The Quarter Cleared Every Test We Set. The Price Cleared Our Fair Value Range.

Results vs. Consensus

Q2 FY2027 Scorecard

Total revenue and EPS consensus figures are the sell-side averages compiled ahead of the print and were cross-checked across four independent compilations, which agreed on both. Snowflake does not guide total revenue, so the product-revenue comparison is against the company's own May 2026 guidance, a primary-source figure. All actuals trace to the Exhibit 99.1 income statement and the GAAP-to-non-GAAP reconciliation, not to figures quoted verbally on the call.

Year-Over-Year Comparison

Non-GAAP diluted EPS growth of 79.4% is computed on the unreconciled figures, $0.6227 against $0.3472, rather than on the rounded $0.62 and $0.35 printed in the release. Both quarters contained 92 days, so there is no calendar artifact in the year-on-year comparison. Percentage-of-revenue lines are computed from the dollar figures in the GAAP-to-non-GAAP reconciliation rather than from the rounded whole-percentage columns the release prints alongside them.

Sequential Comparison

Q1 FY27 figures are derived by subtracting the reported Q2 from the reported six-month totals in the same Exhibit 99.1, which is the company's own arithmetic rather than ours, except for the customer counts and RPO, which are as disclosed at April 30, 2026. Product revenue per day divides reported product revenue by the 92 days in the quarter ended July 31, 2026 and the 89 days in the quarter ended April 30, 2026. It is our calculation and is shown because the raw sequential figure is flattered by three extra days. The Q1 goodwill balance is derived from the January 31, 2026 balance sheet plus the increase disclosed for the first quarter. Customer counts move by more than the net additions management cites because Snowflake restates prior counts for acquisitions, consolidations and spin-offs and re-bases the Forbes Global 2000 cohort to the current-year list: the release discloses 48 net additions to the $1M-plus cohort and 14 net new Forbes Global 2000 customers against stock differences of 49 and 16.

Revenue

The third consecutive acceleration is the fact of the quarter, and it is a harder fact than the first two were. Product revenue grew 36.8% against 33.9% in the quarter before and 32% a year ago. A consumption business that accelerates three times in a row at a $6B run rate is not describing a pipeline; it is describing usage that has already been metered and billed.

"Product revenue came in at $1.49 billion with growth accelerating to 37% year over year. […] After exiting 30% year over year growth, we have now added 7 points of acceleration."
— Sridhar Ramaswamy, CEO

Two adjustments are worth making and they run in opposite directions. The sequential dollar increase of $157.5M is genuinely the largest in the series, against $107.7M last quarter and $93.7M, $67.9M and $68.2M in the three transitions before that. But a fiscal second quarter carries 92 days against a first quarter's 89, so the honest sequential figure is the per-day one: product revenue per day rose 8.2% rather than the raw 11.8%. At 8.2% this is the second-largest per-day step in the series, behind the 12.4% recorded moving into the first quarter, where the same arithmetic runs the other way because an 89-day quarter follows a 92-day one. Set against the three FY26 transitions, which ran 5.8%, 6.2% and 5.9% per day, the last two quarters sit in a visibly different regime. Running the other way, the year-on-year comparison needs no adjustment at all: both quarters carried 92 days, and the Observe contribution is sized by management at roughly one percentage point of full-year growth, leaving something close to 36 of the 36.8 points organic.

The composition changed as well, and this is the part that matters more than the rate. Last quarter management named one product, Cortex Code, as the largest single driver of the raise. This quarter the CEO put a number on the category for the first time.

"I would roughly call it even our AI products. This is a pretty broad swath at this point. […] They contributed approximately half of the acceleration that we are seeing."
— Sridhar Ramaswamy, CEO

Assessment: Approximately half of the acceleration, on management's own estimate, comes from the AI product family and approximately half from the core platform. That is the first quantification of any kind we have received on AI, and it is a share of a delta rather than a revenue level, so it is still not a number a model can be built on. What it does establish is that the core platform is accelerating on its own, which is the more durable half of the story and the half that was in question at initiation.

Margins

The operating margin decomposition is the most informative table in this print, as it has been in every quarter of our coverage, and this time it splits cleanly into a good half and a bad half. Non-GAAP total gross margin fell 120bp, to 71.75% from 72.95%. Non-GAAP operating expense fell 538bp as a share of revenue, to 56.43% from 61.81%. Net the two and the operating line moved 418bp, to 15.32% from 11.14%.

Both halves are larger than last quarter's. Three months ago the same arithmetic ran 42bp of gross margin drag against 354bp of expense leverage for a net 312bp. The drag has nearly tripled and the leverage has grown by half. That is a company converting a mix shift into an operating result, and it is also a company whose incremental gross margin is falling.

The source of the leverage rotated back. Last quarter research and development supplied 248bp of 354bp on the back of seventeen organic hires. This quarter sales and marketing is the largest single contributor at 260bp of 538bp, with research and development at 160bp and general and administrative at 118bp. In dollars, non-GAAP sales and marketing grew 24.8% and non-GAAP research and development grew 25.1%, both roughly ten points slower than revenue.

The gross margin half is the news. Non-GAAP product gross margin of 74.68% is 177bp below the year-ago quarter and 47bp below the prior quarter, and it ends a five-quarter run at or above 75% that ran from Q1 FY26 through Q1 FY27. The full-year target was cut with it, and the CFO gave the reason without being asked.

"For FY 2027, we now expect 74% non GAAP product gross margin. This revised outlook includes a higher revenue mix from fast growing AI workloads which carry a lower contribution margin today. We are delivering continued operating margin expansion as we offset growing cloud costs with slowing headcount expense."
— Brian Robins, CFO

Read the second sentence carefully. The offset for rising cloud cost is now headcount, not the five-year, $6 billion Amazon Web Services agreement that management named as the offset three months ago. The AWS contract was not mentioned once on this call. Whatever rate concession it buys, it is no longer sufficient to hold the metric at 75%.

Assessment: Bear-1 in our standing thesis, gross margin compression, moves back from contained to emerging. This is the first quarter in six in which the central bear point has moved against us, and it moved on the exact mechanism we specified: AI mix reducing incremental margin. The operating margin still expanded because the expense side is doing more work than the gross margin side is undoing. The question the guide now poses is what happens when the expense side runs out of room and the mix shift does not.

EPS and Below the Line

Non-GAAP diluted EPS of $0.62 beat the $0.45 consensus by 37.8% and grew 79.4% year on year on the unrounded reconciliation figures, $0.6227 against $0.3472. Non-GAAP net income attributable grew 82.0%, so share count cost roughly 2.6 points: the non-GAAP diluted base rose 1.5% to 377.9M.

Two below-the-line items deserve attention because they push in opposite directions and net to almost nothing, which is not obvious from the face of the release. Other income (expense), net swung $39.7M, to income of $34.8M from an expense of $5.0M, and the cash flow statement identifies $34.8M of net realised and unrealised gains on strategic investments as the cause. Snowflake's non-GAAP definition excludes expenses associated with strategic investments but does not exclude gains on them, so that mark sits inside non-GAAP net income. Against it, interest income fell 15.1% to $42.0M on a smaller cash balance, and the tax effect of the non-GAAP adjustments rose 74.4% to $75.0M. Net the whole block and the contribution to non-GAAP net income was negative $1.7M this quarter against positive $1.4M a year ago.

So the earnings beat is operational after all, and the arithmetic that proves it also shows the tax rate is a convention rather than a variable: the GAAP provision plus the tax effect of adjustments over pre-tax non-GAAP income comes to exactly 25.0% in both periods. Nothing in the EPS line is being flattered by a rate assumption.

