Snowflake

Snowflake grew revenue more than thirteenfold and still made buyers of its record-breaking first-day price wait five years to break even.

Snowflake priced at $120 on September 16, 2020 after twice raising its range, then closed its first session at roughly double that — the largest software IPO ever at the time, with Berkshire Hathaway and Salesforce anchoring the deal. The S-1 had been unusually blunt about the two things that could break the story: that AWS, Azure and GCP are simultaneously its landlords, its cost structure and its competitors, and that a consumption model gives management no visibility into when revenue lands. It also stated flatly that it might never be profitable. The market bought it anyway, driving the stock to $392 in November 2021 — a $120bn+ valuation on roughly $1.2bn of annualized revenue.

The operating business largely delivered. Revenue climbed from $264.7m in FY2020 to roughly $3.6bn in FY2025, international mix expanded, the platform broadened from warehousing into data lakes, Snowpark, data sharing and later AI/Cortex workloads, and product gross margin improved from the low 60s toward the mid-70s — quietly defusing the fear that cloud vendors would squeeze margins. What did not deliver was the shape of growth or the bottom line. Net revenue retention fell from 169% at IPO to the mid-120s; consumption optimizations, storage-compression efficiencies and customer cost discipline forced repeated guidance resets in 2022-2024, exactly the unforecastability the filing warned about. GAAP net losses widened in absolute dollars past $1bn on heavy stock-based compensation, even as free cash flow turned strongly positive. Databricks, not Amazon Redshift, became the competitive story, and Microsoft Fabric arrived in 2023.

Two shocks defined the middle years. In February 2024 Frank Slootman abruptly retired and Sridhar Ramaswamy took over; the stock fell roughly 20% in a day, validating the S-1's key-person language in the opposite direction from how it was written. Then in mid-2024, attackers used stolen customer credentials to exfiltrate data from Snowflake-hosted accounts belonging to Ticketmaster, AT&T and others — Snowflake maintained its own platform was not breached and the accounts lacked MFA, but the incident dominated headlines and pushed the company to mandate MFA. The stock bottomed at $108.56 on September 1, 2024, below the IPO price. From there the AI-data narrative and re-accelerating product revenue drove a recovery to $328 as of August 2026: +173% versus the $120 IPO price, but only about +37% versus the $240 first-day close, over nearly six years.

The scoreboard is a case study in the gap between a great business and a great entry price. Nearly every operational promise in the best-case narrative was at least partially kept; nearly every warning in the risk factors also proved real. The variable that mattered most for investors was the one the S-1 could not control — what the market paid on day one.

Snowflake is pitching itself as the Cloud Data Platform that replaces legacy on-premises data warehouses and fragmented big-data systems with a single, multi-cloud, consumption-priced service — growing revenue 174% to $264.7 million in fiscal 2020 while still losing money and running entirely on top of the three public clouds (AWS, Azure, GCP) that are also its biggest competitors.

The bull case is that data warehousing is migrating wholesale to the cloud and Snowflake's architecture — separating compute from storage, working identically across all three major clouds, and enabling cross-organization data sharing — becomes the default place enterprises put all their data. Consumption-based pricing means revenue expands automatically as customers migrate more workloads, so a land-and-expand motion into the world's largest organizations compounds: new use cases (data lakes, pipelines, data applications, data exchange) widen the wallet beyond the original data warehousing entry point. Snowflake is early internationally (only 12% of FY2020 revenue from outside the U.S.) and early in the public sector and regulated industries, giving it multiple untapped vectors. A new, IPO-proven management team under Frank Slootman and Michael Scarpelli is being brought in to industrialize enterprise sales, and heavy investment in professional services is presented as an adoption accelerant that will later be repriced for margin.

