Snowflake beat on almost everything. One number went the other way

Snowflake beat expectations on revenue, on profit and on guidance, and its shares rose more than 20% after hours. One number in the release went the other way, and it is the one that looks furthest into the future.

Remaining performance obligations, the contracted revenue Snowflake has booked but not yet recognised, came in at $9.00bn. Brody Ford reported for Bloomberg that analysts had expected $9.37bn.

The numbers that beat

Revenue for the quarter ending 31 July reached $1.55bn, up 35% year on year, against a consensus of $1.48bn. Product revenue, which is 96% of the total, rose 37% to $1.49bn against an expected $1.42bn.

Adjusted earnings came in at 62 cents a share. Analysts had forecast 45 cents.

One point worth getting right, because at least one outlet reported it the other way round. Product revenue excludes professional services. The release says so, and the arithmetic confirms it: $1,491.9m of product revenue plus $54.9m of services gives the $1,546.8m total.

The company raised full-year product revenue guidance to $6.07bn, or 36% growth, up from the $5.84bn and 31% it guided in May. For the current quarter it expects $1.588bn to $1.593bn, comfortably above the roughly $1.50bn consensus.

Net revenue retention was 126%. Snowflake added 692 net new customers and now counts 828 spending more than $1m a year.

The number that missed

Remaining performance obligations grew 30% year on year, which sounds healthy in isolation. Against what analysts modelled it is a shortfall of roughly $370m.

Snowflake’s own release explains why the figure is awkward to read. It says the measure “is not necessarily indicative of future product revenue growth because it does not account for the timing of customers’ consumption”, and lists renewals, contract length and seasonality among the things that move it.

Most coverage on Wednesday evening led on the beat and the share price. Bloomberg was the outlet that put the shortfall in print.

The profit everyone quoted is not the only profit

Snowflake reported non-GAAP operating income of $237.0m, a 15.3% margin, and non-GAAP net income of $235.3m.

On a GAAP basis it lost money. Operating loss was $263.0m and net loss was $191.7m, or 55 cents a share. That is narrower than the $297.9m loss a year earlier, but it is still a loss.

The bridge between the two is mostly one line. Stock-based compensation-related charges came to $456.4m for the quarter, which is larger than the non-GAAP operating income the results are being celebrated for.

Snowflake lists reducing stock compensation as a share of revenue, and achieving GAAP profitability, among its own risk factors.

None of that is unusual for enterprise software, and the loss is shrinking. It is worth stating plainly because the headline figures being quoted are the adjusted ones.

What is driving the growth

Chief executive Sridhar Ramaswamy told Reuters that AI products accounted for “approximately half of the acceleration that we are seeing”.

The coding assistant CoCo passed 9,100 accounts, adding more than 2,000 in the quarter. CoWork, which answers questions about a company’s own data, reached 5,800 accounts. Snowflake counts these by averaging the last four weeks of the quarter.

Finance chief Brian Robins called it the third consecutive quarter in which product revenue growth accelerated, and said margin expanded at the same time.

Running AI is not free. Non-GAAP product gross margin was 74.7%, down from about 76% a year earlier on the company’s own figures. In May, Snowflake signed a five-year, $6bn agreement with Amazon Web Services for Graviton processors, which Ramaswamy said locks in what it pays for storage and compute.

The rival growing faster

Databricks is the comparison Snowflake cannot avoid. We reported last month that it closed a $5bn round at a $190bn valuation.

Databricks has since said it passed $7bn in revenue run-rate, growing more than 80% year on year in its own second quarter. That is more than double Snowflake’s 37%, from a smaller base, and Databricks is private and not obliged to publish audited results.

The wider week

This lands in an earnings run where AI is the explanation for everything. Nvidia reported $96.2bn for its quarter last week, up 106%.

Disclosure is shifting too. We reported on Wednesday that Microsoft will publish an Azure revenue figure each quarter for the first time, while narrowing what counts as Azure.

And the labs themselves are being measured against each other. Anthropic’s revenue passed OpenAI’s for the first time last quarter.

Snowflake has had a harder AI news cycle than this one. In August, a security vendor attributed a Snowflake flaw to AI-written code, a claim GitHub disputed.

What to watch

The stock closed at $305.84 before the results and was up 39% for the year to that point. CNBC noted that if Thursday’s open holds the after-hours move, it would be the fourth largest one-day jump since the 2020 listing.

The question the guidance raises is whether consumption holds. Snowflake bills for what customers actually use rather than for seats, so revenue follows usage rather than contracts, which cuts both ways.

Snowflake now has more than 14,500 customers, and 829 of them are on the Forbes Global 2000 list. Its next scheduled disclosure is the Form 10-Q for the quarter just ended.