Reacceleration to 34% and a Raise 2.6 Times the Beat Clear the Last Upgrade Condition

Results vs. Consensus

Q1 FY2027 Scorecard

Total revenue and EPS consensus figures are the sell-side averages compiled ahead of the print and were cross-checked across two independent providers, which agreed within 0.5% on revenue and exactly on EPS. Snowflake does not guide total revenue, so the product-revenue comparison is against the company's own February 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 67.7% is computed on the unreconciled figures, $0.3946 against $0.2353, rather than on the rounded $0.39 and $0.24 printed in the release. Both quarters contained 89 days, so there is no calendar artifact in the year-on-year comparison this year.

Sequential Comparison

Q4 FY26 comparatives are Snowflake's reported figures for the quarter ended January 31, 2026. Product revenue per day divides reported product revenue by the 89 days in the quarter ended April 30, 2026 and the 92 days in the quarter ended January 31, 2026. It is our calculation and is shown because the sequential optic understates the quarter by three days. Fiscal fourth quarters are seasonally the strongest for renewals, cash collection and net customer additions, so several of the sequential declines above are calendar rather than trend.

Revenue

This is the cleanest beat Snowflake has printed in our coverage, and the important fact is not its size but its direction. Product revenue grew 33.9% against 26.2% a year ago and 30% in the quarter immediately before. A consumption business that accelerates is telling you something different from a subscription business that accelerates, because the revenue was earned by usage that already happened rather than by contracts whose conversion is still in the future. The CFO put the number at "approximately 400 basis points" of acceleration and did not dress it further.

"In Q1, year-over-year product revenue growth accelerated approximately 400 basis points to reach 34%."
— Brian Robins, CFO

Three adjustments are worth making and all three run in the company's favour. The Observe acquisition closed during the quarter and contributed under a point of growth on management's own sizing, so the organic rate is at least 33%. Both this quarter and the year-ago quarter contained 89 days, so unlike last year there is no leap-day artifact to net out in either direction. And the sequential dollar increase of $107.7M landed in a fiscal first quarter, which is Snowflake's seasonal trough for renewals and for net customer additions. Against the four preceding quarters, whose sequential increases were $93.7M, $67.9M, $68.2M and, a year ago, $53.5M, this is a step change in absolute dollars rather than a percentage that flatters a small base.

Assessment: The revenue line resolves the question we have carried since initiation, which was whether this platform compounds at 20-plus percent or decelerates toward it. It now compounds faster than it did a year ago, which is unusual at a $5.8B run rate and is the single largest change in the investment case. The caveat is that the acceleration has one identifiable cause, and that cause is one quarter old.

Margins

The operating margin decomposition is again the most informative table in the print, and this year it has a different author. Non-GAAP total gross margin fell 42bp, to 71.75% from 72.18%. Non-GAAP operating expense fell 354bp as a share of revenue, to 59.84% from 63.38%. Net the two and the operating line moved 312bp, to 11.92% from 8.80%.

What changed is where the leverage came from. A year ago the 575bp of expense leverage was led by sales and marketing at 337bp, with research and development supplying 160bp. This quarter research and development is the largest single contributor at 248bp of the 354bp, sales and marketing supplies 42bp, and general and administrative supplies 65bp. GAAP research and development expense rose 13.2% in dollars, and 19.1% on the non-GAAP basis, while revenue rose 33.5%. The reason is not subtle.

"We added 190 employees this quarter compared to approximately 400 added in the year ago period. Of these 190 employees, 173 joined Snowflake through the Observe acquisition. Excluding Observe, organic hiring was limited to 17 people in the quarter."
— Brian Robins, CFO

Seventeen organic hires against roughly 400 a year earlier is not a hiring pause dressed up as discipline. It is a company that has decided its engineering output no longer scales with its engineering headcount, and both executives spent time on the call describing the internal deployment of Cortex Code that supports that decision. That claim is unauditable from outside. What is auditable is the expense line, and the expense line moved.

On the gross margin, the compression narrowed for a fifth consecutive quarter. Non-GAAP product gross margin of 75.15% is 51bp below the year-ago quarter, against 126bp of compression a year ago and 221bp the quarter before that, and it is 2bp above the prior sequential quarter. The metric has now printed at or above 75% in every quarter since Q1 FY26.

Assessment: Bear-1 in our standing thesis, gross margin compression, moves from emerging to contained. Five consecutive quarters at or above 75% while the AI mix rises is the evidence we said we needed. Bull-2, operating leverage, stays on track and strengthens: the mechanism has now produced 312bp on top of last year's 442bp, and the source rotated from sales-force efficiency to engineering productivity, which is the harder and more durable version of the same claim.

EPS and Below the Line

Non-GAAP diluted EPS of $0.39 beat the $0.32 consensus by 21.9% and grew 67.7% year on year on the unrounded reconciliation figures, $0.3946 against $0.2353, versus 69.6% growth in non-GAAP net income attributable. The 1.9-point gap between those rates is share count: the non-GAAP diluted base rose 1.1% to 375.0M, a much smaller increase than last year's 2.1% because the if-converted contribution from the zero-coupon notes was already in the base.

The quality of the earnings beat is better than the headline in one specific respect. The income tax effect of the non-GAAP adjustments was $50.4M against $23.5M a year ago, more than double. Non-GAAP net income is therefore being taxed harder, not flattered, and the beat is operational rather than a below-the-line artifact.

The GAAP picture improved sharply for the first time in our coverage. GAAP operating loss narrowed to $326.2M from $447.3M. Some of that is the impairment comparison, because the year-ago quarter carried $106.5M of office-exit charges against $17.7M this quarter, but the improvement survives the adjustment: excluding impairment in both periods, GAAP operating margin went to negative 22.2% from negative 32.7%, an improvement of 1,052bp. The gap between GAAP and non-GAAP net income narrowed to $443.6M from $517.4M, which is the first narrowing we have recorded, and stock-based compensation on the cash-flow basis fell to 28.93% of revenue from 36.41%. A year ago the same figure was 36.4% against a full-year target of roughly 37%.

The offsetting fact sits in the share count rather than the expense. GAAP weighted-average shares rose 3.8% to 345.4M, against a 0.3% decline a year ago. Snowflake deployed $300.0M on 1.7 million shares this quarter, an average of about $176.47, versus $490.6M on 3.2 million shares at about $152.63 a year ago. Fewer dollars at a higher price bought roughly half the shares, and approximately $800M remains of the original $4.5B authorization.

