Snowflake
Designed by Oracle Alumni.
Bought by Warren Buffett.
In August 2020, Snowflake filed its S-1. Revenue for FY2020 (ended January 31, 2020) was $264.7 million, up 174% year over year — 2.7x in a single year.
And yet the number that draws the eye most is a different one. Net Revenue Retention Rate (NRR): 158%. Existing customers’ spending grew 58% in the most recent measurement. It is a structure that grows on the expansion of existing customers alone, even with zero new ones (though the S-1 expects this rate to decline over time).
And at the IPO it was disclosed that Berkshire Hathaway (Warren Buffett) would buy $250 million worth of stock (per the S-1/A). An investor known for not investing in technology IPOs placed a bet on Snowflake.
What the S-1 presents is the portrait of a company designed by Oracle alumni, finished by a serial-IPO CEO, and bought by Buffett.
Ten questions. We lay out only what the S-1 says.
S-1 JOURNAL NEWSLETTER
Get the Ten Questions checklist (PDF) and the next S-1 reality check, free by email.
Question 1
What is this company, really?
Primary source · from the S-1
Prospectus Summary (p.1+), Business (p.89+) — from the S-1
A company offering a platform to collect, analyze, and share data in the cloud.
The S-1 defines itself this way:
In plain terms: companies have data — sales, customers, logs, IoT. It sits scattered across AWS, Azure, and GCP (the three big clouds). Snowflake gathers all of it into one place where you can analyze it freely with SQL.
By analogy: a giant shared spreadsheet where every employee can analyze, all at once, the data that outgrew Excel. Except the speed is of a different order — it can aggregate billions of rows in seconds.
Snowflake’s distinction is being multi-cloud. It runs on AWS, on Azure, on GCP. It does not depend on any single cloud vendor.
It puts a company’s data in “one place,” across the walls between clouds.
Question 2
What does it compete on?
Primary source · from the S-1
Business (Competitive Strengths, p.91+) — from the S-1
Snowflake competes in four distinctive ways.
Separating storage and compute
A traditional data warehouse bundled storage and compute together. Adding data forced you to buy more compute too — and vice versa.
Snowflake separated storage and compute completely. Add data and you need not add compute. Run more analytics queries and your storage bill does not change. You pay only for what you use.
Multi-cloud support
The 2020 cloud market was a three-way contest among AWS, Azure, and GCP. Many companies use several in combination. Snowflake delivers the same experience across all three big clouds. Without moving data, you can use the compute resources of a different cloud.
Data Sharing (data exchange between companies)
Customers on Snowflake can share data with one another without copying it. The vision of the “Data Cloud” is to become not a mere database but a network of data.
Consumption-based pricing
Most SaaS charges a fixed monthly fee. Snowflake bills on consumption — pay for what you use. You are charged according to the compute consumed by running queries.
This is the structure that later produces the 158% NRR. Once a customer starts using Snowflake, data sources grow, analysis grows, and spending naturally swells.
Storage/compute separation, multi-cloud, data sharing, consumption pricing — four pillars.
Question 3
How big is the market?
Primary source · from the S-1
Business (Market Opportunity, p.90+) — from the S-1
The market numbers are simple.
In short: an $81 billion market, and Snowflake has not yet captured even 1% of it.
The S-1 stresses that the volume of corporate data is growing exponentially. IoT, mobile, streaming, social media — data flows in from every source, and the appetite for analysis keeps expanding.
| Metric | Figure |
|---|---|
| Data platform TAM (Snowflake’s own estimate, end of Jan 2020) | $81B |
| Snowflake FY2021 H1 revenue (annualized) | $484M |
| Penetration | 0.6% |
It is barely inside the door of the market.
Question 4
Who actually uses it?
Primary source · from the S-1
Business (Customers, p.97+), Prospectus Summary — from the S-1
Here are the customer metrics (as of the end of July 2020).
3,117 companies have begun using it, large accounts are surging, and existing customers’ spending recently grew 58%. Two numbers deserve attention.