The GAAP picture improved for a second consecutive quarter and this time without an impairment asterisk. GAAP operating loss narrowed to $263.0M from $340.3M, and neither quarter carried material office-exit charges, so the 1,272bp of GAAP operating margin improvement is clean. GAAP net loss narrowed to $191.7M from $298.0M and GAAP EPS improved to $(0.55) from $(0.89). Stock-based compensation on the cash-flow basis fell to 27.38% of revenue from 35.30%.

The share count is where the improvement stops. GAAP weighted-average shares rose 4.2% to 349.3M, an acceleration from last quarter's 3.8%, and Snowflake repurchased no stock at all in the quarter. All $300.0M of the first-half repurchase was spent in Q1 at an average around $176. The absolute gap between GAAP and non-GAAP net income was $427.0M against $427.3M a year ago, which is to say it stopped narrowing entirely after narrowing for the first time last quarter.

Assessment: The earnings quality is good and the equity story underneath it is not improving. Bear-2, equity compensation and dilution, stays emerging. The ratio keeps falling because revenue is growing 35%, not because the dollar cost is falling: stock-based compensation on the cash-flow basis rose 4.8% in dollars while the share count rose 4.2% and the buyback went to zero in a quarter through which the stock repriced sharply higher. A company that buys $300.0M of its own stock around $176 and then buys nothing across a quarter that opened with a 36.5% single-session repricing is making a judgement, and it is not a bullish one.

Revenue Composition and Key Business Metrics

Snowflake reports as a single operating segment and does not disclose revenue by vertical, geography or product line. The composition that is disclosed is the split between consumption-based product revenue and the services attached to it, plus the five key business metrics defined in every release. Those are the segment-equivalent disclosures and they are where the mix story lives. This quarter the two halves of that split moved further apart than in any prior quarter of our coverage.

Snowflake restates customer counts for acquisitions, consolidations and spin-offs and re-bases the Forbes Global 2000 cohort to the current-year list, so as-reported counts do not always reconcile arithmetically to the company-computed growth rates. The $10M-plus cohort, the total customer count and the stable-edge share are call disclosures rather than defined key business metrics. The Q1 FY27 share of RPO expected within twelve months is the approximately 50% disclosed in the Form 10-Q for the quarter ended April 30, 2026. The year-ago share is derived from the disclosed 54% share, the disclosed 42% year-on-year growth in that dollar amount, and the $6.9B RPO reported a year ago. Cortex Code was renamed CoCo and Snowflake Intelligence was renamed CoWork on June 2, 2026; both are tracked here under the new names.

Product revenue

The consumption engine accelerated for a third consecutive quarter and, for the first time, management assigned a share of the acceleration to a category rather than to a single product. Approximately half came from the AI family, which now runs wider than the two named agents: CoCo, CoWork, AI functions, document processing, machine learning, notebooks and the newly introduced Cortex AI Gateway. The other half came from the core platform, where the mechanism management describes is migration velocity rather than price.

"In Q2, product revenue once again accelerated to reach 37% year over year growth. This marks our third straight quarter of acceleration. Q2 benefited from continued strength in our core data platform business and a meaningful step up in AI revenue."
— Brian Robins, CFO

The one checkable migration claim on the call was a network equipment manufacturer completing a Teradata migration in under three quarters, against the two-to-three years such a project would previously have taken. Last quarter the reference point was a two-year Teradata migration compressing to one or two quarters. The claim is directionally the same and the specific example is new.

Assessment: The disclosure gap that has run through every quarter of our coverage narrowed for the first time and did not close. "Approximately half of the acceleration" is a share of a second derivative. It cannot be turned into a revenue level, an attach rate or a per-account figure, and management declined a direct request for the consumption uplift among AI adopters. What it does do is settle a question we raised three months ago about whether the raise rested entirely on an eleven-week-old product: it does not, and half of it does not.

Professional services and other

This line has now stopped growing. Revenue of $54.9M was up 0.8% year on year and down 3.0% sequentially, against product revenue up 36.8%. The gap is 36 percentage points, against nine points last quarter. On a non-GAAP basis the gross margin swung from positive 3.04% to negative 7.71%; on a GAAP basis the gross loss widened to $20.7M from $15.0M.

We have flagged this line for two consecutive years as immaterial to earnings and not immaterial as a signal, and this quarter it stops being a footnote. Snowflake's argument is that migrations are compressing from years to quarters, that use cases deployed rose 89% year on year, and that use cases per account executive rose 43%. If that is true, implementation work is being done at record volume. The revenue line that captures implementation work grew 0.8%.

Assessment: There are two readings and they point in opposite directions. The bullish one is that CoCo and the partner channel now perform the migration labour that Snowflake used to bill for, which would make a stalled services line the clearest financial evidence available that the agent products work. The bearish one is that services is being given away to accelerate consumption revenue, which would mean roughly $20M a quarter of undisclosed customer acquisition cost sitting in cost of revenue. The margin moving from positive to negative while the revenue stalls is more consistent with the second than the first. Neither management nor any analyst has mentioned this line on any call in our coverage.

Net revenue retention

126% for a second consecutive quarter, and above the prior-year figure for a second consecutive quarter. The five-quarter sequence now runs 125%, 125%, 125%, 126%, 126%. When we initiated in February 2025 the metric had declined for six consecutive quarters and management was defending a "mid 120s" range that the next print sat at the exact floor of. The metric bottomed at 124%, and our own call that it would exit FY26 at 122% was wrong by three points.

The composition remains healthy on both engines. Net new customers grew 32% year on year to 692 and 48 customers crossed the $1M trailing-twelve-month threshold, so the cohort metric is holding while the base outside the cohort expands.

Assessment: Bull-1, consumption durability, stays on track. Holding 126% while adding 692 customers is a better outcome than rising to 127% on a shrinking cohort, and the metric has now held for a second consecutive quarter at a level management once described as the top of its range. The honest caveat is that a two-year measurement window means this metric is reporting on workloads landed in FY25 and FY26; the AI cohort will not show up in it until FY28.

Remaining performance obligations and the shortening of duration

RPO of $9.00B grew 30%, down from 38% growth last quarter and 42% in Q4, and fell 2.3% sequentially. On the headline that is the third consecutive quarter of deceleration in the metric that is supposed to underwrite the guide. The CFO paired it with a disclosure Snowflake has not previously given.

"Of the $9 billion RPO, we expect approximately 54% to be recognized as revenue in the next 12 months. This represents an approximately 42% year over year growth compared to our estimate in the same quarter last year."
— Brian Robins, CFO

That reframes the deceleration. The near-dated portion of RPO, roughly $4.86B, grew 42% while the total grew 30%, which means the long-dated portion grew slowly or shrank. The twelve-month share was approximately 50% at April 30 and approximately 50% a year ago on the same arithmetic, so it moved roughly four points in a single quarter, against a weighted-average remaining capacity-contract life of 2.6 years reported three months ago. Contract duration is getting shorter, not coverage thinner. The company's own risk disclosures name shortened contract durations as a live factor, and management repeated the Q4-weighted renewal explanation it gave three months ago.

Assessment: Bear-3, visibility erosion, stays contained and gets a better instrument. Current RPO growing 42% against product revenue growing 37% is coverage expanding faster than revenue, which is the opposite of the picture the headline metric paints. The cost of shorter duration is that the annual repricing conversation now happens more often, which cuts both ways in a year when the customer is getting more value and the vendor is absorbing more cloud cost. We would rather have this disclosure quarterly than the headline number.

Customer cohorts

Every cohort grew and one decelerated sharply. Customers above $1M of trailing-twelve-month product revenue reached 828, up 27% year on year, with 48 crossing the threshold in the quarter against 46 last quarter. Forbes Global 2000 customers reached 829, or over 41% of that list, with 14 added. Net new customers of 692 grew 32% year on year, a second consecutive quarter of growth above 30% in the net-add line, though the rate itself eased from 38% last quarter.