Assessment: The earnings improvement is real, operationally sourced and visible in GAAP as well as non-GAAP for the first time. Bear-2, equity compensation and dilution, is challenged rather than resolved. The ratio improved 748bp and the absolute share count still rose 3.8%, which is what happens when the buyback shrinks into a rising stock. With $800M left on the authorization and the shares 36% higher, the anti-dilution program is now materially less able to do the job it was doing a year ago.

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.

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 and the total customer count are call disclosures rather than defined key business metrics. The Q4 FY26 total customer figure is derived by subtracting the 616 net adds from the 13,912 disclosed on this call.

Product revenue

The consumption engine accelerated and, for the first time, management attributed part of the acceleration to a product it sells rather than to migration volume alone. The CFO's framing was that AI works on Snowflake in two directions at once: as a reason customers move data onto the platform faster, and as a revenue line in its own right. Both halves showed up in the same quarter, which is why the growth rate moved rather than held.

"Since last quarter, we've seen a step function change in our AI revenue opportunity led by Cortex Code."
— Brian Robins, CFO

Assessment: The disclosure gap that has run through every quarter of our coverage is unchanged in form and has changed in weight. Management now describes AI revenue as material enough to move a full-year forecast and still declines to size it. A year ago that absence was an inconvenience. Now it is the single biggest input to the guide that an investor cannot verify.

Professional services and other

The smaller line grew 25.1%, roughly nine points slower than product, and the non-GAAP gross margin deteriorated from negative 4.4% to negative 8.2%. On a GAAP basis the gross loss widened to $21.0M from $18.2M. This is the second consecutive year in which the line has moved the wrong way on both counts, and for the second consecutive year neither management nor any analyst mentioned it.

Assessment: Immaterial to earnings and not immaterial as a signal. Professional services is the implementation capacity that turns a signed contract into consumption. If migrations are genuinely compressing from two years to one or two quarters, as the CEO argued on this call, the services line is either being deliberately subsidised to make that happen or is being priced without reference to its cost. Snowflake has never said which, and the widening loss makes the question larger rather than smaller.

Net revenue retention

126% is the first year-on-year increase in the metric we have recorded. The sequence over the last five quarters runs 124%, 125%, 125%, 125%, 126%. When we initiated in February 2025 the metric printed 126% and had been declining for several consecutive quarters; the following quarter it reached 124%, which was the sixth consecutive decline. Management defended a "mid 120s" range throughout; the 124% print sat at the exact floor of it, and we wrote at the time that the growth composition had shifted from cohort expansion toward new-logo acquisition. This quarter both engines ran: the cohort metric turned up and net customer additions grew 38% year over year.

Assessment: The clearest single refutation of our own model in this print. We wrote a year ago that we expected net revenue retention to exit FY26 at 122%, two points below management's stated range, because the trend had been monotonic and nothing on that call suggested an inflection. It exited at 125% and has since gone to 126%. Bull-1 is confirmed without the caveat we attached to it last year.

Remaining performance obligations and the migration of bookings into Q4

RPO of $9.21B grew 38%, against 34% growth in the year-ago first quarter, and fell 5.7% sequentially from the $9.77B that closed FY26. The CFO framed the year-on-year figure and then supplied the structural explanation for both the sequential decline and the shape of the year ahead.

"We continue to see customers favor Q4 renewals. As a result, we expect bookings to be increasingly weighted towards the fourth quarter."
— Brian Robins, CFO

That is a more consequential sentence than it reads. A consumption business whose contracted coverage is increasingly concentrated in one quarter has less useful RPO in the other three, because the metric will fall sequentially in Q1, Q2 and Q3 by construction and then jump. It also means the annual renewal negotiation, and the pricing leverage inside it, concentrates into a single window.

Assessment: Bear-3, visibility erosion, stays contained but the character of the risk changed. Coverage is not thinning; it is bunching. The practical consequence is that the three interim quarters of FY27 will produce an RPO print that looks worse than the business, and the honest read of contractual coverage will only be available once a year.

Customer cohorts

Customers spending more than $1M on a trailing twelve-month basis reached 779, a sequential addition of 46 against 26 in the year-ago first quarter. The CFO disclosed the flow rather than only the stock, which is the more useful figure: 46 customers crossed the threshold in the quarter against 26 a year earlier. Forbes Global 2000 customers reached 813, with 13 added against 4 a year ago. Net new customers of 616 grew 38% year on year, and the $10M-plus cohort added 8 to reach 64.

Assessment: Every cohort metric accelerated at the same time as the revenue rate, which is the combination that distinguishes a demand inflection from a pricing one. The one cohort figure that is not disclosed is the churn side of the net-adds number, so 616 net additions against a base of 13,912 is a net figure whose gross components remain private.

AI adoption

Accounts using Snowflake AI capabilities went from more than 5,200 a year ago to more than 9,100 at the end of FY26 to more than 13,600 now, roughly 98% of the total customer count. Cortex Code, generally available since February 5, was in more than 7,100 accounts by April 30. Snowflake Intelligence accounts more than doubled sequentially from a base of approximately 2,500.

Assessment: The account metric has now saturated to the point of being uninformative. When 13,600 of 13,912 customers touch an AI feature in a given week, the metric measures feature availability rather than adoption depth. The Cortex Code account count is the more useful disclosure because it is a discrete product with a start date, and it is the one metric in this release from which an outsider could begin to infer AI revenue if a per-account figure were ever given. It has not been.

Key Topics & Management Commentary

Overall Management Tone: The most assured call in our coverage, and the assurance was attached to a delivered number rather than to a plan. Prepared remarks opened with the growth rate and the margin, spent the middle on product mechanics rather than customer anecdotes, and closed on a single-sentence thesis. Analyst pushback was narrow and entirely product-economic, concentrated on whether AI consumption carries the same gross margin and whether customers will throttle token spend, and management answered both with mechanism rather than assertion. Nobody asked about cash conversion, macro, competition from the hyperscalers on price, or the guidance risk embedded in extrapolating one quarter of a three-month-old product across a full year.