Rapid growth in large accounts
Customers spending more than $1M grew from 22 to 56 (2.5x) in a single year. Adoption by large enterprises is accelerating.
What 158% NRR means
Existing customers’ spending grew 58% in the most recent measurement period. NRR of 158% means expansion overwhelmingly outweighs churn (the S-1 expects this rate to decline over time).
Customers line up across industries — Capital One (finance), Adobe (tech), Office Depot (retail), Sony (media), Instacart (food), DoorDash (food). Finance, tech, retail, media: a broad spread.
| Metric | Figure |
|---|---|
| Total customers | 3,117 |
| Customers spending $1M+ | 56 (2.5x the prior year’s 22) |
| Fortune 500 customers | 146 (29%) |
| Net Revenue Retention Rate | 158% |
A structure that grows on the expansion of existing customers alone, even with zero new ones.
Question 5
How does it make money?
Primary source · from the S-1
MD&A (p.60+), Business (p.89+) — from the S-1
Snowflake makes money one way: consumption-based pricing.
Customers buy Snowflake Credits in advance (or pay later), and credits are consumed every time a query runs.
The bulk of revenue is Compute. Storing data is billable too, but running analytics queries is the main revenue driver.
Even when a customer prepays for credits, revenue is recognized only when they are used. That makes Remaining Performance Obligations (RPO) — the unconsumed contracted amount — an important metric. At the time of the S-1, RPO was about $688 million.
Because it is consumption-based, the more a customer expands its use of Snowflake, the more it spends. Adding data sources, adding analytics users, adopting real-time analytics — each pushes up compute consumption. “More usage” translates directly into “more revenue.” This is what mechanically drives the 158% NRR.
| What is measured | Description |
|---|---|
| Compute | CPU and memory used to run queries and process data |
| Storage | Volume of stored data (in TB) |
| Data Transfer | Data movement between clouds and between regions |
A structure where “more usage” becomes “more revenue” is what mechanically drives the 158% NRR.
Question 6
Are the numbers healthy?
Primary source · from the S-1
Selected Financial Data, MD&A (p.60+) — from the S-1
Snowflake’s fiscal year ends in January. Two things stand out.
Start by laying out the financial facts.
The anomaly: operating loss exceeds revenue
FY2020 had revenue of $265M against an operating loss of $358M. A company spending more money than it takes in (net loss was $349M).
The reason is Sales & Marketing and R&D. For full-year FY2020, the two together exceed 150% of revenue (150.6%).
The saving grace: gross margin above 60% in H1 + RPO of $688M
Still, two things stand out — gross margin of 56–62% (FY2020 56%, H1 62%). The revenue itself is highly profitable (variable costs are low). And RPO of $688M: nearly $690 million of revenue is already locked in for future use.
In other words: a gross margin of about 60% is a level at which profitability is achievable — it depends on how operating expenses are allocated.
| Item | FY2020 | FY2021 H1 |
|---|---|---|
| Revenue | $264.7M | $242.0M |
| Gross profit | $148.2M | $149.0M |
| Gross margin | 56% | 62% |
| Operating loss | −$358.1M | −$174.1M |
| Net loss | −$348.5M | −$171.3M |
It holds “rapid growth” and a “huge loss” at once, but structurally it is a company that can turn a profit.
Question 7
Who runs it?
Primary source · from the S-1
Management (p.132+), Business — from the S-1
Snowflake has two teams: the one that built the technology, and the one that turned it into a selling machine.
The founding team (the architects of the technology)
Co-Founder and President of Products Benoit Dageville spent about 10 years at Oracle (2002–2012, an architect in the Manageability group) and knows database core technology inside out. With CTO Thierry Cruanes, he led the technical design at founding.
Having mastered the core of the world’s largest database company, Dageville broke at Snowflake the thing Oracle could not break — the coupling of storage and compute.
The CEO (a professional at accelerating growth)
Chairman & CEO (appointed April 2019) Frank Slootman was previously CEO of ServiceNow (2011–2017, leading its IPO), and before that CEO of Data Domain (acquired by EMC in 2009). The industry nickname for him is the “serial-IPO CEO” — an executive brought in to make an IPO succeed.
