Snowflake Inc. SNOW

Key data

The business

Snowflake sells a place to put company data and a way to ask it questions. Large enterprises used to buy, size, and maintain those warehouses themselves. Snowflake separated storage from computing, ran both in the cloud, and charges by consumption rather than by licence, so a customer stores everything and pays only when a query runs. Product revenue reached $4.47 billion in the year ended January 2026, up 29%.

Because the meter runs on usage, revenue moves with what customers actually do rather than what they signed. That is also why AI lands here directly: training and querying models against corporate data runs on exactly this substrate. What changed is the shape of the curve. Growth ran 35.9%, 29.2%, then 29.2% across three fiscal years, and the April quarter printed 33% on $1.39 billion with net revenue retention at 126%.

Business read. A consumption meter on corporate data, reaccelerating at $4.7 billion of scale.

Things you might not know

Remaining performance obligations were $9.772 billion in the January quarter and $9.205 billion three months later, down 5.8% sequentially while up 37.7% year over year. RPO tracks contract signings rather than usage, so it swings with renewal timing and says less about the meter than the headline growth does.

On July 15, 2026 the board granted CEO Sridhar Ramaswamy 1,000,000 performance stock units in five tranches requiring 90 day average prices of $324, $375, $427, $479, and $531, with service dates in September 2029 and 2030. The first hurdle sits below the current price.

In the January quarter Snowflake reported a GAAP operating margin of negative 25% against a non-GAAP operating margin of positive 11%. That 36 point gap is a single accounting item, stock compensation, expressed as margin instead of dollars, and it is the same item the valuation rests on.

Fundamentals

Every consumption line accelerated in the April quarter, with product revenue up 34% and the largest customer cohort up 29% to 779 accounts. The only sequential decline is RPO, off 5.8% from a seasonal January high, which is a bookings artifact rather than a usage one. Retention ticking to 126% means existing customers are still spending more each year.

Valuation

Nearest comparables by business model: Databricks (private), Datadog, MongoDB, Confluent, Elastic.

No peer set was priced in the source report, and market multiples for comparable companies do not appear in filings or investor relations materials, so the peer column stays empty rather than estimated. On its own numbers, no year in visible consensus puts this stock below 55.8 times earnings.

Management

Compensation

The linchpins

Win big if

Consumption compounds near 30% while stock compensation keeps falling as a share of revenue, which is the only route by which $1.12 billion of free cash flow means what a buyer at 98.6 times is paying for. First observable proof is the September 2 print clearing $1.5 billion of revenue with net revenue retention holding above 125%. The confirming signal is the compensation ratio walking from 34.2% toward the low twenties within two years while growth stays above 25%.

Surprised down if

Growth decelerates below 27%, or gross margin compresses while revenue holds, which would mean workloads are being won by reselling cloud compute at a thinner spread rather than by owning the data. First observable proof is a September 2 revenue miss against the $1.5 billion consensus, or retention slipping under 125%. The confirming signal is diluted share count passing 345 million, meaning $873.5 million of repurchases is losing ground faster than before.

Last word

Employees are paid in the same stock that the buyer at 98.6 times free cash flow is holding.