1. The Acceleration Is a Fact Rather Than a Forecast

Every quarter of our coverage has turned on the same question in different clothing: can this platform hold its growth rate, and how much of the answer sits in the future rather than in the print. In February 2025 the answer was a back-half acceleration that had to be delivered. In May 2025 it was a first half that delivered it. This quarter the question stopped applying, because the rate went up.

"Product revenue came in at $1.334 billion with growth accelerating to 34% year-over-year, up from 30% last quarter and 26% a year ago, marking our strongest sequential dollar growth in company history."
— Sridhar Ramaswamy, CEO

The sequential dollar figure is the part worth dwelling on. Snowflake added $107.7M of product revenue between the fourth and first quarters, a period in which the four prior transitions produced $53.5M, $93.7M, $67.9M and $68.2M. A fiscal first quarter is Snowflake's weakest seasonal window for renewals and for net customer additions, and this one produced the largest absolute increase of any quarter in the series.

Assessment: This is the fact that changes the rating. A consumption model cannot manufacture an acceleration through a bookings pull-forward, because revenue is recognised on usage. Whatever caused it, it happened.

2. The Raise Was 2.6 Times the Beat, Which Is the Test We Set

A year ago we withheld an upgrade on an arithmetic point. FY26 product revenue guidance moved to $4,325M from $4,280M, a $45M increase against a $39.3M beat, so the dollars allocated to the three quarters that had not happened moved by $5.7M. We wrote at the time that a raise which confirms the rest of the year would have added to the rest of the year. This one did.

FY27 product revenue moved to $5,840M from $5,660M, an increase of $180M against a beat of $69.8M versus the guide midpoint. The dollars allocated to Q2 through Q4 moved from $4,395.5M to $4,505.7M, an increase of $110.2M, or 2.5%. Expressed as a growth rate, the ask of the unfinished part of the year went from 26.5% to 29.6% against the $3,475.5M those three quarters produced in FY26.

"Given the strength we've observed in both our core data platform business and AI business, we are raising our guidance for the year."
— Brian Robins, CFO

The non-GAAP operating margin guide moved the same way, from 12.5% to 13.5%. This is the first time in our coverage that Snowflake has raised a full-year margin target mid-year rather than clearing it silently in the fourth quarter.

Assessment: This is the half of our February 2025 upgrade condition that was still outstanding after last May, and this print satisfies it. The raise is larger than the beat, and the full-year margin guide moved for the first time. What the raise does not include is any change to the adjusted free cash flow margin, which we take up separately below.

3. Cortex Code Went from General Availability to Forecast Line in One Quarter

Cortex Code, referred to throughout the call as CoCo, reached general availability on February 5, three business days into the fiscal quarter, and finished the quarter in more than 7,100 accounts. The CFO named it directly as the largest single driver of the increase in the full-year forecast. He also explained, without being pressed, why the February guide could not have contained it: Snowflake builds guidance from observed consumption, and a product that had not shipped had no observed consumption to build from.

"CoCo is contributing meaningful AI revenue while also driving increased engagement across the broader platform."
— Sridhar Ramaswamy, CEO

The second-order argument is the more interesting one, and it is also the harder one to test. Management's claim is that the coding agent does not merely earn its own consumption but accelerates consumption of the underlying platform, because a pipeline, an agent or a migration that used to take a quarter now takes weeks, and the resulting workload runs on Snowflake compute. The CEO extended the product to competing data platforms during the quarter, including Amazon Glue, Airflow, dbt Cloud and Databricks, which is a deliberate decision to let the agent pull work toward Snowflake from outside it.

Assessment: Bull-3 in our standing thesis, that AI widens the addressable workload rather than eroding it, moves from at risk to on track. The evidence is a forecast raise management attributes to a named product with a disclosed account count and a known ship date, which is more than we have had in any prior quarter. The unresolved half is that the revenue itself is still unsized, so the raise rests on a number only management can see.

4. Net Revenue Retention Turned Up

126% breaks a pattern that had been intact for our entire coverage and for six quarters before it. When we initiated in February 2025 the metric had declined for six consecutive quarters to 126%, then hit 124% in the following quarter, which was the exact floor management had verbally defended and the level at which we said the revenue bridge lost its foundation. It has since gone 125%, 125%, 125%, 126%.

The composition matters. A year ago the CFO explained a declining cohort metric by pointing at growth from customers outside the measurement cohort, which is an honest answer and a lower-quality growth mix. This quarter the cohort metric rose while net customer additions also grew 38%, so the two engines are running together rather than one compensating for the other.

Assessment: We were wrong on this metric and the error was directional, not marginal. We modelled 122% exiting FY26 on the argument that the decline had been monotonic; it exited at 125%. The lesson we take is that a cohort metric in a consumption business reflects the vintage of the workloads inside the cohort, and a new workload class arriving at scale resets it. That is what AI did.

5. Operating Leverage Rotated into Research and Development

The 312bp of year-on-year operating margin expansion is smaller than last year's 442bp and the source has changed in a way that matters more than the size. Last year sales and marketing supplied 337bp of 575bp. This year research and development supplies 248bp of 354bp, general and administrative 65bp and sales and marketing 42bp, and the reason is seventeen organic hires.

Management connected the headcount decision to the product directly, and did so in the answer to a question that invited the opposite. Asked why the company was not hiring harder into evident demand, the CEO argued that a large share of the functions Snowflake used to add people to are the functions most amenable to automation.

"We will absolutely continue to invest wherever we get strong leverage."
— Sridhar Ramaswamy, CEO

Supporting disclosures, all internal and none independently verifiable, included over 25% faster support case resolution, roughly 40% less engineering time per ticket in the cloud operations team, doubled developer productivity in the data organisation, and more than 100 automated workflows across finance, marketing, sales and human resources.

Assessment: The strongest and the weakest topic on the call simultaneously. The expense line is a fact and the productivity explanation is a story, and the two are consistent but not the same thing. The risk in a seventeen-hire quarter is that the leverage is being borrowed from FY28 capacity rather than created, and there is no external test for which one it is until growth slows and the organisation has to catch up.

6. Product Gross Margin Held at 75% and the AWS Contract Is the Stated Reason

The sharpest question of the call isolated the tension between a rising AI mix and an unchanged 75% full-year product gross margin guide. The CFO conceded the premise immediately and then named the offset.