The meaning of hiring Slootman is plain: the technology is built; now run the growth machine toward an IPO. From his arrival in April 2019 to the S-1 filing was about 16 months.
The S-1’s biography records his ServiceNow years this way:
Oracle’s mind designs it, and an IPO professional finishes it — that is Snowflake’s management structure.
Question 8
Who controls it?
Primary source · from the S-1
Principal Stockholders (p.153+), Description of Capital Stock — from the S-1
Snowflake has a dual-class share structure.
There are two share classes. The Class A sold at IPO carries 1 vote per share. The Class B held by pre-IPO existing shareholders carries 10 votes per share and can convert to Class A at any time. The S-1 itself flags this dual class structure as a risk that voting power concentrates in existing shareholders.
The anomaly: it is VCs, not the founders, who hold the votes
A dual-class structure usually evokes founder control. But Snowflake is different. The personal stakes of founders Dageville and Cruanes are small; the bulk of the 10-vote stock (Class B) is held by VCs.
The largest is Sutter Hill Ventures (20.3%) — the VC that launched Snowflake in something close to co-founder fashion. Per the S-1/A, Class B as a whole accounts for 98.5% of the voting power, and directors, officers, and principal stockholders alone for 70.2%.
The IPO’s special participants: Buffett and Salesforce
Per the S-1/A, Berkshire Hathaway and Salesforce Ventures will each buy $250 million of Class A in a concurrent private placement at the IPO, at the IPO price.
The very fact that Buffett’s company — known for not investing in tech IPOs — participated lifted Snowflake’s valuation a notch.
| Principal stockholders (5%+, pre-IPO) | Ownership |
|---|---|
| Sutter Hill Ventures | 20.3% |
| Altimeter Partners Fund | 14.8% |
| ICONIQ Strategic Partners | 13.8% |
| Redpoint Ventures | 9.0% |
| Sequoia Capital | 8.4% |
It is VCs, not founders, who hold 98.5% of the voting power through dual-class — that is Snowflake’s governance.
Question 9
What is it afraid of?
Primary source · from the S-1
Risk Factors (p.14+) — from the S-1
Here are the five things Snowflake itself says it worries about most.
The landlord starts the same business
The most striking risk in the S-1 is this one.
In other words, the owners of the ground Snowflake runs on (AWS / Azure / GCP) compete directly through their own data-warehouse products. It is the picture of a landlord copying its tenant’s business.
Continuing losses
“We have a history of operating losses and may not achieve or sustain profitability in the future.” “We expect our costs and expenses to increase in future periods.”
The FY2020 operating loss of $358M exceeds revenue. The company writes, in its own words, that when it will reach profitability is unclear.
The uncertainty of consumption pricing
“Because our customers have flexibility in the timing of their consumption… There is a risk that customers will consume our platform more slowly than we expect.”
Most revenue comes from “capacity” contracts, in which customers commit to a certain amount of consumption in advance. But revenue is recognized only as customers actually use the platform, and the pace of that use is up to them. If consumption comes in more slowly than expected, revenue lands more slowly too.
Security and data breaches
A security breach could lead to loss of customer data and reputational harm. Snowflake holds its customers’ most sensitive data; a single breach could cause a fatal loss of trust.
Dependence on management (the newly arrived Slootman and others)
The S-1 cites dependence on recently added executives (Slootman, appointed April 2019, and CFO Scarpelli, appointed August). “Our success depends largely upon the continued services of these and other executive officers.”
If Slootman and the others leave, the operational muscle behind the growth is lost.
The landlord as competitor, continuing losses, customers who control the pace of consumption, data breaches, and dependence on management — five fears.
Question 10
What becomes of the money it raises?
Primary source · from the S-1
Use of Proceeds (p.54+), Cover — from the S-1
Note: This analysis draws only on the facts in the S-1. The final IPO price, the listing date, post-listing share performance and trading demand are not in the filing and fall outside its scope.
