Snowflake

IThe business

Snowflake sells computing and storage for data, billed by use. A company puts its data in Snowflake and runs queries, data engineering, machine learning and, since 2025, a Postgres database on it. The platform runs on Amazon Web Services, Microsoft Azure and Google Cloud, in 53 regional deployments.⁵

Customers usually sign a capacity contract of one to four years and draw it down as they use the platform, so product revenue is recognised as it is consumed.⁵¹² In FY2026 revenue was $4,684M, up 29%, and in the twelve months to July 2026 it was $5,435M.C (calculated) Since FY2021 revenue has grown 51.2% a year. GAAP operating income has been a loss in every year, −$1,435M in FY2026, because stock pay is charged as an expense.⁴⁵

Snowflake reports one segment, so there is no profit by product or region. Three-quarters of revenue comes from the United States, and no customer is 10% of revenue.⁵

How it makes money

• Product revenue, 96% of the total. $5.21B in the last twelve months, up 33%.C (calculated) Customers pay for computing, storage and data transfer as they use them. Remaining performance obligations, the contracted revenue not yet recognised, were $9.0B at 31 July 2026, up 30%, or 1.7 times the last twelve months of product revenue; 54% is due within a year.C (calculated)¹²
• Professional services and other, 4%. Its gross margin was negative 37% in the six months to July 2026.¹²¹⁴
• The customers. 14,554 at 31 July 2026. 829 are in the Forbes Global 2000 and made about 41% of revenue in the six months; 828 pay over $1M a year, up 27% in a year. The two counts come from different lists and happen to be nearly equal.C (calculated)¹²
• AI. The CEO says AI products “contributed approximately half of the acceleration”. CoCo has 9,100 accounts and CoWork 5,800. Neither the filings nor the call give an AI revenue figure.¹⁴²²

Growth is speeding up

Product revenue growth rose from 28.7% in the October 2025 quarter to 36.8% in July 2026, the third quarter in a row of faster growth. Observe, bought on 2 February 2026, adds about 1 point, so growth without it was about 33% in April and 36% in July.E (estimate)¹⁴¹⁵²² Snowflake added a record $157.5M of product revenue over the quarter before, after $107.7M in the quarter to April.C (calculated)¹⁴¹⁵

It beat the top of its guided range by 2.2% to 5.3% in each of the last four quarters, and the proxy says the targets in its pay plan sit above its public guidance, so I treat the guide as a floor.^141516171826 The FY2027 product revenue guide has gone from $5,660M in February to $6,070M in September, from 27% to 36% growth.¹⁴¹⁶

The leading indicator is weaker. Remaining performance obligations grew 30% in the year to July, while product revenue grew 37%. In January they grew 42% while revenue grew 30%, but January is the seasonal peak, so I compare July with July. The CFO estimates that the 54% due within 12 months grew about 42%.C (calculated)^12141622

Growth is 37%, about 36% without Observe, and contracted revenue is growing 30%.

IIThe moat

Snowflake has one moat, a switching cost made of data and workloads, with a thin network effect on top. It is real, and it is narrowing at the edges.

A company that moves its data into Snowflake then builds pipelines, access rules, applications and reports on it. Leaving means rebuilding all of that. The evidence is in retention. Net revenue retention, which measures how much more last year’s customers spent this year, was 178% in FY2022 and has been 125% or 126% for five quarters, on a base of more than $5B.⁹¹²

The network effect is data sharing. The CEO says 43% of customers share data on Snowflake with at least one “stable edge”, a term the call does not define.²² The 10-K says the business benefits from network effects and gives no usage figures, so I count the effect as small.⁵

Snowflake’s own risk factor says there is “less customer ‘lock in’” when its products are used in external environments, and that its support of open data formats may also reduce switching costs.⁵ The cost of delivering the revenue is growing faster than the revenue, and the CFO ties part of it to AI:

AI workloads, which carry a lower contribution margin today.

Cost of product revenue grew 44% in the July quarter against 37% for product revenue, and $113.6M of its $132.1M rise was third-party cloud cost, including AI inference and GPUs.C (calculated)¹² Snowflake cut its FY2027 non-GAAP product gross margin outlook from 75% to 74% in September.¹⁴¹⁶ Databricks, meanwhile, is growing faster, and the gap is widening.

Retention has held at 125% to 126% for five quarters, and the filing’s own risk factor says open formats may lower switching costs.