"You're absolutely right. Our AI products have a lower gross margin than our core platform."
— Brian Robins, CFO

The offset is a five-year, $6 billion agreement with Amazon Web Services entered into during the quarter, which the CFO described as more than double the prior contract signed in FY23 and cited as the source of lower bandwidth cost. The release adds that AWS has committed to an expanded go-to-market investment alongside it and that Snowflake has now passed $7 billion in lifetime AWS Marketplace sales.

"During the quarter, we entered into a 5-year $6 billion contract with AWS more than doubling our prior contract signed in FY '23."
— Brian Robins, CFO

Neither the release nor the call disclosed whether the $6 billion is a minimum spend commitment, how it is shaped across the five years, or what the offsetting rate concession is worth. What was disclosed is the commitment itself and its role in the margin guide, which makes the 75% target contingent on a supplier arrangement rather than on product economics alone.

Assessment: Bear-1 moves to contained on the printed evidence of five consecutive quarters at or above 75%, and the mechanism behind the next five quarters is now known and is not organic. A gross margin held by a purchasing agreement is held; it is also held by something that reprices in 2031 and whose annual obligation is undisclosed. Averaged evenly the commitment runs $1.2B a year against a cost of revenue that on our FY27 estimates is under $1.8B, which makes it one of the larger contractual obligations on this balance sheet and one of the least described.

7. The Cash Margin Goes Down While the Operating Margin Goes Up

This is the unexplained item in the print and the one we weight most heavily against the upgrade. Snowflake raised the FY27 non-GAAP operating margin guide 100bp, to 13.5%, and left the adjusted free cash flow margin guide at 23.0%. Set those against FY26 actuals of 10.45% and 25.46% and the year is shaped as operating margin up roughly 305bp while cash margin falls roughly 246bp.

The obvious candidate explanation is the acquisition, and management supplied its size.

"Our full year outlook for both non-GAAP operating margin and non-GAAP adjusted free cash flow margin continues to include approximately 150 basis point headwind related to our Observe acquisition."
— Brian Robins, CFO

Because the headwind applies to both metrics, it cannot explain the divergence between them. Strip 150bp from each and the operating margin still rises about 455bp while the cash margin still falls about 96bp. Nor is capital expenditure the answer: it fell 76.8% to $10.5M, which is the delivery of a specific commitment the prior CFO made a year ago about office build-outs ending. Within the quarter itself, adjusted free cash flow rose 28.7% in dollars while revenue rose 33.5%, so the margin slipped 71bp.

Assessment: Bull-4, cash generation, stays at risk for a second year, though the reason has changed. Last year the concern was whether a 43.6% collapse in first-quarter adjusted free cash flow was seasonality; FY26 closed at 25.46% against a 25% guide and answered that question in management's favour. This year the concern is structural: a business converting profit to cash less efficiently in a year when profit is expanding faster. The most likely mechanisms are AI compute paid before the consumption it enables is collected, and a renewal base bunching into Q4, but neither was stated and neither was asked.

8. The Guidance Philosophy Says Unchanged and the Behaviour Says Otherwise

The CFO was asked directly whether a 5.5% beat represents a change in how Snowflake sets guidance, and he declined the framing twice, on both occasions attaching the same explanation.

"Let me emphasize that there's been no change in guidance philosophy and we view a 3% beat is a very solid beat."
— Brian Robins, CFO

The defence is internally consistent. Guidance is built from observed consumption; Cortex Code shipped after the February guide was set and therefore contributed nothing to it; the quarter produced a full period of observation, which has now been layered into the remaining three quarters. On that account the 5.5% beat and the 2.6-times raise are both artifacts of a single product launch landing inside a guidance cycle, and the convention itself is intact.

Assessment: Take the explanation at face value and it is still the most important forward-looking statement on the call, because it defines what a normal quarter now looks like. If the philosophy is unchanged, Q2 should beat the $1,417.5M midpoint by roughly 3%, or about $42M, and the full year should be raised by roughly that amount rather than by three times it. That is the test we will apply in August, and it is a test this management team has set for itself in public.

9. Observe Closed, Natoma Signed, and the Acquisition Layer Became Material

Two transactions bracket the quarter. Observe, announced with the FY26 results, closed during Q1: the cash flow statement shows $252.5M paid for a business combination, goodwill rose $342.8M and net intangible assets rose $204.4M. Its contribution was sized narrowly.

"Observe contributed less than 1 percentage point of product revenue growth in Q1, and we continue to expect the acquisition to add approximately 1 percentage point of revenue growth -- product revenue growth for the full year."
— Brian Robins, CFO

Natoma, an enterprise Model Context Protocol platform, was under a definitive agreement signed in May and had not closed at the print. It brings twenty employees and, on the CEO's description, extends the agent control plane from data into the applications where work happens: email, calendars, Slack, ticketing. Neither purchase price was disclosed.

Assessment: The strategic logic is coherent and the disclosure is thin. Observe is now consuming roughly 150bp of both margin guides for a point of growth, which is an explicit and accepted trade. Natoma is small enough not to matter financially and large enough to matter architecturally, because governed action on third-party systems is the part of the agentic pitch Snowflake could not previously make. What an investor cannot do is judge either price.

10. Three of Four Senior Operating Roles Changed Hands Inside Nine Months

The CFO succession that was open and unaddressed for two consecutive calls in 2025 is closed: Brian Robins joined from GitLab in September 2025, and FY27 is the first full year he owns. A new Chief Revenue Officer, promoted from within after more than a decade at the company, completed his first quarter in the role. And the call closed with the departure of the last founder in an operating seat.

"Before I close, I want to acknowledge our Co-Founder and Chief Architect, Benoit Dageville, who will be stepping away from day-to-day operations in mid-June and continuing as a member of Snowflake's Board of Directors."
— Sridhar Ramaswamy, CEO

Product continues under the same executive who has run it for several years, which is the continuity that matters most for a company whose thesis is now a product thesis.

Assessment: Bear-4, key-person and governance risk, moves from emerging to contained. The specific risk we identified, that the author of the conservative guidance convention was guiding a year he would not finish, is resolved: his successor is seated, has publicly restated the convention, and has now run a full guidance cycle. The offsetting observation is that finance, sales and architecture all changed hands within nine months, and the company is choosing this moment to make its most aggressive forecast revision in years.