With hindsight: Revenue went from $264.7m in FY2020 to roughly $3.6bn in FY2025, international expanded, the multi-cloud architecture held, and use cases broadened from warehousing into pipelines, data sharing, Snowpark and AI — the land-and-expand thesis worked. But the compounding was far less automatic than pitched: net revenue retention fell from 169% toward the mid-120s, consumption optimization forced repeated guidance resets, Databricks emerged as a formidable rival, and GAAP profitability never arrived, with net losses exceeding $1bn on stock comp. Anyone who bought at the $240 first-day close earned roughly 37% over six years.

• Revenue grew from $96.7 million in FY2019 to $264.7 million in FY2020, and from $104.0 million to $242.0 million in the comparable six-month periods.

  “Our revenue was $96.7 million and $264.7 million for the fiscal years ended January 31, 2019 and 2020 , respectively, and $104.0 million and $242.0 million for the six months ended July 31, 2019 and 2020, respectively.” source ↗

• The platform has expanded beyond its original data warehousing use case into multiple adjacent workloads, which management frames as the core growth engine.

  “We introduced data warehousing on our platform in 2014 as our core use case. In recent years, customers have begun using our platform for additional use cases, including data pipelines, data lakes, data application development, and data sharing and exchange.” source ↗

• International expansion is a largely untapped growth vector, with non-U.S. customers contributing only 12% of FY2020 revenue.

  “A component of our growth strategy involves the further expansion of our operations and customer base internationally. Customer accounts outside the United States generated 12% of our revenue for the fiscal year ended January 31, 2020 .” source ↗

• Multi-cloud availability across AWS, Azure, and GCP is positioned as a competitive differentiator versus single-cloud offerings.

  “We compete based on various factors, including price, performance, breadth of use cases, multi-cloud availability, brand recognition and reputation, customer support, and differentiated capabilities, including ease of implementation and data migration” source ↗

• Professional services are deliberately run at low margin today to drive adoption, with a stated plan to reprice for profitability later.

  “In the future, we intend to price our professional services based on the anticipated cost of those services and, as a result, expect to improve the gross profit percentage of our professional services business.” source ↗

• Snowflake is investing to unlock federal, state, local and regulated-industry demand, holding FedRAMP, PCI-DSS and ISO 27001 among other certifications.

  “The certifications we maintain and standards we comply with, including the U.S. Federal Risk and Authorization Management Program, PCI-DSS, ISO/IEC 27001, among others, are becoming more stringent.” source ↗

1. The three companies Snowflake runs on are also its main competitors

   partly came true

   platform dependence · structural

   Snowflake's entire platform is hosted on AWS, Azure and GCP — a substantial majority on AWS — and each of those providers sells competing data warehouse products. They control pricing, interoperability and customer relationships, and could bundle, throttle, or disadvantage Snowflake at will.

   “We currently only offer our platform on the public clouds provided by AWS, Azure, and GCP, which are also some of our primary competitors. Currently, a substantial majority of our business is run on the AWS public cloud.” source ↗

   What happened: AWS, Azure and GCP kept selling Redshift, Synapse/Fabric and BigQuery, and Microsoft's 2023 Fabric launch was an explicit bundling threat, but the hyperscalers also became major co-sell partners and never throttled or excluded Snowflake. The competitor that actually compressed Snowflake's narrative was Databricks, not a cloud landlord.

2. Cloud vendor pricing directly determines gross margin

   didn't happen

   profitability · structural

   Cost of revenue is largely the compute and storage Snowflake buys from AWS/Azure/GCP, so gross margin is set by negotiations with competitors who have no incentive to give favorable terms.

   “our costs and gross margins are significantly influenced by the prices we are able to negotiate with these public cloud providers, which in certain cases are also our competitors” source ↗

   What happened: Product gross margin improved from roughly the low 60s at IPO to the mid-70s as Snowflake negotiated better cloud committed-spend terms and improved efficiency. Cloud vendor pricing turned out to be a tailwind, not the margin cap the S-1 feared.

3. Large and growing losses with no stated path to profitability

   came true

   profitability · structural

   Net loss nearly doubled to $348.5 million in FY2020 with an accumulated deficit of $871.6 million as of July 31, 2020, and the company expects costs to keep rising while explicitly stating it may never be profitable.