IIIThe competitive advantage

Snowflake’s rivals sit in two groups, the three clouds that host it and Databricks. The clouds are suppliers and rivals at once. The 10-K says they “generally compete in all of our markets” and that “a substantial majority of our business is run on the AWS public cloud”.⁵ In April 2026 Snowflake amended a cloud agreement to a $6.0B minimum over five years to March 2031, $0.9B to $1.25B in each contract year, and in June 2026 it added storage commitments to May 2036 with a minimum of about $58M.¹²¹³ The 10-Q does not name the provider. The Q1 release describes a new $6 billion AWS agreement, so I link the two, but the link is inferred.¹³¹⁵

None of the clouds reports revenue for its warehouse, so Redshift, BigQuery and Fabric cannot be sized.³⁹⁴⁰⁴¹ Databricks’ figures are its own, unaudited, and on a company-defined run-rate basis.³⁷³⁸

Where Snowflake leads

Neutrality. Snowflake runs the same platform on all three clouds, and it announced $200M partnerships with Anthropic in December 2025 and OpenAI in February 2026. Neither release says who pays whom.⁴²⁴³ A customer that does not want to commit to one cloud or one model lab has a place to put its data.

Reach. 829 of the Forbes Global 2000 are customers, and 692 net new customers joined in the July quarter, 32% more than a year earlier.¹²¹⁴

Cash. Free cash flow was 22% of revenue in the last twelve months, while Databricks says only that its adjusted free cash flow was positive over twelve months.C (calculated)¹²³⁷

Where it trails

Databricks is bigger and, on its own measure, faster. It reports a company-defined revenue run-rate over $7B, growing over 80%, versus Snowflake’s annualised product revenue of $5.97B, growing 37%.C (calculated) Its Lakehouse warehouse product has over $1.5B of run-rate, growing over 100%: a quarter of Snowflake’s size, growing nearly three times as fast, though run-rate is not recognised revenue.C (calculated)¹⁴³⁷ Databricks has over 1,000 customers consuming above $1M of run-rate, and Snowflake has 828.¹²³⁷

Its August 2026 round was at a $190B valuation, 27 times its run-rate, or 26 times after the $5B it raised if the valuation is post-money; the release does not say which. Snowflake’s enterprise value is 19 times its annualised revenue. Divided by growth, that is about 0.5 for Snowflake and 0.3 or less for Databricks. A private round is a mark, not a trade.C (calculated)¹²³⁷

Third-party cloud was about 75% of Snowflake’s cost of product revenue in the July quarter, up from 70% a year earlier, and the 10-Q names Anthropic and OpenAI beside the three clouds as providers whose prices set its costs and gross margins.¹² The clouds also offer warehouses of their own. A cloud that bundles one, which the 10-K lists as a risk, does not have to earn Snowflake’s margin on it.⁵

IVManagement

Sridhar Ramaswamy has been chief executive since February 2024. He joined in 2023 when Snowflake bought Neeva, his search company, for $185.4M in cash. Frank Slootman, chief executive from April 2019 to February 2024, is chairman. The company has since changed its chief financial officer and, twice, its chief revenue officer.^2631323334 The shares rose 63% from the 29 February 2024 close of $188.28 to $306.08 in Q4 2026, versus 49% for the S&P 500, both on price alone.C (calculated)³⁴⁶ The proxy’s five-year line runs the other way: $100 at 31 January 2021, when the shares closed at $272.45, 2.3 times the $120 listing price, was worth $70.73 at 31 January 2026, while the S&P 500 Information Technology index was worth $256.43.C (calculated)³¹¹²⁶

Management earns high marks for execution. Product revenue growth rose from 28.7% to 36.8% in three quarters, the non-GAAP operating margin guide for FY2027 rose two points to 14.5%, and headcount grew by 334 in the first half of FY2027, 173 of them from Observe, versus 935 a year earlier.¹⁴¹⁶²² Stock pay fell from 49.6% of revenue in FY2022 to 30.2% in the last twelve months.C (calculated)⁵⁹¹² It earns low marks for pay: no measure charges management for the stock it issues, and the board granted a new award two weeks after the second failed vote.

