Snowflake put a 22% pop on the record. The tape shows what it really paid for.

After the close on September 2, Snowflake printed its quarter and shares jumped roughly 22% — from a $306.19 regular-session close to about $374 in after-hours trading. The headlines reach for "AI winner." A tape historian reaches for the dates. Q2 fiscal 2027, the quarter that ended July 31, delivered a beat-and-raise, and the pop is real. The record, though, is not a re-rating from a depressed base. It is an acceleration layered on top of a run that had already been going for months. That distinction is the whole investment story.

Start with the numbers, each tied to its basis. Total revenue was $1.55 billion, up 35% from a year ago and ahead of the roughly $1.48 billion analysts expected. Product revenue — the engine — rose 37% to $1.49 billion, marking a third consecutive quarter of accelerating growth. Management raised full-year product revenue guidance to $6.07 billion from $5.84 billion, now implying about 36% growth, and lifted the fiscal-year adjusted operating margin forecast to 14.5% from 13.5%. Adjusted earnings came in at 62 cents a share against 45 cents expected.

That last "earnings" needs a qualifier a first-time reader can catch. The 62 cents is non-GAAP/adjusted profit. On a reported basis Snowflake still lost money in the quarter — a net loss near $192 million — though that loss narrowed sharply from the ~$298 million a year earlier. The company runs an adjusted operating margin of about 15.3% on non-GAAP operating income of $237 million and a real GAAP operating loss. This is a company that is becoming profitable on its own accounting, not one that is there yet. Any valuation of it rests on that future period, not the present.

What makes the acceleration believable rather than a one-off beat is the mechanism underneath. Snowflake's revenue is consumption-based — customers pay as they process data — so the durable signal is where the usage is coming from. Management pointed to CoCo, the AI coding agent, which passed 9,100 accounts after adding more than 2,000 in a single quarter. Net revenue retention held at 126%, meaning existing customers are spending 26% more than they did a year ago, and 828 customers now contribute over $1 million in annual product revenue. Net-new customers grew 32% year over year. That is the absorption evidence: real workloads, real expansion, not one giant contract inflating the print.

Freeze that and step back to the range, because this is where the second leg of the story lives. This 22% move did not come out of a cool tape. Since late May Snowflake has rallied roughly 50% and climbed toward 52-week highs, in the middle of a year that treated big-enterprise software harshly — Microsoft and Salesforce have been down sharply — while data-and-AI platform names took the premium. The proximate spark for the sector was Palantir's blowout, which drove risk-on buying across data and cloud names. So the container shifted: the speculative premium migrated from "software" broadly to "AI-data consumption" specifically, and Snowflake is the named tenant. An analyst at BTIG described the growth as broad-based across the customer base without heavy concentration in any one sector or in AI-native companies alone — the read that says this is structural demand, not the pandemic-era hype concentrated in a few whales.

That is the tape. Now the cost of it. At the after-hours price, Snowflake's market value moved from roughly $106.6 billion at the close to about $130 billion. Against the roughly $6.5 billion of revenue the company is guiding toward this fiscal year, the stock trades near 20 times forward revenue. The technicals tell the same extended story: the share price sits roughly 70% above its 200-day average, and RSI is near 69, territory that has historically meant momentum is rich, not cheap. None of this makes the quarter bad — it is the opposite of bad. It means the re-acceleration is largely already in the price, and the 22% is the market's belief about the future compounding, not a discount being corrected.

Two falsifiers would update the record. The first is mechanical and near-term: the next checkpoint is the fiscal third quarter, where management has guided product revenue around $1.59 billion, a 37–38% increase. If product growth delivers a fourth consecutive acceleration, the container thesis holds. If it merely matches, the multiple carries all the weight. The second is about the container, not the company: watch whether the broad software tape keeps cooling while AI-data names keep absorbing the premium. If it does, Snowflake's premium is a rotation, and rotations can reverse when the next container opens. If data-consumption names go quiet too, then the 22% belongs to the sector's mood, not to Snowflake's fundamentals — and the history books will date the peak to a Palantir echo, not to a balance sheet.

The day is a genuine record: third straight acceleration, a real consumption story, a guidance raise to back it. None of that is in doubt, and no bottom or top is being called on a live tape. The useful question for a retail buyer isn't "was the quarter good" — it was. It is whether the price you pay already includes the growth that has happened, because on a stock 70% above its 200-day, the easy part of this trade has been taken.