Reference price: $376.11 (September 3, 2026, 10:04 AM, with the stock up 23% on the session after results reported the night before). Numbers come from the earnings release and seven years of official annual accounts; price and multiples move in real time.
What this company actually does
It is a cloud data warehouse rented by usage. A large company has its data scattered across a hundred places; Snowflake gathers it in one, organizes it, and charges for every query and every gigabyte stored. No flat fee: the more customers use it, the more they pay. And it now sells the tools for AI to work on that data. $5.4B of trailing revenue, 829 of the world's 2,000 largest companies as customers, $131B market cap.
The thing to understand first: this company has never made money under official accounting. Not one single year since it has existed. What gets published as "profit" is something else — and that is what everything below is about.
The good — and there's more than it seems
It is accelerating, not slowing — three quarters in a row. Product revenue growth: +26% a year ago, +33% last quarter, +37% this one. At this size the normal thing is to slow down; this does the opposite. Full-year guidance raised from +31% to +36%.
Customers spend more every year than they did: a customer that spent 100 last year spends 126 this year, before counting new ones. The product hooks. 828 customers pay over $1M a year (+27%). $9B of work already contracted against $5.4B of annual revenue — and that backlog grows 30% a year.
It is losing less, at a good pace: revenue +35%, operating loss −23% (from losing $30 per $100 of revenue to $17). Opex up 17% with revenue up 35%. More cash than debt ($4.3B against $2.3B of 0% convertible bonds). And 45 analysts sign the next two years — among the best coverage we've seen.
The bad
Problem 1 — and it is the entire analysis: the profit that jumped 23% today is the one you get by not counting what it pays its people. This company publishes two profits. This quarter: what the company says, +$235M (+$0.62/share); what official accounting says, −$192M (−$0.55/share). What sits in between? One thing, almost entirely: $456M of salaries paid in stock. Out of every $100 of revenue, $29 goes to employees in company paper — and that paper is not counted as an expense.
Translated into something physical: imagine a restaurant billing $100 that claims to earn $15 — but it doesn't pay the waiters in cash; it hands them $29 worth of shares in the restaurant. Count those shares as what they are — a salary — and the restaurant loses $17 per $100 billed. And it isn't one quarter: it's six years. Never, in any fiscal year, has it earned a cent under official accounting (2021: −$2.26; 2025: −$3.95; H1 2026: −$1.40). And the loss per share isn't shrinking — it grows, because every year it pays more in stock.
Problem 2: the "free cash flow" it advertises has the same trick. Published: $1.20B of trailing free cash. Paid to staff in stock: $1.64B. Free cash counting that salary: −$447M. The free cash exists because payroll doesn't leave the till — it leaves the shareholder's pocket, as new shares that dilute theirs.
Problem 3: your slice shrinks 4% a year, and more is coming. Share count 335M → 349M in a year (+4.2%), with 382M already promised (+9.4% more pending conversion).
Problem 4: it costs 179 years of a profit that accounting says doesn't exist. On what the company says it will earn this year: 179 years. On 2027: 130. On 2028: 92. On official accounting: can't be computed — it loses money. For scale: a good company costs 15–25 years of what it earns.
Problem 5: up 95% in seven months (from $192.70 at last fiscal year-end to $376.11), and today's 23% is just the last leg. One warning that is measured, not opinion: the companies promised the most growth delivered double what was promised… and were the worst performers — not because growth didn't arrive, but because it arrived already paid for. A +73% expected for this year is in that group.
Problems 6–8: nearly all book equity is acquisitions (of $2.15B declared, $84M is real things — and the acquisitions line grew from $1.19B to $1.64B in six months: it is buying AI capabilities rather than building them); part of this quarter's improvement didn't come from the business ($40M of the $106M came from marking up investment stakes — that's the market rising, not the business); and for a leveraged investor, the published "free cash" yields 0.91% a year (negative counting stock salaries) against 5–6% money — more than five negative points a year before the price moves.
The sum of it
You can't run the usual "cheap or expensive" math here, and that's not a limitation — it's the finding. A multiple divides price by profit, and under official accounting the profit is negative. There is nothing to put underneath.
What you can do is run the math on the company's own number: to earn 2027's expected $2.90, it would need to keep ~$14 of every $100 billed; the company guides 14.5 and just delivered 15. Achievable on the company's number? Yes — it's exactly what it just did. And on official accounting, counting stock salaries? It would still lose $17 of every $100. That's the whole thing: what's expected of this company comes true if — and only if — you keep accepting that paying staff in stock isn't an expense.
Only one of two things can happen: either stock compensation falls from 29% of revenue to something normal — 5 or 10 — and official profit appears while the "company profit" falls with it; or it stays at 29%, official profit never appears, and every shareholder's slice keeps shrinking 4% a year. Not both. And today's price is paying for the company's profit to be true WITHOUT stock comp coming down.
What must be acknowledged without discounting it: the business is real, accelerating, customers spend more every year, $9B is contracted, losses are shrinking. If the only filter were "is this a good product that grows?", it would pass.
The chart says one thing, loudly
Every year since its IPO, without a single exception, the price has traded above what even the company's own generous profit would justify. Today it trades 12 times above: this year's profit would justify $31.50; the price is $376.11. And the expected-earnings line never catches the price: 2029's expectation — the far year, signed by two analysts — would justify $84.45, still 4.5 times below today's price. Translated: even if everything promised comes true four years running, in 2029 this still costs 67 years of profit. And that's the profit that doesn't count payroll.
What we would do
Not buy it. And today less than yesterday. Said precisely: it isn't that the business is bad. It's that the profit the price is built on doesn't exist under official accounting, and what does exist — the 29% of revenue paid in stock — comes out of the shareholder's pocket.
| Price | What you would be paying (on the company's own generous profit figure) |
|---|---|
| Below $87 | Very attractive. |
| $87–131 | Attractive. |
| $131–189 | Fair value, being generous. |
| Above $189 ← it is here, at $376 | Expensive. Double the ceiling. |
And that band is already generous: the ceiling is 65 years of the profit expected two fiscal years out. Even so, it trades at double that ceiling today. The entry price is no fantasy: seven months ago this traded at $192.70, and two days before these results at $305.84. For a leveraged portfolio there's an added reason: holding this costs more than five points a year — in a company priced at 179 years of a profit that doesn't exist, every passing day works against you twice.
The three things we would watch
1. Stock pay per $100 of revenue. THE number for this company. Down from 39 to 29 in a year. If it keeps falling toward 20 over two more quarters, official profit appears and the whole thesis changes. If it stays at 29, there is no company underneath the paper.
2. Whether growth keeps accelerating. +26% → +33% → +37%. While it accelerates, an absurd price can hold. The first quarter below 30%, a 179x multiple has nothing to hold on to.
3. The share count. 349M now, 382M guided. Every quarter that number rises 1%, each shareholder's slice is worth 1% less even if the company does well.
And the standing warning, concretely here: always read the line that starts with "net loss", not the one that starts with "adjusted profit". Same company; they say the opposite.
In one sentence
It has grown 35% a year for six years and has never once made money under official accounting: the profit that jumped twenty-three percent today is what's left if you don't count the $1.6 billion it pays its people in stock.