Capital allocation

Snowflake spends cash to hold its share count level, and the count still rises. Since February 2023 it has bought back 25.4M shares for $3.70B, an average of $145.69, worth $7.77B at $306.08.C (calculated) It also paid $1.87B in taxes on shares withheld from vesting awards from FY2024 to July 2026.C (calculated) Shares outstanding went from 333.9M in January 2025 to 343.9M in January 2026 and 352.5M in July 2026, up 2.5% in six months.⁵⁶⁷¹²

Acquisitions total about $2.1B across eight deals since FY2023, about half paid in stock. The largest were Streamlit at $650.8M and Observe at $595.8M.C (calculated)⁵¹² Observe was a related-party deal: the lead independent director, the chairman, a former director and Sutter Hill Ventures, where the lead director is a managing director, together received about $250M, $208.2M of it in Snowflake stock for Sutter Hill.C (calculated)²⁶³⁵ A committee of independent directors approved it with a fairness opinion.

The 0% convertible notes are a cost of capital that does not show in interest. In September 2026 Snowflake spent $548.3M to retire $261.8M of 2027 notes, 2.1 times their principal, because those notes had become worth far more than par.²³²⁴

Incentives

The pay plan rewards revenue, the non-GAAP operating margin and non-GAAP adjusted free cash flow: for FY2027, executives’ awards vest on three years of product revenue growth (70%), non-GAAP operating margin (15%) and adjusted free cash flow (15%).²⁶ Both profit measures add stock pay back, so management can raise them by paying more in stock. None measures stock pay, dilution or return on capital.

On 15 July 2026 the compensation committee granted the CEO 1,000,000 performance units with share-price hurdles of $324 to $531, designed to add up to $100B of value from the $271.87 close on the grant date.³⁰ The average close over the 90 days to Q4 2026, $330.63, is already above the first hurdle, so the first tranche is met on price; the service date of 15 September 2029 and the committee’s certification still apply.C (calculated)³³⁰ The award pays on the market’s willingness to pay, not on the business reaching my bull case: the stock needs to reach $531 for the full award, 73% above today, and my bull value is $299.C (calculated)

Shareholders objected to the previous grants. Say-on-pay failed in 2025 with 30% in favour and again in 2026 with 44%, and the board made the July grant two weeks after the second vote.²⁸²⁹³⁰ All directors and executive officers together own 4.8%, most of it options and founders’ holdings; Vanguard is the only holder above 5%.²⁶

Management is paid to grow revenue and to add stock pay back, and the July award pays on the share price, not on cash an owner keeps. The CEO’s directly held shares are worth about $53M, and all directors and officers own 4.8%, mostly options.²⁶ The buybacks cost $3.70B at an average of $145.69 and look good at $306.08, but the share count still rose 2.5% in six months.¹²

VFinancials

Revenue rose from $592M in FY2021 to $4,684M in FY2026. The GAAP operating margin improved from −91.9% to −30.6% and was −22.8% in the last twelve months. Free cash flow turned positive in FY2022 and was 22.0% of revenue in the last twelve months.⁴⁵¹²

Free cash flow is $1.2B. After the $1.6B of stock pay that helps produce it, it is negative $447M, and the Rule of 40 falls from 54 to 24.

Stock is not free: Snowflake spent $1.39B on repurchases and share-withholding taxes in the last twelve months, more than the $1.20B of free cash flow, and the share count still rose.C (calculated)¹² Free cash flow after stock pay is negative in every year in the table below.C (calculated)

Stock pay as a share of revenue has fallen 19 points since FY2022, from 49.6% to 30.2%, while the dollar amount rose from $605M to $1.6B.C (calculated)⁵⁹¹² The July quarter’s charge, including payroll tax, was 29.5% of revenue, down from 38.1% a year earlier.C (calculated)¹⁴ Salesforce, ServiceNow and Workday spent 10.5%, 19.2% and 21.4% of revenue on stock pay in the year each passed $5B of revenue.C (calculated)⁴⁷⁴⁸⁴⁹

The non-GAAP profit in the headline adds back $1,756M of stock-related charges in the last twelve months, 32.3% of revenue.¹⁴

Efficiency

Sales and marketing fell from 61.0% of revenue in FY2022 to 42.4% in the last twelve months. Each dollar of the previous year’s sales and marketing produced $0.70 of new product revenue in the last twelve months, up from $0.57 in FY2025, and the incremental non-GAAP operating margin was 23%.C (calculated) The Rule of 40, revenue growth plus free cash flow margin, is 54 as reported and 24 after stock pay; it was 13 after stock pay in FY2025, so the recovery is recent.C (calculated)⁴⁵¹²¹⁴