The exception is the top of the pyramid. Customers above $10M of trailing-twelve-month product revenue went to 65 from 64, an addition of one, against eight in the prior quarter and eight in the quarter before that. This is a call disclosure rather than a defined metric and is subject to the same restatements as the rest, so a single quarter of it is weak evidence. It is also the cohort where an AI consumption uplift, if it exists at the scale management describes, should appear first.

Assessment: The land motion is accelerating and the very top of the expand motion paused. With net revenue retention flat at 126% and the largest cohort adding one customer, the growth this quarter came from breadth rather than from depth at the top accounts. That is not a problem at 37% growth. It is the specific thing to watch if the AI thesis is that existing large customers consume structurally more.

AI product adoption

CoCo, the coding agent that reached general availability on February 5 and finished its first quarter in more than 7,100 accounts, surpassed 9,100 accounts, adding more than 2,000 in the quarter. CoWork reached 5,800 accounts. Snowflake launched over 330 product capabilities to general availability in the first half of FY27, 35% more than in the first half of FY26, and introduced Cortex Sense for business context and Cortex AI Gateway for policy-based model routing, the latter incorporating the Natoma acquisition to let agents act outside Snowflake.

One metric disappeared. Accounts using Snowflake AI capabilities, disclosed at more than 13,600 last quarter and more than 9,100 the quarter before, was not given this quarter. We argued three months ago that the metric had saturated to the point of being uninformative, since it covered roughly 98% of the customer base. Its removal is defensible on those grounds and is still a disclosure going backwards in a quarter when everything else went forwards.

Assessment: Bull-3, that AI widens the addressable workload rather than eroding it, stays on track. Two products with disclosed account counts, ship dates and sequential growth is more than we had a year ago and the account counts are now the most useful AI disclosure the company makes. What is still absent is any denominator: 9,100 accounts using CoCo tells you nothing about what an account spends, and the company has now declined that question directly.

Key Topics & Management Commentary

Overall Management Tone: Confident and noticeably less detailed than three months ago. Prepared remarks led with the growth rate and the margin, spent the middle on product architecture rather than on financial mechanics, and closed on a slogan repeated four times. Analyst pushback was almost entirely absent: twelve questions produced one probe on gross margin, one on whether the AI consumption is durable, and ten invitations to elaborate, and nobody asked about cash conversion, the buyback, the AWS contract, macro or the guide's construction. Management was least convincing where it declined to quantify, offering a share of an acceleration in place of a revenue figure and refusing a direct request for the consumption uplift among AI adopters.

1. The Third Acceleration Is the Hardest One to Explain Away

Each quarter of our coverage has turned on the same question in different clothing: can this platform hold its growth rate. In February 2025 the answer was a back-half acceleration that still had to be delivered. In May 2025 it was a first half that delivered it. In May 2026 the rate went up 400bp. This quarter it went up another 290bp, and the sequence from the fourth quarter of FY26 now reads 30%, 34%, 37%.

"We have just achieved 37% year over year product revenue growth, marking our third straight quarter of acceleration. While expanding our non GAAP operating margin 400 basis points year over year to 15%."
— Sridhar Ramaswamy, CEO

The sequential dollar figure is the part that resists a benign explanation. Snowflake added $157.5M of product revenue between the first and second quarters, against $107.7M in the prior transition and $93.7M, $67.9M and $68.2M in the three before that. Adjusting for the three extra days a second quarter carries, the per-day increase was 8.2%. That is the second-largest step in the series, behind the 12.4% of the preceding transition, where an 89-day quarter following a 92-day one flatters the same arithmetic in the opposite direction. It is roughly twice the per-day dollar addition of the equivalent quarter a year ago, and against the 5.8%, 6.2% and 5.9% of the three FY26 transitions the last two quarters are a different regime rather than a single outlier.

Assessment: One acceleration is a product launch. Two is a trend. Three, with the sequential dollar record broken twice in a row and the company sizing the AI contribution at half, is a change in the trajectory of the business. This is the fact that keeps the rating out of Underperform territory despite everything the valuation section will say.

2. The 3% Convention Failed Its First Test

Three months ago the CFO declined twice to concede that a 5.5% beat represented a change in how Snowflake sets guidance, and offered a specific defence: guidance is built from observed consumption, CoCo had shipped after the February guide was set, and a 3% beat is the design point. We wrote at the time that this defined a public benchmark, that Q2 therefore carried a stated expectation of roughly $42M above the $1,417.5M midpoint, and that a second 5%-plus beat would mean the convention had changed.

Product revenue came in at $1,491.9M, $74.4M above the midpoint, a beat of 5.25%. The full year went to $6,070M from $5,840M, a raise of $230M against a beat of $74.4M, or 3.1 times. Dollars allocated to the two quarters that have not happened rose $155.6M. The CFO's framing of the guidance methodology was unchanged and, this time, unchallenged.

"As always, our forecast is based on observed consumption patterns. There are no changes to our forecast methodology or our guidance philosophy."
— Brian Robins, CFO

Last quarter's version of this raise moved $110.2M into the unfinished year and lifted the implied growth ask from 26.5% to 29.6%. This quarter's moves $155.6M into a shorter remainder and lifts the implied second-half growth to 36.0% against the $2,385.0M the second half of FY26 produced. That is a full-year guide that now asks the back half to grow faster than the front half did, which is the opposite of every guide this company has issued in our coverage.

Assessment: The test we set in May has been answered, and the answer is that the convention no longer describes the behaviour. Two consecutive beats above 5% with raises at 2.6 times and 3.1 times the beat mean the guide is a floor. That is bullish for the next two quarters and corrosive for the information content of guidance, because a floor cannot tell you what management actually expects. The company is discovering the size of its own AI revenue in real time, which is a better problem than the alternative and a worse one for anyone underwriting the forecast.

3. The 75% Product Gross Margin Floor Broke and the Company Cut Its Own Target

This is the single most important thing in the print and it moved against us. Non-GAAP product gross margin printed 74.68%, 177bp below the year-ago quarter and 47bp below the prior one, ending a run of five consecutive quarters at or above 75% that we cited three months ago as the evidence for moving this bear point to contained. The full-year target went to 74.0% from 75.0%.

"For FY 2027, we now expect 74% non GAAP product gross margin. This revised outlook includes a higher revenue mix from fast growing AI workloads which carry a lower contribution margin today."
— Brian Robins, CFO

Three months ago the same executive conceded that AI products carry a lower gross margin than the core platform and named the offset: a five-year, $6 billion agreement with Amazon Web Services entered into during that quarter, more than double the prior contract. This quarter the AWS agreement was not mentioned by management or by any analyst, and the offset named for rising cloud cost was headcount. When pressed on the margin implications of model routing, the CFO was direct about the order of priorities.

"And so you saw our non GAAP product gross margin go down to 74%. Because we have increased our guidance so much. […] But we are still committed as we guided to increase our overall operating margin"
— Brian Robins, CFO

There is a coherent version of this: the margin fell because the mix shifted toward products that are growing so fast they raised the whole forecast, so the dollar gross profit is higher than it would have been at 75% on a smaller base. The arithmetic supports that. Non-GAAP product gross profit was $1,114.1M against $833.6M, up 33.6%, and the guide raise adds revenue at a contribution margin below 75% but well above zero.

There is also a less coherent part. The FY27 guide of 74.0% against a first half that printed 74.90% requires roughly 73.2% across the second half, which is a further 146bp of compression from the quarter just reported. Management gave no bridge for that, and the supplier agreement that was the stated offset one quarter ago went unmentioned.