11. What Snowflake Stopped Talking About

Three topics that dominated prior calls did not appear once. Iceberg and open table formats, the subject of an entire section of our last two notes, were not mentioned by management or by any analyst. Snowpark, the last new product Snowflake ever sized, at roughly 3% of FY25 product revenue, was not mentioned. Macroeconomic conditions, which occupied a full exchange a year ago and produced the best answer that management had given on the subject, were not mentioned by anyone on either side of the call.

Assessment: The silence on macro is the most notable of the three, because it is the analysts' silence as much as management's. A call in which no one asks a $5.8B consumption business what its guide assumes about enterprise budgets is a call being held at a moment of unusual confidence. The Iceberg silence is now three calls long and continues to leave unanswered the question we first raised at initiation, which is whether compute over externally governed tables earns the same gross margin as compute over native storage. Five consecutive quarters of stable product gross margin is mild evidence that it does.

Guidance & Outlook

Two of the four full-year lines moved and the two that moved are revenue and operating margin. That is the mirror image of last May, when revenue moved by exactly the size of the beat and every margin line was left untouched, and it is the reason this print reads differently from that one despite a superficially similar structure.

Implied quarter-over-quarter ramp. The full-year product revenue target of $5,840M less the $1,334.3M delivered and the $1,417.5M guided at the Q2 midpoint leaves $3,088.2M for the second half, or 29.5% growth against the $2,385.0M Snowflake produced in the second half of FY26. The first half, actual plus guide, runs at 31.8%. The year is therefore shaped as a mild and orderly deceleration from 34% in Q1 to 30% in Q2 to roughly 29.5% across Q3 and Q4, which is a materially less demanding structure than the back-half step-ups Snowflake guided to in each of the two prior years.

Implied margin path. Q1 printed 11.92% and Q2 is guided to 12.5%, so a 13.5% full-year figure requires roughly 14.6% across the second half. Unlike last year, when the unchanged full-year margin implied a second half below both the quarter reported and the quarter guided, the arithmetic here requires genuine sequential expansion in Q3 and Q4. That is a guide with something in it rather than a guide that has been left alone.

Implied cash-flow path. A 23.0% full-year adjusted free cash flow margin on our estimated FY27 revenue base implies roughly $1.17B across Q2 through Q4 against the $986.4M generated in the same three quarters of FY26, an increase of about 18% on a Q2-to-Q4 revenue base growing about 33%. That is not a demanding number in the way last year's was; it is a low number, and the question is why it is low rather than whether it is achievable.

Street at: Snowflake does not guide total revenue, and the Q2 product revenue guide of $1,415M to $1,420M represents 30% growth against the 33.9% just delivered. The full-year target moved 3.2% above the guide the Street had been modelling against since February, and the growth rate attached to it moved 400bp. The sell-side price target cluster entering the print averaged approximately $229, which the reaction session cleared.

Guidance style: Stated as unchanged and demonstrably different in effect. The CFO's position is that a 3% beat is the design point and that this quarter's 5.5% is the residue of a product launch that fell outside the observation window. Both halves of that claim are testable in August. If Q2 beats by roughly 3% and the full year rises by roughly the size of that beat, the convention is intact and this quarter was a one-off adjustment. If Q2 beats by 5% again, the convention has changed and the market will reprice the guide as a floor rather than a forecast.

Analyst Q&A Highlights

What Actually Inflected in the Quarter

The opening question refused to accept the beat at face value and asked instead which of three possible causes was responsible: a change in the demand environment, organic expansion within the core platform, or the AI products. The answer separated all three and, unusually for this company, ranked them. The CFO then supplied the mechanical version, saying that Cortex Code was the largest driver of the increase in the forecast and that the acceleration in the core business informed the outlook as well.

Q: "But the simple question is like what sort of inflected in the quarter on like two fronts, I would say, maybe from a market backdrop demand perspective? And then from like a -- within the Snowflake portfolio, between, let's say, maybe vibrations, organic customer expansion in the core data platform and then the AI story. Can you talk about where specifically you're seeing the inflection?"
— Sanjit Singh, Morgan Stanley

A: "So I would break this up into three parts. First, AI is accelerating the value that people can get from the data that they have put into Snowflake or that they can put into Snowflake."
— Sridhar Ramaswamy, CEO

Assessment: The three-part structure is the most useful thing management said on the call, because it separates a secular tailwind that is unmeasurable from a product line that is measurable but undisclosed. The order also matters: the tailwind was named first and the products second, which is a claim that the core platform, not the AI SKU, carries most of the acceleration. If that is right, the growth is more durable than a new-product ramp. If it is wrong, the guide depends on a three-month-old product.

Whether a 5.5% Beat Represents a Change in Guidance Philosophy

The most direct challenge to the quarter came from an analyst who had been tracking the company's stated beat cadence and noticed that this print broke it. The framing was careful, acknowledging both that the new product helped and that consumption revenue is inherently variable, and it asked management to reconcile the outcome with the convention. The answer conceded nothing on the convention and explained the exception, and the explanation is the most important sentence for anyone modelling the next three quarters.

Q: "Brian, like if you think about the last couple of quarters, you've been telling us about like the beat cadence that you think about. Obviously, this quarter, you beat by much more and CoCo is helping there, but it's also a consumption model. Like how do you think about this going forward? And how should we think about the guidance philosophy that you have here?"
— Raimo Lenschow, Barclays

A: "Let me emphasize that there's been no change in guidance philosophy and we view a 3% beat is a very solid beat. The difference that happened this quarter was CoCo was launched in the quarter."
— Brian Robins, CFO

Assessment: A complete answer to a fair question, and one that sets a public benchmark. Management has now defined a normal quarter as a 3% beat, which means the August print carries a stated expectation of roughly $42M above the Q2 midpoint and a full-year raise of roughly that size. Anything materially above that will indicate that the company is still discovering the size of its own AI revenue, which is a better problem than the alternative and a worse one for the reliability of the guide.

Whether AI Products Dilute the 75% Product Gross Margin Guide

The most analytically precise question of the call worked backwards from the guidance table rather than forward from the narrative. It observed that a large apparent increase in a lower-margin product had produced no change in the full-year product gross margin guide, and asked management to reconcile the two. The concession was immediate and the offset was named, which is the pattern we prefer to a denial.