Accounting to watch

1. Stock pay. This page uses two measures. Stock pay is the stock-based compensation expense, 30.2% of revenue in the last twelve months. The non-GAAP bridge and my model also charge employer payroll tax on stock transactions, which takes the charge to 32.3%. Free cash flow already bears that tax, because it is paid in cash, so free cash flow less the 30.2% expense is the like-for-like check on my model.C (calculated)⁵¹²¹⁴
2. Non-GAAP. The 12.4% non-GAAP operating margin adds back the full 32.3%, so it is not comparable with a peer’s GAAP margin.C (calculated)¹⁴
3. Interest income. $171M of interest income sits inside the last twelve months’ operating cash flow, 14% of free cash flow.C (calculated)⁵¹² My cash flows leave it out and add the cash once.
4. Timing. The fourth quarter produced $765M of FY2026’s $1,120M of free cash flow, and the July 2026 quarter $84M. In the six months to July 2026 receivables fell $601M and deferred revenue fell $807M, because customers renew in the fourth quarter.C (calculated)¹²¹⁶
5. Restated metrics. Snowflake adjusts customer counts and retention after the fact, as the first table below shows.⁵⁷¹²
6. Convertible notes. Snowflake owes $5.79B of zero-coupon notes after the September 2026 sale. The 2027 and 2029 notes, which convert at $157.50, were worth $4.45B at 31 July on $2.30B of principal; capped calls recover only part of the difference.¹²²³²⁴
7. A growth rate. The FY2026 10-K says revenue grew 29% in each of FY2024 to FY2026. The arithmetic for FY2024 is 35.9%, and I use it.C (calculated)⁵⁸

VIValuation

The price implies product revenue growth of 27.7% a year from year two to year ten. My range is $24 bear, $128 base and $299 bull, and the margin of safety price is $96. Values are dated Q4 2026 and use a 10.5% discount rate.¹¹² The shares fell 7.3% on 30 October 2026. Snowflake has not reported and filed no 8-K;⁵⁵ other software and cloud stocks fell with it, after Alphabet posted third-quarter free cash flow of −$9.1B and Amazon −$42.4B over twelve months.⁵²⁵³⁵⁴ The S&P 500 fell 1.8% the same day.⁵⁶ My cases stay as they were.

First, how this goes badly wrong

Before the numbers I asked how an owner at $306.08 loses money, and worked back.

• Growth fades sooner than the price needs. The price needs 27.7% a year from year two to year ten. Salesforce grew 23.1% and ServiceNow 22.5% from the year each passed $5B, and both were winners.⁴⁷⁴⁸ If Databricks and the clouds take share, I get my bear case, $24.
• Stock pay stays high. It took 30.2% of revenue in the last twelve months.¹² My base needs 12% by year ten. If it stops at 18% after year five, my year-ten value is about $173, not $239.
• AI does not carry the growth. The price needs AI workloads to become most of it, while open data formats lower lock-in.

Free cash flow is $1.2B on the company’s definition. After $1.64B of stock pay it was negative $447M, so stock pay is the whole gap between the headline and what an owner keeps.¹² Capital spending was only $58M, so standing still costs stock, not plant. Buybacks and share-withholding taxes took $1.39B, more than the free cash flow, and the share count still rose.¹²

I found no sourced market size and no measure of how often software companies at $5B of revenue follow each path, so my 25%, 50% and 25% weights are judgment. A beta of 1.32 gives a 10.5% discount rate and 2.46 gives 15.3%. So read $128 as a range: $140 at 10.0% and $100 at 12.0%.⁴⁴⁴⁵

What the price implies

At $306.08 the market value is $114.5B on the 374.1M shares my valuation uses: 352.8M outstanding, 8.4M net options and 12.9M from the 2027 and 2029 notes, which convert at $157.50. On basic shares it is $108.0B.C (calculated)¹¹² Enterprise value is $111.2B, 20.5 times the last twelve months’ revenue, 21.3 times product revenue and 93 times free cash flow.C (calculated)^1122324 Owner free cash flow is negative, so it has no yield, against 5.00% on the ten-year Treasury.⁴⁴

What the market is paying for

As I read it, the price asks for AI workloads to become most of Snowflake’s growth. So far the CEO says AI products supplied about half of the acceleration, CoCo has 9,100 accounts and CoWork 5,800, 43% of customers share data, and Snowflake has $200M partnerships with Anthropic and OpenAI and a $390M commitment to an AI service provider.^1214224243 To justify the price, AI workloads must carry most of a $64.1B product revenue line by July 2036 while Snowflake keeps the data layer as open formats lower lock-in. Three tests would show it: the non-GAAP product gross margin outlook back at 75% from 74%, net revenue retention above 130%, and CoCo adding at least 2,000 accounts a quarter.E (estimate)