Assessment: Bear-1 moves from contained back to emerging. We moved it to contained in May on five quarters of printed evidence while noting the mechanism had become a purchasing contract rather than product economics. One quarter later the printed evidence reversed and the purchasing contract was not mentioned. We were right to note the mechanism and wrong to move the tag on the evidence, and the correction is this: a gross margin held by a supplier concession is a gross margin that tells you about the supplier, not about the business.

4. Operating Margin Expanded 418bp and the Leverage Rotated Back to Sales

Non-GAAP operating margin of 15.32% is 418bp above the year-ago quarter and 282bp above the company's own guide of 12.5%. That is the second consecutive quarter of clearing the quarterly margin guide by roughly 285bp, after Q1 cleared a 9.0% guide by 292bp.

The decomposition changed source again. Last quarter research and development supplied 248bp of 354bp of expense leverage. This quarter sales and marketing supplies 260bp of 538bp, research and development 160bp and general and administrative 118bp. In dollars, non-GAAP sales and marketing rose 24.8% and non-GAAP research and development rose 25.1% against revenue up 35.1%, so both lines are growing at roughly seven-tenths the rate of revenue rather than one line being frozen.

The company also raised the full-year operating margin target for a second consecutive quarter, to 14.5% from 13.5%, having taken it to 13.5% from 12.5% in May. Snowflake has now moved its own full-year margin guide upward twice inside one fiscal year, which it did not do once in the two years before.

Assessment: Bull-2, operating leverage, stays on track and is now the most reliable line in the model. The mechanism has produced 442bp, then 312bp, then 418bp of year-on-year expansion across three of the four quarters we have graded, and it has done so through two different sources. The thing to watch is that the gross margin drag has gone from 42bp to 120bp in one quarter while the expense leverage went from 354bp to 538bp. The second number cannot keep growing indefinitely; the first one can.

5. Hiring Restarted, Which Retires Last Quarter's Most Fragile Claim

Three months ago the headline operating fact was seventeen organic hires in a quarter, against roughly 400 a year earlier, and we wrote that the risk in a seventeen-hire quarter is that the leverage is borrowed from FY28 capacity rather than created. The CFO gave the first-half figure this quarter.

"Our outperformance was driven by strong revenue growth and disciplined headcount management. Year to date, we have added 334 employees, which includes a 173 from our Observe acquisition."
— Brian Robins, CFO

Subtract the 190 added in the first quarter, of which 173 arrived with Observe, and the second quarter added 144 people, all of them organic. Organic hiring went from 17 to 144 in one quarter while the operating margin expanded 340bp sequentially. That is a company that hired into demand and got leverage anyway, which is a materially better outcome than getting leverage by not hiring.

Assessment: The FY28 capacity risk we flagged in May is smaller than it was. The absolute level of hiring is still low relative to 35% revenue growth, and the sales organisation in particular is carrying a bookings base that concentrates into the fourth quarter with a headcount that grew a fraction of revenue. But a 144-person organic quarter is not a company running on fumes, and the 260bp of sales and marketing leverage was earned against real additions rather than against a freeze.

6. AI Got Sized, as a Share of a Derivative Rather Than a Level

Asked to separate the contribution of the newer products from the flywheel effect they have on the core, the CEO gave the first quantification of AI that Snowflake has ever offered.

"I would roughly call it even our AI products. This is a pretty broad swath at this point. Absolutely, it is CoCo and CoWork. But it is also things like AI functions that make data operations proceed at an impressive scale or even newer products like the AI gateway. They contributed approximately half of the acceleration that we are seeing."
— Sridhar Ramaswamy, CEO

Note precisely what is and is not being said. Half of the acceleration, not half of the growth and not half of the revenue. The acceleration this quarter was roughly 290bp of year-on-year growth rate, so on management's framing the AI family contributed something on the order of 145bp of it and the core platform contributed the rest. That is a much smaller claim than the headlines around this print suggest, and it is a more credible one.

The definitional boundary is also doing work. "AI products" now includes AI functions, document processing, machine learning and notebooks alongside the two named agents. Several of those existed before CoCo and were previously counted as core platform consumption. A category whose perimeter is set by the company and whose revenue is never disclosed can be sized at half of anything.

Assessment: Progress on the disclosure that matters most, and not enough of it to change how the model is built. We continue to treat AI revenue as unmodellable and to underwrite the platform on core-platform economics, with AI as the reason the core is growing faster rather than as a line of its own. The specific number to demand is a per-account consumption figure for CoCo, which the company has the data for, has been asked for, and has now declined to give.

7. The Cash Divergence Widened, and This Time the Mechanism Is Visible

Three months ago we wrote that a cash margin guided down while an operating margin was guided up was the one number in the print that nobody explained and nobody asked about, and we made it the deciding item against the upgrade. Nothing has been explained and nobody asked again. What has changed is that the cash flow statement now shows where the money is going.

The headline looks benign. Adjusted free cash flow of $92.3M gave a 5.97% margin against 5.92% a year ago, essentially flat, and a fiscal second quarter is Snowflake's seasonal trough for collections. On a first-half basis the picture is worse: adjusted free cash flow margin of 12.18% against 12.53%, down 35bp, while non-GAAP operating margin rose 369bp to 13.71%. Operating cash flow grew 10.3% across the first half against revenue up 34.3%.

The gap is in deferred revenue. It consumed $807.2M of cash across the first half of FY27 against $323.9M in the first half of FY26, two and a half times as much, while accounts receivable released $600.6M against $276.1M. Total deferred revenue fell to $2,596.2M at July 31 from $3,361.4M at January 31. Customers are burning down prepaid capacity faster than they are recontracting, which is exactly what the Q4-weighted renewal pattern and the shortening contract duration would predict, and it is exactly what an acceleration in consumption against a fixed contracted base looks like.

"We are increasing our FY 2027 non GAAP operating margin guidance from 13.5% to 14.5%. […] We are reiterating our full year non GAAP adjusted free cash flow margin guide of 23%."
— Brian Robins, CFO

The guide compounds the point rather than resolving it. The $230M product revenue raise came with a 100bp increase in the operating margin target and no change at all to the 23.0% adjusted free cash flow target. Holding professional services revenue roughly flat, that raise adds approximately $94M to guided non-GAAP operating income and approximately $53M to guided adjusted free cash flow: about 41 cents of incremental operating income per incremental revenue dollar against about 23 cents of incremental cash.

Assessment: Bull-4, cash generation, stays at risk for a second year and the diagnosis has improved. This is a working-capital effect of growth rather than a conversion failure, which is the more benign of the two readings we set out in May, and it is self-limiting: a deferred revenue balance can only drain once before the renewal cycle refills it in the fourth quarter. What we cannot get comfortable with is that management has now had two quarters to say any of this and has said none of it, while raising the profit guide twice and leaving the cash guide untouched. In a business valued on cash, that is a choice about disclosure.

8. The Buyback Stopped and the Acquisition Was Paid in Stock

Snowflake repurchased no shares in the second quarter. All $300.0M of first-half repurchases occurred in the first quarter, at an average around $176 per share, and roughly $800M of the original $4.5B authorization remained at that point. The quarter opened with the stock repricing away from that average: the shares closed at $175.26 on May 27, the session before last quarter's print, and at $239.20 on May 28, a 36.5% single-session move four weeks into the quarter. We hold no price capture for the balance of May through July, so the range across which the company declined to buy is not established here beyond its starting point.

At the same time the balance sheet shows an acquisition. Goodwill rose approximately $101.8M sequentially and cash paid for business combinations in the quarter was $2.0M, against $254.4M for the first half. The Natoma transaction, signed in May and integrated into Cortex AI Gateway during the quarter, was therefore settled almost entirely in something other than cash. Neither the price nor the consideration mix was disclosed.