Q: "I think there might also be a perception that products like Cortex Code come at generally a lower gross margin than the rest of the business. But one thing I noted from your guidance for the full year is that you stuck with the product gross margin guidance of 75% despite a big apparent uptick in Cortex Code, which suggests perhaps that the gross margin drag is minimal, if any."
— Karl Keirstead, UBS

A: "You're absolutely right. Our AI products have a lower gross margin than our core platform. The one thing that we want to do with our AI products when we launch a new product like CoCo is make sure that we develop a great product that we get massive adoption."
— Brian Robins, CFO

Assessment: The premise of the question was wrong in the way that produces the most useful answer. The drag is not minimal; it is being paid for out of a supplier renegotiation. That makes the 75% guide a composite of two moving parts rather than a statement about product economics, and it means the metric will hold for as long as the AWS rate concession outruns the AI mix shift. Neither quantity is disclosed, so the guide is now less informative than it was when it was simply the product margin.

Whether Context and Harness Engineering Constitute a Moat Against the Model Labs

A question about defensibility approached it from the technical side rather than the commercial one, asking what role a data platform plays in the context and harness layers that determine how well an autonomous agent performs, and whether that role is defensible against the frontier labs. The answer came in two parts, and the product executive's half was the more concrete: Snowflake already uses query metadata and platform activity to optimise performance, and the same telemetry now feeds agent context.

Q: "There's been a lot of talk, especially from an autonomous AI perspective, the importance of context engineering and harness engineering. And Sridhar, you mentioned that in your prepared remarks. I guess I'm wondering what role does Snowflake have in that? And how do we think about that from a moat perspective from some of the AI labs?"
— Matt Hedberg, RBC Capital Markets

A: "We have a track record of using metadata and activity inside of Snowflake to drive better results. Oftentimes, it used to be query optimization and performance. And we are now using that same type of civil activity in Snowflake to provide better context to AI."
— Christian Kleinerman, EVP of Product

Assessment: The strongest version of the competitive argument available to this company, and it is an argument from accumulated telemetry rather than from features. Query history and access patterns are an asset a model lab does not have and cannot buy. The limit of the argument is that it is unquantified and unverifiable from outside, and the promise to demonstrate it at a user conference the following week is a promise rather than a disclosure.

How Cortex Code Changes Migration Velocity and the Go-to-Market Motion

A two-part question asked how a coding agent changes the rate at which customers get data onto the platform, and what it does to the shape of the sales organisation. The answer produced the most checkable operating claim on the call: migration timelines compressing from years to quarters, with the company's own two-year Teradata reference deal cited as the old benchmark.

Q: "Sridhar, I want to dive a little bit more into CoCo just in terms of -- how does that sort of change your customers' ability to get more data out of the platform at a faster rate?"
— Kirk Materne, Evercore ISI

A: "I talked again in my remarks about how we did a 2-year Teradata migration. We are engaged in more migrations, but the time lines for doing those now run between a quarter and 2 quarters."
— Sridhar Ramaswamy, CEO

Assessment: If migration cycles genuinely compress from eight quarters to one or two, the revenue effect is not a growth rate but a pull-forward of the entire migration pipeline into a shorter window. That is bullish for FY27 and FY28 and it raises a question about what the pipeline looks like once it has been drained faster than it refills. It also sits awkwardly beside a professional services line whose gross loss is widening, since accelerated implementation is exactly the work that line performs.

Why Sales Hiring Did Not Follow the Demand

The one question that pushed on the cost side observed that a quarter of this quality had not produced a hiring quarter to match, and asked whether the company was under-investing in the go-to-market motion at the moment it should be leaning in. The answer reframed the question around productivity rather than capacity and, notably, described sales efficiency as an AI outcome rather than a management one.

Q: "Sridhar, just on the sales and marketing side, if you the backlog observed, we didn't have a really big S&M hiring quarter. And I'm just curious, based on the demand and everything you're seeing, why not lean a little harder in the go-to-market side."
— Brent Thill, Jefferies

A: "I think part of what we need to understand right now is that there are many, many places in which AI is making Snowflake a lot more efficient. ... We will absolutely continue to invest wherever we get strong leverage."
— Sridhar Ramaswamy, CEO

Assessment: The right question and an incomplete answer. Use cases won per account executive rose 86% year on year, which is the supporting statistic and which management supplied unprompted in the prepared remarks, so the efficiency claim has at least one number behind it. What the answer avoided is the capacity question underneath: a sales organisation that does not grow while bookings concentrate into a single renewal quarter is a sales organisation making a bet on productivity holding. No follow-up tested it.

Whether the Trusted-Data Position Survives Better Models

The closing question was the most existential and the least answerable. It observed that customers and partners describe Snowflake in terms of governance and security guardrails, and asked what protects that position as the underlying models improve fast enough to close functional gaps. The reply enumerated infrastructure that takes years to build, and the product executive added the argument that matters commercially: the configuration already exists and rebuilding it elsewhere is work no customer wants.

Q: "But the AI world is moving so fast. And assuming the competition out there gets better with all this. What makes you confident that Snowflake's position as the trusted enterprise data and AI partners secure over the long term?"
— Koji Kada, Bank of America

A: "Because there are a set of deep infrastructure capabilities that just take a lot of time to develop, whether it is role-based access control and role-level access control at massive scale."
— Sridhar Ramaswamy, CEO

Assessment: A switching-cost argument rather than a capability argument, which is the honest one and the durable one. Access policies, masking rules and identity configuration accumulate over years and carry real migration cost, and that is what has protected data platforms through three technology cycles. The answer's weakness is that it defends the storage and governance layer, which is roughly a tenth of revenue, rather than the compute layer, which is the rest of it.