The reverse DCF

Run my base model backwards, holding year 1 at $7.10B and the margins and stock pay of my base case. The price then needs product revenue to grow 27.7% a year from year 2 to year 10, to $64.1B, 12.3 times the last twelve months, and to $116.2B in year 15 as growth fades to my base case’s 3.0%; my base case reaches $41.6B.C (calculated) Snowflake’s product revenue grew 40.7% a year over the last four fiscal years, from a base about a fifth the size.C (calculated)⁵⁹ Salesforce grew 23.1% a year from FY2015, its first year above $5B, to FY2024, with acquisitions. ServiceNow grew 22.5% a year over the four years from 2021, its first year above $5B. Salesforce is the only peer with a nine-year record.C (calculated)⁴⁷⁴⁸

The price needs product revenue of $64.1B in ten years, 12.3 times today’s and 1.5 times what Salesforce reports now.

On my bull margins and bull year 1, the price needs 22.7% a year through year 10, which reaches $46.7B, under my bull case’s $50.4B. The difference comes after year 10: my bull case exits that year at 13.5% growth, and the price needs the faster pace to carry on into years 11 to 15, to $78.9B in year 15, not $72.2B. The price pays for growth after year ten more than for the first ten years.C (calculated)¹²¹⁴

No plausible margin closes the gap alone: it takes a non-GAAP margin 28 points higher, 61% in year 10, and with no stock pay in any year my base case is worth $216, 29% below the price. Counting stock pay at half its size gives $173.C (calculated)¹¹⁴

The discount rate

I use 10.5%, the capital asset pricing model at Snowflake’s 60-month beta of 1.32 to the S&P 500: the 5.00% ten-year Treasury yield plus 1.32 times Damodaran’s 4.20% implied equity risk premium for 1 October 2026.C (calculated)^3444546 His premium assumes a 5.29% risk-free rate; at that rate the model gives 10.8%. The 36-month beta, 2.46, gives 15.3%, and I give it no weight, because the swing from the 10 April 2026 close of $121.11 to $306.08, a rise of 2.5 times, sets it.C (calculated) The rate decides how far the price sits from my cases: at 10.0% my base case is worth $140 and my bull case $328, 7% above the price; at 10.5% they are worth $128 and $299.

Bear, base and bull

The model runs fifteen years. Revenue is product revenue times 1.043, the ratio of total to product revenue in the last twelve months. Growth is explicit for ten years, then fades in a straight line to a terminal rate of 2.5% in the bear case, 3.0% in the base case and 3.5% in the bull case. Owner operating income is revenue times the non-GAAP operating margin, less the stock pay charge including employer payroll tax. Cash tax is 23% on income after $7.3B of federal losses, which offset 80% of taxable income each year.⁵ Working capital brings in 20% of each year’s revenue increase, compared with 28% in the last twelve months.C (calculated)⁵¹² Capital spending equals depreciation. Cash flows are discounted to 31 July 2026 and rolled forward to Q4 2026.

Cash is pro forma for the September 2026 notes, which convert at $500.38 and $483.98, above every case, so I subtract them at principal.¹²²³²⁴ Shares are the 352.8M outstanding plus 8.4M net options, 361.2M. I do not count the 25.1M unvested restricted stock units, because the stock pay charge already pays for them; that values existing grants at their grant-date price, below today’s, so it leans toward a higher value. The 2027 and 2029 notes ($2.04B, converting at $157.50) are repaid at par in the bear and base cases and convert into 12.9M shares in the bull case, which uses 374.1M shares.C (calculated)¹²²⁴ The cases charge no acquisitions beyond the one point Observe adds to the guide; the table of other treatments shows the cost.

The last twelve months are the check on year 1: free cash flow less the stock-based pay expense was negative $447M, or negative $618M without interest income. Free cash flow already bears the payroll tax on stock, so this is the like-for-like base. My base year 1 is negative $495M, an improvement of $123M on the base without interest, on $2.0B of added revenue.C (calculated)⁵¹²

My base case starts from the guide. The September guide implies $3.24B of product revenue for the second half of FY2027, and I add 3%, below the 3.8% average by which Snowflake beat the top of its range in the last four quarters. For the first half of FY2028 I assume growth of 33%, down from 36.8% in the July quarter. Year 1 is $7.10B, up 36.2%.E (estimate)^1415161718