   “We generated net losses of $178.0 million and $348.5 million for the fiscal years ended January 31, 2019 and 2020 , respectively, and $177.2 million and $171.3 million for the six months ended July 31, 2019 and 2020, respectively.” source ↗

   What happened: Snowflake remained GAAP-unprofitable every year after the IPO, with net losses widening in absolute terms past $1bn by FY2025, driven largely by stock-based compensation. The company did reach substantial positive free cash flow and non-GAAP operating profit, but the S-1's 'may never achieve profitability' line remained literally true on a GAAP basis.

4. Consumption-based model means no revenue visibility

   came true

   growth · structural

   Unlike a subscription model, revenue is recognized when customers actually consume compute and storage, so Snowflake cannot predict the timing of revenue and warns investors may mis-model the business against SaaS comparables.

   “Because our customers have flexibility in the timing of their consumption, we do not have the visibility into the timing of revenue recognition that a typical subscription-based software company has.” source ↗

   What happened: Snowflake repeatedly reset guidance as customers optimized spend and as platform efficiency gains (including compute and storage improvements) reduced credits consumed for the same workloads, most visibly in 2022-2024. The stock's largest single-day declines were tied to consumption-driven guidance disappointments, exactly the modeling mismatch the filing flagged.

5. Entire executive team is new and untested together

   came true

   key person · serious

   The CEO joined April 2019 and the CFO August 2019, following the prior CEO's resignation; the filing concedes the leadership team's ability to work with employees and lead the company is unproven.

   “Many of our executive officers and other members of our management team have been with us for a short period of time, including Frank Slootman, our Chairman and Chief Executive Officer, who joined us in April 2019, and Michael P. Scarpelli, our Chief Financial Officer, who joined us in August 2019.” source ↗

   What happened: Frank Slootman abruptly retired as CEO in February 2024, handing the job to Sridhar Ramaswamy, and the stock fell roughly 20% on the announcement; CFO Michael Scarpelli subsequently transitioned out of the role as well. The Slootman-Scarpelli team executed strongly through 2023, but their departure proved to be a material, market-moving key-person event.

6. Customer concentration, and that customer is an underwriter affiliate

   didn't happen

   operations · serious

   Capital One Services was 17% then 11% of revenue in FY2019 and FY2020, and an affiliate of that customer is one of the offering's underwriters — a notable related-party overlap.

   “Capital One Services, LLC (an affiliate of Capital One Securities, Inc, one of the underwriters in this offering) accounted for approximately 17% and 11% of our revenue, respectively” source ↗

   What happened: Capital One's share of revenue fell below disclosure thresholds as the customer base scaled past 10,000 accounts; no single customer concentration issue has been reported as material post-IPO.

7. Extremely short operating history relative to valuation

   came true

   growth · serious

   Founded 2012 and first selling in 2014, Snowflake tells investors flatly not to extrapolate its historical growth rates.

   “We were founded in 2012 and first offered our platform for sale in 2014.” source ↗

   What happened: Management's warning not to extrapolate growth was correct: revenue growth decelerated from 174% in FY2020 to roughly 26-30% by FY2025. Investors who paid the November 2021 peak of $392 waited more than four years to recover, with the stock trading below the $120 IPO price as late as September 2024.

8. Security breach exposure with admission of prior attacks

   came true

   tech & security · serious

   The platform stores customers' personal, health and financial data on third-party clouds, and Snowflake discloses it has already been the target of cyber-attacks and may not have adequate insurance.

   “We have previously been, and may in the future become, the target of cyber-attacks by third parties seeking unauthorized access to our or our customers’ data or to disrupt our operations or ability to provide our services.” source ↗

   What happened: In mid-2024 attackers used stolen customer credentials to exfiltrate data from Snowflake-hosted environments belonging to Ticketmaster, AT&T, Santander and roughly 100+ other accounts. Snowflake maintained its own platform was not compromised and that affected accounts lacked MFA, but the episode dominated coverage and led the company to mandate multi-factor authentication.

