Why the Street Is Wrong About Snowflake's Consumption Slowdown

It is useful to state the bear case precisely because most bear cases are looser than they appear when enumerated.

The bear argument is three-part. First, product revenue growth has decelerated from 60%+ to the mid-20s. Second, net revenue retention has trended from 165% down to roughly 126%. Third, GAAP operating loss has widened from $1.1 billion to $1.44 billion even as revenue has grown, implying that operating leverage is not coming through and that the long-term margin target is a fiction.

Each of these data points is accurate. The conclusion the bears draw is that Snowflake is a low-quality software business whose economic model is fundamentally impaired. The sell-side has followed this logic, with several shops moving to sub-consensus 2027 revenue estimates and broadly cautious commentary.

We think the bears are working the wrong income statement. Here is what the cash flow statement actually shows.

Point one on consumption deceleration. Consumption growth from 60% to 25% is a function of denominator effect plus enterprise maturity, not economic impairment. The customer base that drove the 60% growth period was a cohort of early-adopter tech and fintech customers whose data workloads grew exponentially off a small base. Today's customer base is much larger, more diversified across industries, and more deeply embedded in mission-critical workflows. The growth rate is lower because the base is larger. This is not new; every successful enterprise software company has gone through this transition. The ServiceNow analogue: growth decelerated from 60% to 25% between 2015 and 2018, and the stock re-rated higher, not lower, over that period because the FCF trajectory supported it.

Point two on net revenue retention. The NRR trend from 165% to 126% has been the single most cited bear data point. What the bears underweight is that NRR decline is largely a denominator effect as the numerator (existing customer consumption) scales from a much larger base. 126% is still in the top decile for enterprise software and is consistent with a healthy consumption expansion business. The reference point here is Datadog, which has held NRR above 110% for most of its public life and has compounded revenue at 30%+ CAGR regardless. 126% at Snowflake's current scale is structurally fine.

Point three on widening operating loss. The GAAP operating loss has widened from $1.1 billion to $1.44 billion. That is unambiguously true. It is also almost entirely SBC. SBC ran at approximately $1.5 billion in FY26, roughly 32% of revenue and a step-up from the 28% prior-year level. If you strip out SBC to look at the cash operating profile, the business generated $1.12 billion of FCF on $4.68 billion of revenue, a 24% FCF margin. That is a 400 basis point FCF margin expansion over three years. The non-SBC operating economics are improving, not deteriorating.

The bear case, properly enumerated, is a dilution argument, not a profitability argument. The SBC run-rate is high. Diluted share count has grown roughly 2-3% annually. That is the real cost to shareholders. At current multiple compression, however, the dilution is partially absorbed by the lower share price, and the pace of SBC as a percentage of revenue has started to decline (it was 38% three years ago and is now 32%). The trend is in the right direction.

The bearish case has been in place for six months. The data hasn't changed the calculus.

Here are the data points that the bearish sell-side commentary systematically underweights.

First, remaining performance obligations (RPO). RPO closed FY26 at approximately $6.9 billion, up 34% YoY. RPO is the single best leading indicator for enterprise software revenue because it reflects contract value that has been booked but not yet recognised. A 34% RPO growth rate is inconsistent with a business in fundamental trouble. It is consistent with forward revenue growth of 25-28% in FY27.

Second, customer count at the $1 million+ annual consumption threshold. This cohort grew approximately 22% YoY in FY26. The enterprise adoption curve is still producing new large accounts at a healthy pace. This metric has historically been the leading indicator for revenue reacceleration within enterprise software.

Third, Cortex AI consumption. Snowflake's native AI platform, which allows customers to run large language models against their own Snowflake-resident data, is in early commercial traction. FY26 was the first full year of meaningful Cortex revenue contribution. The attach rate to existing Snowflake customers is running higher than originally modelled and provides a second growth vector beyond core data warehousing consumption. None of the current Street models adequately incorporate Cortex uplift.

Fourth, the cash position. $3.9 billion of cash and equivalents, no debt. The balance sheet is not a constraint on execution.

Fifth, capital efficiency. FCF-to-revenue of 24% for FY26 is consistent with a software business in healthy middle maturity. Adjusting for the fact that working capital benefits from deferred revenue in growth years, the underlying cash profitability is strong.