The step from 40% growth to 33% is the sharpest slowdown in the path. About a point is Observe; the rest is a smaller sequential add than the record second half. The 40% is already a beat on the 36% the guide implies, and smoothing the step to 36% would make year 1 up 37.9% and the base value $129. The year 1 non-GAAP margin of 16% is above the guide, which implies 15.2% for the second half; it needs 16.7% in the first half of FY2028, 3.0 points above the first half of FY2027, compared with the 3.7 points gained a year earlier. The July quarter’s stock pay charge was 29.5% of revenue, against 28% in my year 1.E (estimate)¹⁴¹⁶

Growth then falls from 30% in year 2 to 9.5% in year 10, and product revenue reaches $31.7B. That is 6.1 times today’s, and 2.7 times in the first four years, versus 2.5 times for Salesforce, 2.3 for ServiceNow and 1.9 for Workday.C (calculated)⁴⁷⁴⁸⁴⁹ The non-GAAP operating margin rises from 16% to 33% and stock pay falls from 28% to 12% of revenue, an owner margin of 21%. That is about Salesforce’s GAAP margin today, 20.1%, at a lower revenue. Owner free cash flow turns positive in year 3 and reaches $5.9B in year 10.E (estimate)⁴⁷

The 10-K gives no market size, only that the opportunity is large, and I found no sourced estimate I would defend. The base case rests on the revenue comparison with Salesforce, ServiceNow and Workday, not on a share of a market.⁵

The bear case is Workday’s path. Databricks passes Snowflake’s core warehouse by FY2029 and the clouds take share. Year 1 is $6.84B, up 31.2%, with growth fading to 5% by year 10, when product revenue is $16.0B, 3.1 times today’s; in four years it is 2.0 times, close to Workday’s 1.9. The non-GAAP margin settles at 24% and stock pay at 15%, an owner margin of 9%, between Workday’s 7.5% and ServiceNow’s 13.7%.E (estimate)⁴⁸⁴⁹ The bear is worth $24 and reaches the price only at a discount rate of 3.4%, below the 5.00% Treasury yield.C (calculated)⁴⁴ It takes both halves: bear growth on base margins is worth $61, and base growth on bear margins $55. Owner free cash flow falls from $1.70B in year 12 to $1.45B in year 14 because the $7.3B of losses run out in year 13 and cash tax rises from $75M to $400M.C (calculated)

The bear is a value on my method, not a floor the market would set. It values year 15 at 12.8 times owner free cash flow. Workday has an enterprise value of $50.2B, 4.9 times its last twelve months’ revenue and 42 times its owner free cash flow, with revenue growing 13%.C (calculated)⁴⁹⁵⁰⁵¹ Each extra turn on the year 15 multiple adds $0.93 to the bear value.C (calculated)

The bull case is Snowflake as the data layer for AI. Year 1 is $7.43B, up 42.5%, the guide beaten by 5.2% as in the last two quarters, then growth of 38% in year 2, 26% in year 5 and 13.5% in year 10, when product revenue is $50.4B. The non-GAAP margin reaches 38% and stock pay 9%, an owner margin of 29%, nine points above Salesforce’s today.E (estimate)¹⁴⁴⁷

Weighted 25%, 50% and 25%, the three cases give $145 a share.E (estimate)¹²¹⁴ The weights are my judgment; I found no sourced measure of how often software companies at $5B of revenue follow each path. The price sits above all three cases, so the conclusion does not depend on them, and the margin of safety uses the base value because the weights cannot be measured.

Base growth that ends year 10 at 12% is worth $144 on base margins, and at 20% $210.E (estimate)¹²¹⁴

The comparison with Salesforce, ServiceNow and Workday rests on three companies that succeeded, and survivors flatter any reference class. I use them for the shape of the path, not as a forecast. The part I cannot take from them is stock pay: at $5B of revenue Snowflake paid 30% in stock, where they paid 10% to 21%.C (calculated)⁴⁷⁴⁸⁴⁹

Of my $128 base value, $4 is cash after the old notes are repaid at par, and 50% arrives after year 15.E (estimate)¹²²⁴ One point on the discount rate moves my base value by 15% to 20%.