9. Service disruptions already experienced and largely outside Snowflake's control

   didn't happen

   operations · serious

   Snowflake commits to minimum availability SLAs but depends on public cloud uptime; it admits to past outages and to lacking contractual recourse against cloud providers in some cases.

   “In some cases, we may not have a contractual right with our public cloud providers that compensates us for any losses due to availability interruptions in the public cloud.” source ↗

   What happened: No publicly reported outage since the IPO has been material enough to drive disclosed SLA credits, customer loss or a significant stock reaction.

10. Unprofitable professional services drag on margins

    partly came true

    profitability · serious

    Professional services have been priced to drive adoption rather than profit and are growing alongside product revenue, pressuring overall margins until repricing succeeds.

    “our sales efforts have focused on helping our customers realize the value of our platform rather than on the profitability of our professional services business” source ↗

    What happened: Professional services remained a low- or negative-margin line and continued to dilute total gross margin relative to product gross margin, but product margin expansion more than offset it and total gross margin rose after the IPO.

11. Dual-class structure and index exclusion

    didn't happen

    governance · serious

    Class B shares carry ten votes each and are held entirely by pre-IPO holders, concentrating control; the structure also makes Snowflake ineligible for major stock indices.

    “Our Class B common stock has ten votes per share, whereas our Class A common stock, which is the stock we are offering in this offering, has one vote per share.” source ↗

    What happened: The ten-vote Class B structure sunsetted under the charter's automatic conversion provisions within roughly two years of listing, collapsing Snowflake into a single class of common stock and removing the index-eligibility problem. No post-IPO controversy over insider voting control emerged.

12. Active trademark infringement lawsuit

    didn't happen

    legal · serious

    Yeti Data sued Snowflake weeks before the filing seeking a permanent injunction against infringement, and the company acknowledges its patent portfolio is relatively undeveloped versus large competitors.

    “on July 24, 2020, Yeti Data, Inc. (Yeti Data) filed a lawsuit against us in the U.S. District Court for the Central District of California alleging trademark infringement and other ancillary claims” source ↗

    What happened: The Yeti Data trademark case was resolved without any publicly reported injunction, rebranding, or financially material outcome for Snowflake.

13. Privacy regulation could reshape the product itself

    didn't happen

    regulation · serious

    GDPR, CCPA, data sovereignty rules and possible AI regulation could change how the platform is designed and how customers may use data, with GDPR fines up to 4% of worldwide revenue.

    “if regulators assert that we have failed to comply with the GDPR, we may be subject to fines of up to EUR 20 million or 4% of our worldwide annual revenue, whichever is greater” source ↗

    What happened: No GDPR fine or regulatory action against Snowflake has been publicly reported, and data-sovereignty requirements were addressed through regional deployments rather than product redesign. If anything, regulated-industry and government demand grew, consistent with the bull case.

14. Enterprise-heavy sales model creates long cycles and Q4 seasonality

    partly came true

    growth · boilerplate

    The move upmarket brings long, unpredictable sales cycles, upfront costs with no guaranteed payoff, and increasingly pronounced fourth-quarter order concentration.

    “Historically, we have received a higher volume of orders from new and existing customers in the fourth fiscal quarter of each year. We expect this seasonality to become more pronounced as we continue to target large enterprise customers.” source ↗

    What happened: Q4 remained the seasonally strongest bookings quarter and enterprise sales cycles lengthened during the 2022-2023 IT spending downturn, contributing to slower net-new-customer growth. This was a persistent drag but never a thesis-breaking one.

15. Emerging growth company reduced disclosure and delayed auditor attestation

    didn't happen

    governance · boilerplate

    Snowflake elects JOBS Act exemptions, deferring auditor attestation on internal controls and using private-company accounting transition periods; management's own Section 404 report only comes as of January 31, 2022.