What They're NOT Saying

1. Any dollar figure for AI revenue, after calling it a step-function change. The CFO described a step-function change in the AI revenue opportunity and named Cortex Code as the largest driver of a $180M full-year raise. There is still no revenue figure, no consumption figure and no attach rate. A year ago the absence was a gap in the disclosure; now it is the largest single input to the guide and it is invisible from outside.
2. Why adjusted free cash flow margin is guided down while operating margin is guided up. FY27 operating margin goes to 13.5% from 10.45% delivered in FY26, and adjusted free cash flow margin goes to 23.0% from 25.46%. The Observe headwind is stated at approximately 150bp and applies to both, so it cannot explain the divergence. Capital expenditure fell 76.8%, so it is not capital intensity. Nobody asked.
3. What the $6 billion AWS contract obligates Snowflake to, and when. The agreement was disclosed as a five-year contract more than double the prior one, and cited as the offset that holds the 75% product gross margin guide. Whether it is a minimum purchase commitment, how it is shaped across the five years, and what rate concession it buys were all left out of both the release and the call.
4. The Snowflake Intelligence account count. Disclosed only as having more than doubled sequentially, from a base the company itself put at almost 2,500 three months ago. Every other AI adoption figure in the release carries an absolute number. This one does not.
5. Why remaining performance obligations decelerated to 38% growth from 42%. The CFO framed the metric against the year-ago first quarter, where the comparison is favourable, and did not address the sequential deceleration from the fourth quarter. The Q4-weighted renewal explanation covers the sequential dollar decline; it does not by itself explain a four-point drop in the growth rate.
6. The price of Natoma, or of Observe. The cash flow statement shows $252.5M paid for a business combination in the quarter, which is the closest an outsider can get to the Observe consideration, and even that figure is net of cash acquired and excludes any equity component. Natoma's terms were not disclosed at all.
7. Iceberg and open table formats, for a third consecutive call. The question we have carried since initiation, whether compute over externally governed tables earns the same gross margin as compute over native storage, has now gone unasked for three quarters while the AI mix has become the dominant margin variable.
8. Macro, in any form, by anyone. A year ago the macro exchange produced the best answer management had given on the subject. This quarter neither prepared remarks nor a single analyst question referenced enterprise budgets, spending environments or the possibility of an optimization cycle. The guide's macro assumption is therefore entirely unstated.
9. Professional services, whose gross loss nearly doubled as a share of revenue. The non-GAAP margin on the line went to negative 8.2% from negative 4.4% while the segment grew nine points slower than product. Unmentioned by management, unraised in Q&A, for the second consecutive year.
10. What happens to the buyback with $800M left. The original $4.5B authorization is roughly 82% consumed, the quarter's $300M bought half as many shares as last year's $491M, and GAAP weighted-average shares rose 3.8%. No comment on whether the authorization will be extended, and no question about it.
11. Any path or timeline to GAAP profitability. Absent from prepared remarks and Q&A for a third consecutive quarter, in a quarter where the GAAP operating loss narrowed $121.1M year on year and the GAAP-to-non-GAAP gap closed for the first time. The company had a better GAAP story to tell this quarter than in any prior one and chose not to tell it.

Market Reaction

• Pre-print setup: SNOW closed at $175.26 on May 27, down 20.1% year to date, down 15.1% over the trailing twelve months and up 21.5% over the trailing thirty days, against a 52-week closing range of $121.11 to $277.14. The S&P 500 was up 9.9% year to date over the same period. The pre-print close sat in the lower third of the 52-week range, and the sell-side price target cluster averaged approximately $229, roughly 31% above the price.
• After-hours move: shares rose approximately 36% in the extended session on May 27, with the move attributed across sources to four items landing together: the product revenue beat and the acceleration to 34%, the $180M full-year raise, the $6 billion AWS agreement, and the Natoma acquisition. Essentially the entire eventual move was priced before the 2:00 p.m. Pacific call concluded.
• Next-day session: the stock opened at $237.00 on May 28, a 35.2% upside gap, traded a $229.49 to $244.92 range, and closed at $239.20, up 36.5% or $63.94 on the day. The close sat $5.72 below the session high.
• Volume: 39.6M shares against a 30-day average of 7.4M, roughly 5.4 times normal.
• Index context: the S&P 500 rose 0.6% on the session, so essentially the entire move was idiosyncratic.

The shape is the inverse of both prior prints we have covered and it inverts them in an unusual way. In February 2025 the stock gapped 12.5% higher and closed on its session low. In May 2025 it gapped 7.2% higher and closed within $0.82 of its high. This time it gapped 35.2% higher, which is to say the extended session did all the work, and the regular session added 0.9% on top of the opening gap while trading a fifteen-point range. A move that is fully priced overnight and then goes sideways on 5.4 times volume is a repricing rather than a discovery: the market read the release, concluded the growth rate had changed, and spent the following session transferring stock between holders at the new level rather than arguing about it.

What the tape priced was the number that requires no interpretation. Product revenue acceleration and a full-year raise more than double the beat are both visible on the first page of the release, and neither requires reading the reconciliation. The second-layer facts, which are the 312bp of margin expansion on seventeen organic hires and the five-quarter product gross margin floor, are the ones that normally get priced during the following session. This time there was nothing left for them to price.

The setup entering the print is the reason the move was as large as it was. The shares came in down 20.1% year to date and down 15.1% over twelve months, in the lower third of a 52-week range, in a market up 9.9% year to date. Snowflake was a de-rated name carrying a consensus-long positioning of 45 buys against 6 holds, which is the configuration that produces a violent upside move when the fundamental question resolves in the bulls' favour. Leaving the session the stock is up 9.0% year to date and still 13.7% below its 52-week closing high, having closed above the entire pre-print sell-side target cluster.

Street Perspective

Debate: Is the CoCo-Driven Raise Durable or a One-Quarter Extrapolation?

Bull view: the bull case being made on the Street is that the acceleration is broader than the product. Management ranked the secular tailwind to the core platform ahead of the AI products themselves, net revenue retention turned up, and every customer cohort metric accelerated at the same time. A raise built on a broad-based inflection is more durable than one built on a single SKU.

Bear view: the bear camp contends that the company has extrapolated eleven weeks of a brand-new product across three quarters it has not observed, using a forecasting method it says depends on observed consumption. The CFO's own explanation for why the February guide missed by 5.5% is precisely the reason to distrust the extrapolation: nobody, including management, knows the steady-state consumption curve of a coding agent that shipped in February.

Our take: both are right and they resolve on a different timescale than the debate assumes. The near-term forecast risk is low, because usage is already running and Q2 is guided with four weeks of the quarter observed. The medium-term risk is real, because coding-agent consumption in an enterprise is a function of seat rollout and cost governance, and management spent a full answer describing the token limits it is building to constrain exactly that spend. We size the durable part at the core-platform acceleration and treat the AI product line as a call option that has been partially exercised.