What closes the gap

Swap one bull input at a time into my base case and growth does most of the work. Bull growth alone lifts the base from $128 to $207, the bull non-GAAP margin alone to $159, the bull stock pay alone to $147 and bull growth after year 15 alone to $134.E (estimate) Taken together, in the order the chart shows, they reach the $299 of the bull case. The price then needs a discount rate of 10.37% instead of 10.5%.¹

No row of the base case reaches the price; at 8.0% it is worth $211. The bull case is worth $405 at 9.0% and $328 at 10.0%.⁴⁴⁴⁵

The margin of safety

• Price$306.08Close, Q4 2026 [1]
• Base value$128My estimate E (estimate)
• Margin of safety below$9625% under my base value C (calculated)

My rule is 25% below the base value for a business with net cash and no customer above a tenth of revenue, 30% when one cycle or two customers drive results, and 35% when fixed charges such as leases sit ahead of the shares. Snowflake fits the first tier: it has net cash after its convertible notes, $7.1B of cash and $5.8B of notes, and no customer above 10% of revenue.⁵¹² That gives $96, and 25% is the rule’s output. Consumption risk is in the cases, not in the discount: the bear case holds growth to 5% by year 10. The price is 219% above that line, so the stock would have to fall 69% to reach it; at 10.0% the line would be $105.C (calculated) The lowest close of the past year, $121.11 on 10 April 2026, was 26% above it and 5% below my base value; the stock has risen 153% since.C (calculated)³

Other uses of the same dollar

The ten-year Treasury pays 5.00%.⁴⁴ My base returns about 0% a year from this price, so a bond pays more. The closest comparison on this site is ServiceNow, large subscription software with its own stock pay question. Its base returns 11.4% a year, at a price 11% below its base value.

Two questions are separate. The business is growing fast: product revenue rose 37% in the July quarter and net revenue retention is 126%.¹² The price already asks for close to my bull case. This looks like a good business at a price with little room for disappointment. The bull case needs growth of 13.5% in year ten, a 38% margin and stock pay of 9%. At a 60% chance each, all three come to 22% E (estimate). They are linked and 60% is my guess.

Where this could be in 5 and 10 years

Value in year 5 is the present value, at July 2031, of the owner free cash flow after that date and of everything after year 15, at 10.5%. Year 10 is the same at July 2036. It is per share, on today’s share count, in that year’s dollars. I leave out today’s cash and notes.E (estimate)

For the base to be right at year 5, product revenue has to reach $17.2B, 3.3 times today’s,C (calculated) with a non-GAAP margin of 26.5% and stock pay at 18% of revenue. That is an owner margin of 8.5%. At year 10 it needs $31.7B, a 33% margin and 12% stock pay, an owner margin of 21%. The base is then worth about $239 a share, and $72 of cash has been paid out on the way.

The bear and the bull differ on growth and on owner margin. In year 10 the bear has growth of 5% and an owner margin of 9%. The bull has 13.5% and 29%. Year-10 product revenue is $16.0B against $50.4B. All three use 10.5%, and one point on it moves the base value by 15% to 20%.

The most likely way the base is wrong is stock pay. It has to fall from 32% of revenue today to 12% while revenue grows six times. If it stops at 18% after year 5, my year-10 value is about $173, not $239.C (calculated) At $306.08 the base returns about 0% a year to year 10. The price is 2.4 times my base value today.C (calculated)

These are estimates of value, not forecasts of the share price. The market price can sit above or below value for years.

VIIThe case for the price

What most people miss

Management has put a date on GAAP profit, the kind that counts stock pay as a cost: the quarter to January 2028. It set that in June and repeated it in September.⁵⁷²² My base case is slower. Its owner margin, the margin after stock pay, is −6% in the year to July 2028 and turns positive only in year 4 E (estimate).

There is also room inside customers Snowflake already has. In FY2026 its top 25 averaged $34.1M of product revenue, and its average Forbes Global 2000 customer paid $2.4M.⁵⁷ That is a 14-times gap C (calculated).

The speed-up also looks broad. The CEO says AI-native companies are still a small part of revenue. Projects deployed rose 89% in a year.²²

The bull case

Management makes the fullest case. AI agents are only as good as the data they can read, and much of that data already sits in Snowflake. On 2 September Sridhar Ramaswamy called AI “a structural multiplier for our business”.²² Christian Kleinerman said many customers who committed to one model company later wanted another, and Snowflake lets them switch.²²

Morgan Stanley raised its price target to $470 from $300 on 3 September, per Schaeffer’s. I have not seen the note.⁶⁰

Brad Gerstner has said the most. Altimeter led an early $45M round. On IPO day in 2020 he told CNBC the cloud “fundamentally rearchitects where all data sits”.⁶¹ In March 2022 he said that if he couldn’t beat Snowflake’s return over a decade, he’d “put all my money in Snowflake and go surf with my kids”.⁶²

Against reality

The price is about my bull case, $306.08 against $299. So if Snowflake keeps beating its guide by about 5%, grows 42% in the first half of FY2028, reaches $50B of product revenue by July 2036 and takes stock pay down to 9% of revenue, the price is possible E (estimate).