The share count reflects both facts. GAAP weighted-average shares rose 4.2% year on year, an acceleration from 3.8% last quarter, and proceeds from option exercises rose 141% to $67.8M as employees monetised into the run.

Assessment: Bear-2, equity compensation and dilution, stays emerging and the character of it has sharpened. Last quarter the concern was that a shrinking authorization buys fewer shares at a higher price. This quarter it bought none, and the company issued equity to buy a business while it did so. The ratio metrics keep improving because revenue is compounding at 35%; the absolute share count is rising faster than it was a year ago, and the GAAP-to-non-GAAP net income gap stopped narrowing at $427M.

9. GAAP Profitability Finally Has a Date

For three consecutive quarters we listed the absence of any stated path to GAAP profitability among the things Snowflake was not saying, in a period when the GAAP loss was narrowing sharply. That gap is closed. At the Investor Day held alongside Snowflake Summit on June 2, management set a target of GAAP profitability in the fourth quarter of FY28, and the CEO reaffirmed it on this call.

"And we are pairing this investment in growth with continued operational discipline. We remain on track for GAAP profitability in Q4 fiscal 28, and the operating leverage we build along the way strengthens the durability of that outcome."
— Sridhar Ramaswamy, CEO

The arithmetic required is demanding but no longer implausible. The GAAP operating loss of $263.0M this quarter is a 17.00% negative margin against 29.72% a year ago, an improvement of 1,272bp, and stock-based compensation-related charges of $456.4M are 29.50% of revenue against 38.10%. To reach a GAAP operating profit six quarters from now, the same 860bp-a-year improvement in the equity compensation ratio has to continue while revenue compounds at roughly 30%.

Assessment: A commitment with a date is worth considerably more than a trajectory, and this is the first hard financial target this management team has put on a horizon longer than the fiscal year. It also converts a soft bear point into a checkable one: from here, every quarter can be graded against the ratio path that Q4 FY28 requires. We treat it as a genuine positive and note that it is the one item in this print where the company chose to be more specific rather than less.

10. Outcome-Based Selling Enters the Model

The most commercially novel disclosure on the call was not a product. Asked about the Frontier engineering programme, the CEO described a class of engagement in which Snowflake commits to a business outcome and is paid on delivery.

"It is combining a knowledge of what is possible with the data platform with the harnesses like Cocoa and Cowork, with the industry specific knowledge needed to drive meaningful outcomes to our customers. […] Obviously, we get paid only when we when we deliver outcomes in situations like this."
— Sridhar Ramaswamy, CEO

Snowflake has not sized this, has not said how the revenue is recognised, and has not said whether the specialist hires sit in cost of revenue, sales and marketing or professional services. It is worth noticing that this is being built at the same time the professional services line has stopped growing and its margin has turned negative.

Assessment: Strategically sensible and financially opaque. Outcome-based contracting moves Snowflake up the value chain and gives it a reason to be in the room for budget decisions that a consumption vendor is normally excluded from. It also introduces delivery risk, lumpier revenue recognition and a services cost base into a company whose entire margin story rests on consumption revenue with 75% gross margin. Small today. Worth a line in the model the moment it is sized.

11. What Snowflake Stopped Talking About

Four topics that shaped prior calls did not appear. The $6 billion AWS agreement, named three months ago as the offset holding the product gross margin at 75%, was not mentioned in the quarter that the gross margin target was cut. Iceberg and open table formats went unmentioned for a fourth consecutive call. Macroeconomic conditions went unmentioned for a second consecutive call in our coverage, by management and by all twelve analysts. And the approximately 150bp Observe headwind to both margin guides, stated explicitly on the last call, was dropped this quarter even as the revenue contribution from Observe was explicitly reaffirmed at approximately one percentage point.

Assessment: The AWS silence is the notable one. A supplier agreement introduced as the reason a margin would hold, and then not mentioned in the quarter that margin failed to hold, is either immaterial to the outcome or awkward to discuss. Neither reading is good, and no analyst asked which it is. The macro silence is now a three-call pattern and means the guide's assumption about enterprise budgets remains entirely unstated at a $6B run rate.

Guidance & Outlook

Three of the five full-year lines moved and they do not point the same way. Revenue went up 3.9%, operating margin went up 100bp for the second consecutive quarter, product gross margin came down 100bp for the first time in our coverage, and the cash margin was left alone for the third consecutive quarter. The share count guide also went up 4M, which is a small line that matters given the buyback went to zero.

Implied quarter-over-quarter ramp. The full-year product revenue target of $6,070M less the $2,826.2M already delivered leaves $3,243.8M for the second half, or 36.0% growth against the $2,385.0M Snowflake produced in the second half of FY26. The first half grew 35.4%. This is the first guide in our coverage that asks the back half to grow faster than the front half rather than slower, and it does so without a back-half step-up story attached: the Q3 midpoint of $1,590.5M implies 37.3% and leaves $1,653.3M for Q4 at 34.8%, so the shape is a mild deceleration inside a raised level rather than an acceleration that has to be delivered.

Implied margin path. The first half printed 13.71% and Q3 is guided to 15.5%, so a 14.5% year requires roughly 14.9% in Q4 against the 15.32% just delivered. That is the least demanding margin guide of the three we have graded, and this management has cleared its own quarterly margin guide by 292bp and then 282bp in the two quarters since it was set.

Implied gross margin path. This is the line that got harder. The first half printed 74.90% non-GAAP product gross margin. A 74.0% full year on the guided revenue base requires roughly 73.2% across the second half, which is a further 146bp of compression from the 74.68% just reported. Management gave no bridge and no analyst asked for one.

Implied cash path. Holding professional services revenue roughly flat year on year in the second half, the 23.0% adjusted free cash flow margin guide implies roughly $1,089M across Q3 and Q4 against the $918.6M those two quarters produced in FY26, an increase of about 19% on a second-half revenue base growing about 34%. As in May, this is not a demanding number. It is a low number, and the question remains why it is low rather than whether it is achievable.

Street at: Snowflake does not guide total revenue. The Q3 product revenue guide of $1,588M to $1,593M sits roughly 1.3% above the approximately $1.57B the sell-side was carrying, and the full-year target moved 3.9% above the guide the Street had modelled against since May with the growth rate attached to it moving roughly 500bp. Entering the print the sell-side average price target was $317.38 across 36 analysts with 33 buy ratings and 3 holds and no sells, a configuration that leaves little room for the ratings themselves to move.

Guidance style: No longer describable as conservative-with-a-3%-design-point. Two consecutive beats above 5% and two consecutive raises at multiples of the beat, 2.6 times and now 3.1 times, mean the annual guide functions as a floor. Management's methodology explanation has not changed and is probably accurate: guidance is built from observed consumption, and consumption keeps outrunning the window it was observed in. The consequence for a modeller is that the guide now tells you the minimum rather than the expectation, and the burden of estimating the expectation has moved from the company to the reader.

Analyst Q&A Highlights

What Share of the Acceleration the AI Products Actually Contributed

The second question of the call was the most valuable one, and it was a request for decomposition rather than for enthusiasm. It asked management to separate the direct revenue of the newer products from the flywheel effect those products have on core consumption, and volunteered the questioner's own prior that the core, being far larger, was probably the bigger factor. The answer defined the AI category broadly, put it at roughly half, and named the rest.

Q: "I was wondering if you guys could try to separate out a little bit or give us a little bit of color on how we should think about you know, what portion of the acceleration is coming from these newer products that are obviously, you know, getting really rapid adoption, versus sort of the flywheel of those newer products on the core?"
— Kirk Materne, Evercore ISI

A: "I would roughly call it even our AI products. This is a pretty broad swath at this point. […] They contributed approximately half of the acceleration that we are seeing"
— Sridhar Ramaswamy, CEO

Assessment: The single most useful disclosure on the call and the first quantification of AI that this company has ever given. It is also carefully bounded. Half of an acceleration is not half of growth, the perimeter of "AI products" is set by the company and now includes capabilities that predate the agents, and no revenue level was attached. Taken at face value it means the core platform is accelerating on its own, which is the more durable claim and the one that supports a higher multiple. Taken sceptically it means the company can size an undefined category against a self-reported delta.