Debate: Does the 75% Product Gross Margin Guide Survive the AI Mix?

Bull view: five consecutive quarters at or above 75%, compression narrowing from 221bp two years ago to 126bp to 51bp now, and a supplier agreement that more than doubles in size and explicitly funds the offset. The metric that was the central bear point at initiation has held through the fastest mix shift in the company's history.

Bear view: the margin is being held by a purchasing contract, which management said in as many words when asked. That makes 75% a negotiated outcome rather than a structural one, and it defers rather than answers the question of what a mature AI mix does to incremental margin. The full-year guide also sits below the 75.76% the company delivered in FY26, which is the company forecasting its own metric lower.

Our take: the bulls have the printed record and the bears have the mechanism. We move Bear-1 to contained because five quarters of evidence is what we said we needed and we are not going to move the goalposts, but we note that the reason changed underneath us. A gross margin defended by a five-year supplier commitment is defended for five years; the underlying product economics of AI consumption remain undisclosed and will matter at renewal.

Debate: Why Does the Cash Margin Fall While the Operating Margin Rises?

Bull view: the guide is conservative, as this company's guides always are, and a 23% adjusted free cash flow margin against a 150bp acquisition headwind is roughly flat with FY26 on a like-for-like basis. FY26 was guided at 25% all year, printed a 6% second quarter, and finished at 25.46%. The pattern says take the guide as a floor.

Bear view: operating margin is guided up 305bp and cash margin down 246bp in the same year, from the same base, with the same 150bp acquisition drag on each. Something is consuming cash that is not consuming accounting profit, capital expenditure fell 77% so it is not that, and management volunteered no explanation. In a business that is valued on cash, an unexplained widening of the profit-to-cash gap is the item that should have led the Q&A.

Our take: the bears have the arithmetic and the bulls have the track record, and this time the track record is a strong argument, because the identical concern a year ago resolved in management's favour within the same fiscal year. We do not model above the guide and we treat the divergence as the single most important thing to resolve on the August call. The most plausible mechanism is that AI compute is purchased and paid for ahead of the consumption revenue it enables, which would be a working-capital effect of growth rather than a conversion problem, but that is our inference and not a disclosure.

Debate: Is 14x Forward Product Revenue Cheap After a 36.5% Session?

Bull view: the multiple went down, not up. On enterprise value to forward product revenue the stock closed at roughly 14.3x against the 16.1x it carried a year ago, on a business now growing 33.9% instead of 26.2%, with an operating margin of 11.9% instead of 8.8% and a product gross margin that has stopped compressing. Paying a lower multiple for a faster and more profitable business is the definition of the re-rating not having happened yet.

Bear view: a stock that adds $24B of market value in a single session on eleven weeks of data from one product has priced a great deal of a future that has been observed once. The GAAP company still loses $296M a quarter, stock-based compensation runs at 28.9% of revenue and the share count is rising again, and roughly 60 times forward adjusted free cash flow is not a cheap number in absolute terms whatever the revenue multiple says.

Our take: the revenue multiple comparison is the correct one and it is decisive. The reason the stock can rise 36.5% and get cheaper is that the forward revenue base rose faster than the price, which is what happens when a guide moves 3.2% and the growth rate moves 400bp at the same time. We take the bull side while acknowledging the bear's real point, which is that the entry price is now four times better if you bought it the day before.

Model Update Needed

Valuation framework. At the $239.20 close and the guided 376M non-GAAP diluted share count, the equity is worth approximately $89.9B. Cash, cash equivalents and short- and long-term investments were $4,387.5M at April 30. 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 $85.6B. That is roughly 14.3x our FY27E product revenue of $5,975M and roughly 60x our FY27E adjusted free cash flow of about $1.43B. The same conventions applied to the $175.26 pre-print close give roughly 10.3x and 43x, which is the repricing the session performed. A year ago, at the $203.18 post-print close, the stock carried 16.1x forward product revenue and roughly 64x forward adjusted free cash flow.

Fair value range: $235 to $285. The low end corresponds to about 14.1x enterprise value to FY27E product revenue, which is roughly where the stock closed and where we think it finds support if Q2 beats by the 3% management called normal and the full-year raise is correspondingly small. The high end corresponds to about 17.2x, achievable if the adjusted free cash flow margin recovers toward the FY26 level, the Q2 beat again exceeds 3%, and the second-half margin path required by the 13.5% guide is delivered rather than deferred. The midpoint of $260 sits roughly 9% above the reaction close, which is the arithmetic behind an upgrade rather than a hold at a higher fair value.

What would change the number. Upside: an adjusted free cash flow margin above 25% in Q2, a second consecutive raise larger than the quarter's beat, a first quantified disclosure of AI or Cortex Code revenue, product gross margin printing above the 75% guide for a sixth quarter, or long-term financial targets introduced at the June Investor Day that exceed the current framework. Downside: net revenue retention returning to 124%, adjusted free cash flow margin below 20% in Q2, product gross margin below 74.5%, a Q2 beat inside the 3% convention accompanied by no full-year raise, or any disclosure that the AWS commitment carries a minimum spend the current cost base cannot absorb.

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 strengthened materially and for the first time on more than one axis at once. Three of the four bull pillars are confirmed, three of the four bear points are challenged, and the guide-raise condition, the one of the two upgrade conditions from our February 2025 initiation still outstanding after last May, is satisfied by this print. The one pillar that did not improve is the one we chose as the honest valuation denominator: cash conversion is guided down in a year when profitability is guided up, and the reason has not been given.

Action: Upgrade to Outperform from Hold. A consumption business that accelerates 400bp at a $5.8B run rate, raises its full year by 2.6 times the quarter's beat, and gets cheaper on forward revenue while doing it is a different company from the one we rated Hold a year ago. Accumulate on any retracement toward the low end of the $235 to $285 range. Move back to Hold if the August print beats by the 3% management called normal and the full year is raised by no more than the beat, which would mean this quarter was a single product launch rather than a change in trajectory. Downgrade to Underperform if adjusted free cash flow margin prints below 20% in Q2 with the 23% full-year guide reaffirmed and still unexplained.