Today’s numbers are some way short. Growth is 36.8%, not 42%. The stock pay charge was 29.5% of revenue in the July quarter. Net revenue retention is 126%, and this story needs more than 130%.¹²¹⁴ GAAP profit is the one on pace. The operating margin was −17.0% in the July quarter, up 12.7 points in a year, and the target needs 17 more points in six quarters C (calculated).¹⁴

How long growth lasts is the constraint. If Snowflake hits the GAAP date through stock pay alone, my base rises from $128 to about $134 E (estimate). Bull growth alone takes it to $207 E (estimate).

The 13Fs give counts, not reasons. Altimeter reported 32,221,080 shares at 31 March 2021 and 1,926,250 at 30 June 2026, 5.0% of its listed value. That is 94% fewer C (calculated), though a 13F can’t show whether shares went to the market or out to fund investors.⁶³⁵⁸ Berkshire paid $120 a share around the 2020 IPO and reported 6,125,376 shares up to 31 March 2024.⁶⁴⁶⁵ Its filings from 30 June 2024 on list none.⁶⁶ Gavin Baker’s Atreides reported 756,767 shares at the end of 2024 and none at 30 June 2026.⁶⁷⁶⁸ ARK reported 269,539 at 30 June, after none at 31 March.⁵⁹ ARKK’s file for 30 October lists none. The latest ARKW file I could open, from 6 October, lists 63,641.⁶⁹⁷⁰

What they have to hit

Six numbers, each published on a set schedule. Where my signposts cover the same number, the thresholds match.

Snowflake has not dated the Q3 report; last year’s came on 3 December.¹⁷

Where I come down

The GAAP date moved me most, because it makes my stock pay path look slow. It does not change my $24, $128 and $299, since a date is not a result and stock pay is the smaller lever. A July 2027 quarter above 40%, with RPO growth caught up and GAAP profit on time, would raise my values.

VIIIRisks, and what would change my view

What would move me

The first test is the Q3 FY2027 report. The documents I read give no date; the Q3 FY2026 release came on 3 December 2025.¹⁷ Snowflake guided product revenue of $1,588M to $1,593M, up 37% to 38%, and a non-GAAP operating margin of 15.5%.¹⁴

Risks

The risk Snowflake names that I weight most is its own consumption model:

…customers may continue to optimize consumption, rationalize budgets, and prioritize cash flow management.

1. Consumption. Product revenue follows usage, so a pause in usage reaches revenue in the same quarter, and the 10-K says better compression or more efficient computing can lower revenue if new workloads do not replace it.⁵ Receivables and deferred revenue each moved by $600M to $800M in the half-year to July 2026.¹²
2. AI economics. AI workloads carry a lower contribution margin today, cost of product revenue grew faster than product revenue in the July quarter, and the gross margin outlook came down by a point.C (calculated)¹²¹⁴²² If the mix keeps shifting, the non-GAAP margin in my base case, 33% in year 10, is the number that gives way first.
3. The clouds. AWS hosts a substantial majority of the business and competes with it. Snowflake has $6.0B of cloud spending to commit to over five years, with shortfalls payable; the yearly minimum of $0.9B to $1.25B is a floor that matters only if usage stalls. A cloud could bundle a competing product or price against Snowflake.⁵¹³
4. Databricks. It is growing at over 80% from over $7B of company-defined run-rate, and its warehouse product at over 100% from over $1.5B.³⁷ The 10-K does not name it. If its Lakehouse product passes Snowflake’s core, my bear case is closer than my base.
5. Stock pay, dilution and governance. 25.1M restricted stock units and 9.3M options are outstanding, together 9.7% of shares.C (calculated) Shares outstanding rose 2.5% in six months. Two say-on-pay votes failed, the board is divided into three classes, and the fund of the lead independent director took $208M in stock in the Observe deal.^12262829
6. Security and litigation. Attackers reached customer accounts that lacked multi-factor authentication in May 2024, and the class actions are in discovery. Two securities class actions name the company and two former executives, a copyright class action covers the training of its language model, and a labour board matter has a possible loss of up to $25M.¹²
7. The multiple. At 20.5 times revenue the price already assumes much of the good outcome. The shares rose 2.5 times from April to October, fell 7.3% on 30 October 2026 with no Snowflake filing, and have moved by −11.4%, +36.5% and +16.6% in a day on results.C (calculated)³