Wow, this was not exactly the Snowflake report I was hoping for. But not everything was bad, either.
Below are the notes that I found stood out most from this earnings report, with some fundamentals for context in the beginning.
- Product revenue for the quarter was $590.1M, 50% YoY growth and 5.6% sequential growth. Q1 was better than expected and beat guidance by 3%. Sequentially, product revenue is decelerating significantly: 9.7% > 18.2% > 12.1% > 6.2% > 6.3%.
- Q2 guidance calls for $625M, which (assuming no beat) would result in 5.9% QoQ.
- While this looks okay for a consumption-based business, what really turned investors off is the revised full-year guidance - for the second time!
- Initially, Snowflake guided for 47% YoY growth in fiscal 2024 (preliminary, so they didn’t have to do that).
- Then they revised that down to 40% YoY, only to reduce it again to now 34% YoY.
- The market hates that.
- Net Retention Rate declined further from 158% to 151%, which is in line with especially large customers optimizing their SNOW bill and buying only as many credits as they need in the shorter term.
- SNOW added 43 large customers (> $1M), which is similar to the record added in Q4 2023). Strongest customer metric of those SOW reports.
- Total customers: I dislike. SNOW needs total customers so that they can expand them into new large customers, but the total customer net add of 317 customers is at a low since about 3 years ago.
- Forbes 2000 Global customer net added is also comparatively low, but this is a very lumpy metric.
- Remaining performance obligations were $3.4B, 31% YoY, dropping to -6.8% sequentially. This doesn’t give any outperformance upside for the near-term (even if Q1 RPO is also seasonally low, but much more pronounced this Q).
- Non-GAAP adjusted free cash flow was a record $287M, up 58% YoY, representing a record FCF margin of 46%.
Notes on Snowflake AI
- SNOW has a data advantage since many models have been primarily trained with Internet and public data, but with SNOW, enterprises can benefit from customizing these models with their own data. Slootman called out a large US financial institution using SNOW data to train its model.
- Data science, ML + AI use cases keep growing. In Q1, 1,500+ customers used SNOW for any one of these workloads, up 91% YoY.
- SNOW acquired Applica last year. Its language model helps to understand unstructured data, so users can turn unstructured data (like documents, contracts, etc.) into structured properties and reference them for analytics, data science, and AI.
Acquisition of Neeva:
- Why Neeva? —> Team is great + its tech can combine LLM + generative AI with traditional tech which is interesting to get more precise answers.
- “This will enable Snowflake users and application developers to build rich, search-enabled and conversational experiences. We believe Neeva will increase our opportunity to allow non-technical users to extract value from their data.”
- “help us accelerate the efforts around Snowflake as a platform for search and conversational experiences, but most important within the security perimeter of Snowflake with the customers’ data so that they can leverage all this new innovation and technology, but with the safety on the privacy and security of the data.”
- “asking really hard analytical questions that take people weeks and weeks or even months to figure out, that will take some workforce software to do that in a matter of seconds to be productive that way. So, we’re sort of at the top of the hype cycle. The real work really starts now.”
- Analyst: “Do you guys envision kind of building up GPU clusters and offering training and inference on your platform?” → SNOW: “We’re doing all of it. We alluded in the prepared remarks to Applica, which it is a multi-model collection of models being built at Snowflake that requires GPUs.”
The market hates uncertainty, and on this Call, it got a lot of it (bolded is mine):
- “This may well continue near term, but cycles like this eventually run their course. Our conviction in the long-term opportunity remains unchanged.”
- “It is challenging to identify a single cause of the consumption slowdown between Easter and today.”
- “[…] customers remain hesitant to sign large multi-year deals.”
- “Contrary to last quarter, the majority of this underperformance is driven by older customers.”
- “As a result, we’re reining in costs until we see a consistent change in consumption.”
- “[…] and that we don’t have real strong visibility in terms of, “Okay, when is it all going to be different?”
- “This, in the past, is we literally saw four weeks in April where there was no week-over-week growth per se or not material.”