Whether an Acceleration This Fast Repeats the Last Cycle's Mistake

The opening question refused to celebrate and instead asked whether this growth resembles the pre-IPO era, when the revenue was driven by unprofitable cloud-native customers, and whether management was seeing poor consumption hygiene from customers running the new agents. The answer went to concentration first and to cost governance second, and the most interesting part of it was that Snowflake's own optimisation tooling is now one of the most used features of the product that drives the consumption.

Q: "And then on sort of the durability of the of the growth, like, are you seeing any sort of irrational behavior or, you know, poor operational hygiene when it comes to con consuming both cocoa and CoWork."
— Sanjit Singh, Morgan Stanley

A: "First, I think we see the acceleration come from a very broad swath of customers. It is not concentrated, for example, with let's say, AI native companies."
— Sridhar Ramaswamy, CEO

The CFO added the two data points that bear on durability: guidance is set on observed behaviour and the behaviour has now been observed for a couple of quarters, and gross retention has been "relatively flat across the last several quarters," a metric Snowflake has never disclosed the level of.

Assessment: A fair question, answered on the right axis. Concentration is the correct thing to test in a consumption model, and a broad-based acceleration is far more valuable than a narrow one. The unfalsifiable part is that "broad swath" carries no distribution: with 828 customers above $1M and 65 above $10M, a great deal of a $157M sequential increase can be broad-based and still be concentrated in the top two hundred accounts. The genuinely reassuring detail is the cost-management skill being a top-ten use of the agent, because it means the optimisation reflex that broke the 2023 consumption cycle is now inside the product rather than outside it.

Whether Model Routing Helps or Hurts Gross Margin

The only question that pushed on margin came from a line of enquiry about model neutrality, and it did the useful thing of asking what a mix shift between open-weight and frontier models does to Snowflake's own economics. The product answer is that the company earns different economics on open models because it runs the inference itself, and that automatic routing gives it latitude to optimise. The CFO's unprompted addition was the concession.

Q: "I am guessing the bulk of tasks, completed by COCO are being directed to Frontier Labs. But I am just curious, during the quarter, did you detect any interesting behavioral shift, let's say, a mix shift from open class models to sonnet class models if that happens, Brian, is there any effect potentially positive on gross margins to Snowflake's financials?"
— Karl Keirstead, UBS

A: "And so you saw our non GAAP product gross margin go down to 74%. Because we have increased our guidance so much. […] But we are still committed as we guided to increase our overall operating margin"
— Brian Robins, CFO

Assessment: The answer inverts the question. It was asked whether model mix could help gross margin and answered with why gross margin fell and why that is acceptable. Both halves of the reply are honest, and together they establish the order of priorities: adoption first, operating margin second, gross margin last. That ordering is defensible while revenue compounds at 35% and expense leverage is running at 538bp. It becomes the central risk the moment either of those slows, because it says gross margin is the line management is willing to spend.

Whether AI Consumption Is Incremental or Pulled Forward

The sharpest analytical question of the call asked for the number that would settle the AI thesis, and got a refusal. It asked what the consumption uplift is for AI adopters against comparable non-adopters, how that has developed across the earliest cohorts, and what evidence distinguishes higher lifetime consumption from projects merely being brought forward.

Q: "And so what is the consumption uplift for AI adopters relative to comparable nonadopters? How is that developed across the earliest cohorts? And what evidence are you seeing, or maybe what is giving you the confidence that all of this reflects higher lifetime consumption rather than projects just being pulled forward."
— Koji Ikeda, Bank of America

A: "At this time, we are not ready to share the exact Uplift numbers, but we do measure cohort behavior"
— Sridhar Ramaswamy, CEO

The substantive part of the answer offered breadth and depth of use cases as the evidence: work that previously required a specialist application and a multi-quarter implementation is now done conversationally, and the conversations Snowflake is having about supply chain, support and fraud detection are ones it would not previously have been in.

Assessment: The company measures the number, was asked for it directly, and declined. That is a materially different posture from not having the metric, and it is the second consecutive quarter in which the single most decision-relevant AI disclosure has been withheld while the AI narrative carried the guide. The qualitative answer is not empty, but breadth of use cases cannot distinguish durable consumption from a migration pipeline being drained faster, which is precisely what the question asked.

Whether the Raise Is Core or AI, and Why the Answer Changed

The last question of the call was the one that connected this quarter's framing to last quarter's. Three months ago management named the coding agent as the largest single driver of the full-year increase; this quarter the framing was balanced between core and AI, and the questioner wanted to know whether something changed in core consumption or whether the confidence simply reflects being halfway through the year.

Q: "I think the last quarter, you had mentioned most of the full year guide increase was on account of COCO. This quarter, it sounds a lot more balanced between core and AI and your confidence in forecasting acceleration and product revenue growth also seems to be much higher. Just wondering if there is something fundamentally that changed during the quarter in terms of consumption of the core from your customers that is driving that higher visibility?"
— Dharmik Jhaveri, J.P. Morgan

A: "We base our guidance based on observed behavior up until the call that we have."
— Brian Robins, CFO

Assessment: The question spotted the shift and the answer did not explain it. Management's position is that the same methodology produced a different attribution because the observed behaviour changed, which is internally consistent and unfalsifiable from outside. The important content is what the shift itself implies: a raise attributed to a single new product is a call option, and a raise attributed half to the installed base is an earnings stream. If the second framing holds through the fourth quarter it is worth more to the multiple than the first was.

Outcome-Based Selling and the Size of the Budget Now in Reach

A question about the addressable budget produced the most commercially novel disclosure of the call. It asked how large the opportunity is now that the conversation reaches strategic priorities rather than data infrastructure, and asked for detail on the Frontier programme and outcome-based selling. The answer described industry specialists paired with the agent products and, in passing, a contracting model Snowflake has not previously discussed.

Q: "So maybe can you just talk about how much what is the ambit of opportunity that you are now able to access and see in terms of budget dollars? And maybe weave in. We have heard some really exciting tales of your Frontier program and some of the you know, exceptional, traction that is getting out there in the marketplace particularly on the outcome based selling"
— Alex Zukin, Wolfe Research

A: "It is combining a knowledge of what is possible with the data platform with the harnesses like Cocoa and Cowork, with the industry specific knowledge needed to drive meaningful outcomes to our customers. […] Obviously, we get paid only when we when we deliver outcomes in situations like this."
— Sridhar Ramaswamy, CEO

Assessment: A strategically significant answer that no follow-up examined. Outcome-based contracting changes the risk profile of the revenue, the timing of its recognition and the cost structure required to deliver it, and it is being built in the same quarter the professional services line stopped growing and its gross margin went negative. Unsized, undisclosed and probably immaterial today. The reason to flag it now is that if it scales, the 75% product gross margin conversation becomes a different conversation entirely.

Whether Application Vendors Turning Into Databases Threatens the Position

The most existential question asked about the pattern of large software companies partnering with model providers and taking on more of the system-of-record data, with the model layer owning the user interface, and whether Snowflake should respond by holding more operational data itself. The two-part answer conceded the competitive framing and then went at it from the customer's side.