The IPO plan is as follows.
Investors already committed
At the S-1/A stage, two large investments are locked in. Berkshire Hathaway and Salesforce Ventures will each buy $250 million of Class A in a concurrent private placement at the IPO.
Buffett participating in a tech IPO is extremely unusual. The very fact that this participation is spelled out in the S-1/A is a message to the market about Snowflake’s standing.
Use of proceeds
The use is “general corporate purposes” and “working capital, operating expenses, capital expenditures.”
No specific allocation (so much for multi-cloud support, so much for overseas expansion) is spelled out. The posture is “raise it, and the company decides how to use it.”
Not a stopgap but a scale-up IPO
Snowflake already runs at a gross margin above 60% and holds $688M of unconsumed RPO. This is not an IPO driven by a cash crunch.
The meaning of the IPO is “capital-market access to run Slootman’s growth machine at full throttle,” and “the validation of Buffett’s participation.”
| Item | Detail |
|---|---|
| Listing | NYSE (ticker: SNOW) |
| Assumed price | $75–$85 (initial S-1/A, Sep 2020; final offering price was $120) |
| Lead underwriters | Goldman Sachs, Morgan Stanley, JPMorgan |
The IPO is not about cash but about capital access for the growth machine and the validation of Buffett’s participation.
In closing
Beyond the ten questions
What Snowflake is selling is not just a “cloud data warehouse.” FY2020 revenue of $264.7M (up 174%), 3,117 customers, 146 of the Fortune 500, Net Revenue Retention of 158% (a trailing six-month figure; the S-1 expects it to decline) — a company with a growth curve where existing customers’ expanding spend physically compounds through consumption pricing.
But the ground it runs on is owned by AWS, Azure, and GCP (the three big clouds), each of which competes directly through its own data-warehouse product. The FY2020 operating loss of $358M exceeds revenue (net loss $349M). Sales & Marketing and R&D consume more than 150% of revenue. Gross margin tops 60% in H1, and RPO of $688M underpins “profitability is achievable,” but the company itself will not say when.
Designed largely by Oracle alumnus Dageville (about 10 years there), and brought to the S-1 filing in about 16 months by Frank Slootman, who finished two IPOs at ServiceNow and Data Domain. And Berkshire Hathaway’s $250 million participation in a tech IPO is spelled out in the S-1/A — that is the self-assessment Snowflake presents to the market.
LIGHT
The bull case
· FY2020 revenue of $264.7M (+174%), NRR 158% (trailing figure, expected to decline), 3,117 customers, 56 accounts at $1M+, 146 of the Fortune 500.· Gross margin above 60% in H1 (FY2020 56%), RPO of $688M.· Berkshire Hathaway and Salesforce each participate with $250M at the IPO (per the S-1/A).· Oracle alumnus Dageville led the design; serial-IPO CEO Slootman finished it.
SHADOW
The bear case
· Operating loss of $358M exceeds revenue; Sales & Marketing + R&D consume more than 150% of revenue.· Landlords AWS/Azure/GCP compete directly through their own data-warehouse products.· Because pricing is consumption-based, revenue is recognized only as customers consume — if they use the platform more slowly than expected, revenue slips too.· The company discloses its own dependence on newly arrived management like Slootman.· VCs hold the bulk; the founders are not the largest shareholders.
The S-1 writes of “freeing a company’s data with the Data Cloud.” The market expects “the next ruler of the cloud.” At the intersection of the two, Snowflake sells stock. It is the moment Oracle alumni broke, in the cloud, the structure Oracle could not break.
S-1 JOURNAL NEWSLETTER
Event-driven S-1 breakdowns and new deep dives, delivered free by email. Subscriber bonus: the Ten Questions checklist (PDF). No spam. Unsubscribe anytime.
By subscribing you agree to receive emails from S-1 Journal.
That judgment is now in your hands.
Source: All facts are drawn from Snowflake Inc.’s S-1 (accession 0001628280-20-013010, filed 2020-08-24), together with the company’s IPO roadshow presentation and factsheet where noted.