration.

Horizontally, the problem is price and rivals: the absolute valuation is expensive (forward P/E 112x, 90th percentile in tech), and after the 36% surge on 2026-05-28 the reacceleration is priced in one move with zero margin of safety; the moat is 3/5 — the core stronghold is real, but on the fastest-growing AI/ML frontier it trails Databricks by an order of magnitude, and self-promoting Iceberg de-locking + the Fabric squeeze leave the moat trend narrowing. It is "not a bubble in relative terms" (growth-adjusted cheap, low historical percentile, a 21% sell-side discount), but the reference frame for "relatively cheap" (sell-side consensus) is chased higher by momentum and rests on the optimistic premise of "the reacceleration is sustainable + share is held."

The combined conclusion: the business quality deserves attention, and even deserves to be held long term at a reasonable price, but the current price does not deserve it — expensive + a stronger competitor + no margin of safety after the surge, with the relative advantage insufficient to offset the absolute valuation and competitive risk. Hence the rating of Watch (rather than Hold/Buy): quality better than prior 3/5 narrow-moat names, but certainty and margin of safety short of the "Hold" tier of a wide moat + reasonable valuation. Ideal buy ≤ $170; wait for the valuation to pull back or for further validation of AI monetization before turning more positive.

Research Uncertainties

• TAM definitions diverge by an order of magnitude (the company's broad definition of $170 billion vs. the third-party narrow definition of $10-23 billion); the judgment on penetration and the long runway depends heavily on the choice of definition, and this report presents both side by side rather than picking one.

• Databricks is a private company, and its ARR/growth/AI run-rate are taken from official press releases and secondary reports, with the IPO timing (reported Q3-2026) at the rumor level; its "AI revenue" is not fully comparable with Snowflake's (the order-of-magnitude gap holds, but there is uncertainty in the precise ratio).

• The Rule of 40 varies by definition (about 54 on the FCF basis vs. about 44 on the EBITDA basis), as noted in this report; the EV/Rev historical median of about 22x is on a since-IPO basis of about 5.5 years (not 10 years).

• The sustainability of the reacceleration is unverified — whether the single-quarter +34% is a pulse or a trend needs confirmation over the next 2-3 quarters; the guidance is on the company's basis and unrealized.

• The DCF fair value of ~$168 is sensitive to the discount-rate and terminal-value assumptions and should be treated as a central reference, not a precise level.

• The valuation scenario ranges are a framework-level estimate, not a price forecast; this report is research and analysis, not investment advice.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.