Q: "Sridhar, I wanted to ask your take on some of the moves we have seen from traditional SaaS companies partnering with LLMs and sort of becoming more of a of a database themselves as the LLMs sort of take the UI layer How do you see this playing out? Does it make sense for Snowflake to take on more of the system of record data?"
— Tyler Radke, Citi

A: "You can definitely expect to hear a lot more about things like hybrid tables, and Postgres because they are the foundation we think, for a new generation of agentic applications, some of which will have UI and some of which will not."
— Sridhar Ramaswamy, CEO

The product executive supplied the commercial version: application vendors becoming database players is not a new trend, and what customers say is that they will not copy their data into three platforms to use three applications, so consolidation onto a central platform wins.

Assessment: The consolidation argument is the right one and it has held through three technology cycles. The new content is the operational-database ambition, which is a genuine expansion of scope and a genuine expansion of the competitive set. Hybrid tables and Postgres put Snowflake into a market with entrenched incumbents and lower gross margins than analytic compute, and the company has just cut its product gross margin guide. It is a coherent strategic response and it is not a free one.

What They're NOT Saying

1. The consumption uplift for AI adopters, which the company says it measures. Asked directly for the uplift against comparable non-adopters and for the cohort development behind it, the answer was that Snowflake is "not ready to share the exact Uplift numbers." This is the number that separates a durable consumption step from a migration pipeline being drained faster, and it exists internally.
2. Any dollar figure for AI revenue, for a second quarter after calling it a step-function change. The CFO described "a meaningful step up in AI revenue" and the CEO put the AI family at approximately half of the acceleration. There is still no revenue level, no per-account consumption figure and no attach rate, in a quarter where the category is credited with roughly half of a 290bp move in the growth rate.
3. Why the adjusted free cash flow margin guide did not move when the operating margin guide moved for a second time. The $230M revenue raise added approximately $94M to guided non-GAAP operating income and approximately $53M to guided adjusted free cash flow. The first-half cash margin fell 35bp year on year while the first-half operating margin rose 369bp. Two consecutive calls, no explanation, and no analyst question either time.
4. The $6 billion AWS agreement, in the quarter the margin it was supposed to defend failed to hold. Three months ago it was named as the offset holding product gross margin at 75%. This quarter the target was cut to 74.0%, the offset cited was headcount, and the agreement was not mentioned by anyone. Its shape across five years, whether it carries a minimum spend, and what rate concession it buys all remain undisclosed.
5. A bridge from the 74.68% just printed to the roughly 73.2% the second half needs. The full-year 74.0% product gross margin guide requires another 146bp of compression from the quarter just reported. Management supplied the direction and the cause and no arithmetic.
6. Anything at all about the buyback. Zero shares repurchased in the quarter, after $300.0M in the first quarter at an average around $176 and with roughly $800M of the $4.5B authorization remaining. GAAP weighted-average shares rose 4.2%. No comment on whether the authorization will be extended and no question about why it went unused while the stock ran.
7. What Natoma cost, or what it was paid for with. Goodwill rose approximately $101.8M in a quarter in which $2.0M of cash went to business combinations, so the consideration was almost entirely non-cash. The company issued equity to buy a business in the same quarter it declined to buy back its own.
8. Professional services, for a third consecutive year. Revenue grew 0.8% against product revenue at 36.8% and the non-GAAP gross margin swung from positive 3.04% to negative 7.71%. Unmentioned by management, unraised in Q&A, in a quarter whose central operating claim is that implementation timelines are collapsing.
9. The level of gross retention, offered as reassurance without a number. The CFO said gross retention "has been relatively flat across the last several quarters." Snowflake has never disclosed the level, so a flat unknown is being used to support a durability argument.
10. Regional performance, in response to a direct question. Asked specifically about Europe and Asia against a business still indexed to North America, the answer was that "all regions are performing" and that all are factored into the guide. No growth rate, no mix, no relative commentary. This is the least informative answer on the call.
11. The count of accounts using Snowflake AI capabilities. Disclosed at more than 13,600 last quarter and more than 9,100 the quarter before, and absent this quarter. We argued in May that the metric had saturated and stopped being informative, so its removal is defensible; it is still a disclosure going backwards.
12. The 150bp Observe margin headwind, dropped while the revenue contribution was reaffirmed. The CFO restated Observe at approximately one percentage point of full-year product revenue growth and did not repeat the approximately 150bp headwind to both margin guides that was stated explicitly on the last call.
13. Iceberg and open table formats, for a fourth consecutive call. The question we raised at initiation, whether compute over externally governed tables earns the same gross margin as compute over native storage, has now gone unasked for four quarters while AI mix became the dominant margin variable.
14. Macro, in any form, by anyone. Neither prepared remarks nor any of twelve analyst questions referenced enterprise budgets, spending environments or the possibility of an optimisation cycle. The guide's macro assumption is entirely unstated at a $6B run rate.

Market Reaction

• Pre-print setup: SNOW closed at $305.84 on September 2, up 39.4% year to date, up 30.9% over the trailing twelve months and down 0.5% over the trailing thirty days, against a 52-week closing range of $121.11 to $337.38. The S&P 500 was up 12.0% year to date over the same period. The print-day session itself fell 4.4% from $319.80 on volume of 11.6M shares against a roughly 4M-share average, and the stock had given back 8.1% from its August 21 close of $332.78 across the eight sessions into the print.
• After-hours move: shares rose roughly 22% to 23% in the extended session on September 2, with the move attributed across sources to the product revenue beat and acceleration to 37%, the $230M full-year raise, and the 100bp increase in the operating margin target.
• Next-day session: the stock opened at $377.25 on September 3, a 23.3% upside gap, traded a $355.47 to $384.56 range, and closed at $356.47, up 16.6% or $50.63 on the day. The close sat $28.09 below the session high and $1.00 above the session low.
• Volume: 21.8M shares against a 30-day average of 4.5M, roughly 4.8 times normal.
• Index context: the S&P 500 rose 1.1% on the session, so essentially the entire move was idiosyncratic.

The shape is the opposite of last quarter's and the difference is the whole story of the session. In May the stock gapped 35.2% higher and closed within $5.72 of its high, which is a repricing: the market read the release, concluded the growth rate had changed, and spent the following session transferring stock at the new level. This time it gapped 23.3% higher, traded up to 25.7%, and then gave back 29.1% of the opening gap to close a dollar off the low. A gap that fades on 4.8 times volume is distribution, not discovery.

Two things explain the fade and neither of them is in the release. The first is that the news was worse the further into the document a reader got. The first page carries a 37% growth rate, a $230M raise and a 100bp operating margin increase. The financial outlook section, four paragraphs down, carries a product gross margin guide cut from 75.0% to 74.0% and an unchanged cash margin. The opening print was made by the first page and the session was made by the rest of it.

The second is where the price started. Entering the print the stock was already up 39.4% year to date against an S&P up 12.0%, having risen more than two and a half times from the $121.11 low of its 52-week closing range, and the sell-side average target of $317.38 sat only 3.8% above the pre-print close with 33 of 36 ratings already at buy. There was no rating capacity left and very little target capacity, which is a setup in which good news has to be paid for entirely by multiple expansion rather than by positioning. The stock cleared its 52-week closing high by 5.7% and finished 12.3% above the entire pre-print target cluster, and the post-print target raises that followed clustered from the high $300s to $450.

The pre-print tape is worth noting as well. The stock fell 4.4% on the day of the print on nearly three times normal volume and had drifted 8.1% lower over eight sessions. Somebody de-risked into this print, and the 23.3% gap is partly the cost of having done so.

Street Perspective

Debate: Is a Third Acceleration a Trend or the Top of One?

Bull view: the bull case being made on the Street is that three consecutive accelerations at a $6B run rate, with the sequential dollar record broken twice, cannot be a product artifact. Management sized the AI family at roughly half the acceleration, which means the core platform is accelerating on its own, and the second half is now guided to grow faster than the first half rather than slower. Net revenue retention is holding at 126% while net customer additions grow 32%, so both engines run.