    “We will be required, pursuant to Section 404, to furnish a report by management on, among other things, the effectiveness of our internal control over financial reporting as of January 31, 2022.” source ↗

    What happened: Snowflake's size pushed it out of emerging growth company status quickly, bringing full large-accelerated-filer disclosure and auditor attestation on internal controls. No material weakness or restatement has been reported.

How the S-1 reads

The filing is unusually direct about the two things that most threaten the business — that its suppliers are its competitors, and that its consumption-based revenue is inherently unforecastable — devoting dedicated risk headings to each rather than burying them. Where most S-1s hedge, this one uses flat declaratives ('we may never achieve profitability,' 'we have previously been... the target of cyber-attacks,' 'conducting business virtually is unproven'), and it volunteers a granular, self-critical COVID list including employee morale and stranded real estate costs. Governance is maximally founder-and-insider protective: ten-vote Class B stock, a classified board, for-cause-only removal, a supermajority amendment threshold, and Securities Act forum selection, paired with JOBS Act reduced disclosure. The compensation section is striking for the scale of new-hire option grants relative to modest cash salaries, and for the fact that the prior CEO's exit package — acceleration and vesting modification worth over $16 million — dwarfs the salary lines it sits beside.

• “There is risk that one or more of these public cloud providers could use their respective control of their public clouds to embed innovations or privileged interoperating capabilities in competing products, bundle competing products, provide us unfavorable pricing, leverage its public cloud customer relationships to exclude us from opportunities” source ↗

  The most consequential admission in the filing: Snowflake's landlords are its rivals.

• “Further, investors and securities analysts may not understand how our consumption-based business model differs from a subscription-based business model, and our business model may be compared to subscription-based business models.” source ↗

  Explicit warning that the consumption model may be misunderstood by the market.

• “We have incurred substantial losses during our history, do not expect to become profitable in the near future, and may never achieve profitability.” source ↗

  Blunt statement on profitability prospects buried in the tax discussion.

• “However, you should not rely on the revenue growth of any prior quarterly or annual period as an indication of our future performance.” source ↗

  Management disclaims its own growth trajectory as a guide to the future.

• “Market opportunity estimates and growth forecasts included in this prospectus, including those we have generated ourselves, are subject to significant uncertainty and are based on assumptions and estimates that may not prove to be accurate.” source ↗

  Concedes its market-size figures are self-generated and may not be reliable.

• “As compared to our large competitors, our patent portfolio is relatively undeveloped and may not provide a material deterrent to such assertions or provide us with a strong basis to counterclaim or negotiate settlements.” source ↗

  Acknowledges its IP position is weak relative to incumbents.

• “Many of our existing security holders have substantial unrecognized gains on the value of the equity they hold based upon the price of this offering, and therefore, they may take steps to sell their shares or otherwise secure the unrecognized gains on those shares.” source ↗

  Unusual acknowledgment that insiders may rush to sell into the offering's aftermath.

What this one teaches

• An S-1 that names its worst structural risk in plain language does not neutralize it — but it does let you check the scorecard later. Snowflake's cloud-dependence warning proved overstated while its consumption-visibility warning proved exactly right; candor and accuracy are different things.
• A 100% first-day pop is a transfer of value, not a validation. Snowflake's business grew more than thirteenfold in six years while buyers at the first-day close earned roughly 37% — the IPO price, not the business, determined the return.
• 'We may never achieve profitability' should be read literally, not as boilerplate. Six years and $3.6bn of revenue later, Snowflake was still GAAP-unprofitable because stock-based compensation scaled with the business.
• Key-person risk factors are written about departures the company does not anticipate. The S-1 worried that a brand-new executive team was untested; the actual damage came when that team, having proven itself, left — costing roughly a fifth of the market cap in a day.

The paper trail

1. 2020-08-24 S-1 filing index ↗ document ↗
2. 2020-09-08 S-1/A filing index ↗ document ↗
3. 2020-09-14 S-1/A filing index ↗ document ↗
4. 2020-09-16 424B4 filing index ↗ document ↗

Filed as Snowflake Inc.. All documents are public domain, served by SEC EDGAR.