Snowflake lost $348.5m in the year to January 2020, which is more than the $264.7m it took in revenue.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020 On Wednesday its shares closed at $253.93, up 111% from the $120 at which they were sold the night before, and the company was worth about $70.4bn.2CNBCSnowflake more than doubles in market debut, largest ever software IPO16 Sept 2020 Nobody was forced to buy.
The listing was the biggest software flotation on record. Snowflake sold 28m shares at $120, above a range that had been raised from $75-85 to $100-110, and took in $3.36bn before fees.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020,3Renaissance CapitalLet it SNOW: Data unicorn Snowflake prices $3.4 billion IPO above the range15 Sept 2020 Salesforce Ventures and Berkshire Hathaway each bought a further $250m of stock at the same price in private placements.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020 The firm had been valued at $12.4bn in a funding round in February.4The Irish TimesSnowflake completes largest-ever software IPO16 Sept 2020 Seven months later the market put the figure at nearly six times that.
Spending more, a little less quickly
Snowflake, net revenue retention rate, %
| Net revenue retention | |
|---|---|
| Jan 2019 | 180 |
| Jul 2019 | 223 |
| Jan 2020 | 169 |
| Jul 2020 | 158 |
Source: Snowflake prospectus (Form 424B4)1
We think buyers were not paying for earnings, of which there are none, but for a single ratio. Snowflake sells data warehousing in the cloud and charges by use. Its “net revenue retention rate” compares what customers spent in the latest 12 months with what the same customers spent in the 12 before. It stood at 158% on July 31st.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020 Revenue in the first half of its fiscal year rose 133%, to $242m, and the firm itself credits that growth mainly to existing customers consuming more.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020 The price is an opinion about how long that continues.
Put a number on the opinion. Revenue over the past 12 months was about $403m, by our arithmetic from the prospectus.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020 Suppose the market eventually pays a generous 20 times sales; $70.4bn then requires about $3.5bn of revenue, nearly nine times today’s. If existing customers alone supplied the growth, 158% retention would get there in five years. At 130% it takes more than eight. The ratio behaves like a half-life, not a constant, and it has already been decaying: it was 223% a year earlier.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020
The strongest objection is that consumption pricing is a virtue, not a flaw. Customers pay only for what they use, so there is no shelfware to cancel, and contracted future revenue, at $688.2m on July 31st, was three times its level of a year before.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020 A rising retention figure on this model means people are genuinely doing more with the product. And, as the chief executive, Frank Slootman, said after the debut, “a stock is worth exactly what somebody wants to pay for it”.2CNBCSnowflake more than doubles in market debut, largest ever software IPO16 Sept 2020
All of that is fair, and none of it answers the point. The same pricing that lets revenue climb without a salesman’s signature lets it fall without one. The prospectus says plainly that revenue is recognised on consumption, “which is inherently variable at our customers’ discretion”.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020 A subscription firm sees a cancellation coming; a consumption firm sees a smaller bill. Customers who have grown used to spending freely can tune their queries just as freely, and a ratio of 169% or 158% has not been tested by a downturn. Berkshire’s cheque, for its part, was written at $120, not $254.
What, then, must a buyer at $254 believe? That retention stays above 130% for the better part of a decade; that the customer base, 3,117 now against 1,547 a year ago, keeps doubling; and that losses will turn into margin before the patience of the market runs out.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020 Each is plausible. All three at once is a harder sale. The firm burned $176.6m of cash in operations last fiscal year, so it is not short of reasons to want the market’s goodwill.1SEC EDGARSnowflake Inc. final prospectus, Form 424B416 Sept 2020
None of this makes the stock a fraud or the company a poor one. Snowflake looks a good business, as the customer count and the growth show. We think it is a good business at a price that leaves no room for ordinary disappointment. Investors who bought on Wednesday have paid for the next ten years in advance, and the ratio they are relying on has been falling.
Mr Slootman is right that a stock is worth what somebody will pay. He did not say for how long.