Bear view: the bear camp contends that migration pull-forward is the mechanism and that it is finite by definition. If AI compresses a two-year Teradata migration into three quarters, the revenue does not increase, it arrives sooner, and the pipeline is drained faster than it refills. The company was asked directly whether the consumption is incremental or pulled forward and declined to answer with data it says it has.

Our take: the bulls have the better of it for the next four quarters and the bears have the better question. A pull-forward would show up as a rising growth rate with falling contracted coverage, and instead current RPO grew 42% against product revenue at 37%, which is coverage building faster than revenue. That is inconsistent with a drain. What it does not settle is FY29, when the migration cohort has landed and the growth rate has to come from consumption depth rather than from new workloads arriving.

Debate: Does the Gross Margin Cut Matter, or Is It the Right Trade?

Bull view: the argument in the market is that a gross margin falling because a higher-margin-dollar mix is growing faster is a good problem. Non-GAAP product gross profit rose 33.6% in dollars, the operating margin expanded 418bp anyway, the full-year operating margin target went up 100bp in the same breath as the gross margin target came down, and management has said plainly that adoption comes first and margin comes after. On that framing 74% is the price of a 36% growth rate and it is cheap.

Bear view: the bear camp contends this is the first crack in a thesis that has always rested on 75% product gross margin as the floor of an unusually good business model. The metric has now printed below 75% for the first time in six quarters, the company cut its own target by 100bp, the supplier agreement introduced as the offset one quarter ago went unmentioned, and the second half requires another 146bp of compression with no bridge. Operating leverage is finite; mix shift is not.

Our take: the bears have this one and the market has not priced it. The important fact is not the 74.68% print, which is small, but that the company revised its own full-year target downward in a quarter when everything else went up. A management team that raises revenue and operating margin and cuts gross margin is telling you which line it has the least control over. The trade is right at 36% growth. It is the line to watch when growth is 25%.

Debate: Why Does Cash Conversion Keep Lagging Profit?

Bull view: the bull case is that this is seasonality plus conservatism, as it was last year. Snowflake guided 25% all through FY26, printed a 6% second quarter, and finished at 25.46%. The second quarter is structurally the trough, this one printed 5.97% against 5.92% a year ago, and the 23.0% full-year guide requires only 19% dollar growth in the second half against revenue growing 34%. Take the guide as a floor, as every other guide from this company has proved to be.

Bear view: the bear camp contends the divergence is now structural and quantified. Deferred revenue consumed $807.2M of cash in the first half against $323.9M a year ago, first-half operating cash flow grew 10.3% against revenue up 34.3%, and the incremental economics embedded in the raise are 41 cents of operating income against 23 cents of cash per revenue dollar. Management has had two quarters to explain it and has said nothing.

Our take: the mechanism is a working-capital effect of growth rather than a conversion failure, which is the more benign of the two readings, and it is visible in the filing even though nobody named it. Consumption is outrunning the contracted base while renewals bunch into the fourth quarter, and a deferred revenue balance can only drain once. We take the bull side on the arithmetic and the bear side on the disclosure: the company is valued on cash, this is the second consecutive quarter in which the cash guide alone did not move, and the silence is now a choice rather than an oversight.

Debate: Is 21 Times Forward Product Revenue Defensible After a 49% Quarter?

Bull view: the argument for paying up is that the growth rate keeps rising and the terminal value keeps getting bigger. On our FY28 estimates the stock is at roughly 17 times product revenue, the operating margin is guided up twice in one year, GAAP profitability has a date on it for the first time, and a company accelerating through 37% at a $6B run rate is scarce enough to deserve a scarcity multiple. The post-print target raises ran to $450.

Bear view: the bear camp contends the multiple has done the work, not the business. Enterprise value has gone from roughly 14 times our forward product revenue at the May close, to roughly 18 times entering this print, to roughly 21 times leaving it, on a business whose forward product revenue estimate rose 3.8%. That is a 49% price move against a 4% forecast move in one quarter. On cash the number is worse: roughly 89 times our FY27 adjusted free cash flow estimate, against roughly 60 times three months ago.

Our take: the bears have the arithmetic and it is decisive for the rating. Three months ago we upgraded on the observation that the stock could rise 36.5% and get cheaper, because the forward revenue base rose faster than the price. That is no longer true and it inverted violently: this quarter the price rose 49% and our forward revenue base rose 3.8%. The business has done nothing wrong. Everything that has changed in the risk/reward has been paid for in the price, and at 18.6 times our next-twelve-month product revenue the stock sits at the top of the band we can defend.

Model Update Needed

Valuation framework. At the $356.47 reaction close and the guided 380M non-GAAP diluted share count, the equity is worth approximately $135.5B. Cash, cash equivalents and short- and long-term investments were $4,329.2M at July 31. Under the if-converted treatment already embedded in the guided share count, the zero-coupon notes are equity rather than debt, so enterprise value is approximately $131.1B. That is roughly 21.2 times our FY27E product revenue of $6,200M, roughly 16.7 times our FY28E product revenue of $7,875M, and roughly 89 times our FY27E adjusted free cash flow of about $1.48B.

The same conventions applied to the $305.84 pre-print close give approximately 18.0 times FY27E product revenue, and applied to the $239.20 close after last quarter's print give approximately 14.3 times against the then-current estimate. The multiple has therefore expanded roughly 48% in one quarter while our forward revenue estimate moved 3.8%. Every prior quarter of our coverage has been a story about whether the business would earn its multiple. This one is a story about the multiple.

Fair value range: $305 to $365. The framework is enterprise value to product revenue over the next twelve months, which spans the third and fourth quarters of FY27 and the first and second of FY28 and comes to approximately $7,032M on our estimates. The low end corresponds to roughly 16 times that base and the high end to roughly 19 times. At the $356.47 close the stock trades at approximately 18.6 times, near the top of the band. The midpoint of $335 sits roughly 6% below the close.

What would change the number. Upside: a third consecutive beat above 5% with a raise larger than the beat; a first quantified disclosure of AI revenue or of the per-account consumption uplift; product gross margin printing above 74.68% in the third quarter against a guide that requires 73.2% across the second half; adjusted free cash flow margin above 30% in the third quarter; or a resumption of the buyback with an extended authorization. Downside: a beat inside the 3% convention with a raise no larger than the beat; product gross margin below 73.5% in the third quarter; first-half-to-third-quarter cumulative adjusted free cash flow margin below 15% with the 23.0% full-year guide reaffirmed and still unexplained; net revenue retention below 125%; or the current RPO growth rate falling below the product revenue growth rate.

Thesis Scorecard Post-Earnings

The pillars below are the same eight established at initiation in February 2025, graded against what this quarter's print and call revealed. The status tags are the ones carried in our thesis of record.

Overall: the thesis is unchanged in direction and the balance of evidence shifted for the first time in three quarters. Three of the four bull pillars are confirmed and the operating record is the best in our coverage. Against that, the central bear point moved against us on the exact mechanism we specified, the cash pillar has now been at risk for two consecutive years on two different explanations, and the equity-count problem got worse rather than better in a quarter when the company had every reason to lean on its buyback. Nothing here breaks the investment case. The case is simply no longer cheap.

Action: Downgrade to Hold from Outperform. We upgraded in May at $239.20 on the argument that the stock could rise 36.5% and get cheaper because the forward revenue base was rising faster than the price. Three months later the price has risen another 49% and the forward revenue base has risen 3.8%, so the same argument now runs the other way. Take profits into strength rather than adding. We would return to Outperform below roughly $310, which is approximately 16 times our next-twelve-month product revenue and roughly where the shares traded before this print, or on a third consecutive beat above 5% accompanied by either a first quantified AI revenue disclosure or a resolution of the cash divergence.