On Optimization & Usage
Snowflake is facing a multitude of different optimizations. On a positive note: some of these help its customers do more with less which is a customer-centric approach and should pay off in the long run.
Currently, they call out 3 optimization patterns:
- Optimizations by cloud vendors due to better hardware, result in better performance.
- SNOW’s own & regular optimizations improve performance and make things cheaper for customers. Optimization 1 +2 create an annual 5% headwind to revenue.
- A few of the largest customers change their storage retention policies —> SNOW loses storage revenue + queries run quicker which also is a headwind for revenue. Change in retention policy was also seen by one of the hyper scalers (I think it’s AWS). But, this should revert at some point since this lower data retention (earlier archiving) comes with a cost for customers.
- “[…] CFO being in the business and[…] So really artificially constraining the demand because of the general anxiety that exists in the economy. So that really needs to start lifting. And that will happen.”
- Customers outside of financial services are the ones optimizing
- Snow reiterated its 10B goal by 2029. From where is confidence? → “What I would say is we have a lot of customers. We have only moved a fraction of their data that we know they have multi-year plans to go on Snowflake, and that’s what gives us the confidence as well as the pipeline of deals. And I’m not just talking to pipeline now. There’s deals for next year that I know their long sales cycles, these big customers. That’s what gives us the pipeline on top of a lot of the new products we have coming out over the next couple of years.”
- The number of queries is up 57% YoY in the quarter, outpacing revenue. Queries are running more efficiently.
- Snowpark: SNOW went from 20% in “one quarter to 30% of our customers using it on at least a weekly basis. We think that’s going to go to 100%.”
Data sharing is an important metric for SNOW’s network effects and is measured via the customers having at least 1 stable edge. The number of customers using stable edges increased from 23% to 25% and grew 61% YoY. We don’t have that much historic data yet, but seasonally, Q1 and Q4 tend to be stronger than the rest of the year. Considering Q1 should be seasonally strong, stable edge growth is okay-ish, but decelerated from previous Q1 sequential growth numbers:
- Q1 2024: 13.4% QoQ
- Q1 2023: 18.4% QoQ
- Q1 2022: 26.4% QoQ
- Powered by Program: This allows customers to build their apps directly on top of Snowflake and reduces potential data redundancies. The numbers here look pretty solid: Power by Program customers grew 128% YoY and 18.1% sequentially. Here, we have even less historic data and it looks like it is also decelerating (probably also due to initially very high numbers when first reporting this metric), but it looks still relatively stable.
- Finally, my personal lowlight of Data Cloud Metrics - Data Market Place listings. These grew only 3% QoQ, showing a significant slowdown. This was not addressed on the conference call, so I am a bit puzzled on why:
- See here the QoQ slowdown starting at Q1 2022: 31.0% > 31.7% > 40.9% > 20.9% > 21.6% > 13.3% > 10.6% > 8% > now 3% QoQ.
Is it “only” macro and law of large numbers? Or is it the competition?
- SNOW mentioned that multi-year bookings are not down “due to competitive pressures, but because customers remain hesitant to sign large multi-year deals.”
- SNOW vs. Microsoft: “And yes, we will compete, we will compete with Microsoft from day one, and that will – and we’ve been very successful in that regard for a whole bunch of different reasons. But people keep on coming, and that’s – and we expect that. […] The good news is that I think the relationship is relatively mature, meaning that when there is friction or people who are not following the rules, we have good established processes for addressing and resolving that.”
- “Where you see huge differences is in the total cost of ownership, and that is not the cost of computing and storage. And that is like, what is the cost to run that technology? And this is where [indiscernible] has a huge advantage. And our customers know that. It’s just – it’s reduced skill sets, far fewer people, not having to touch the complexity of the underlying platforms, I mean, on and on and on.”
- Frank Slootman on Snowpark: “[…] what we’re seeing relative to competitive platforms, Spark and by Spark, we’re seeing Snowpark being only better performance, but better price performance. So, interestingly enough, we see customers giving us technical wins and wanting to migrate because of the better economics of the competitive dynamics.”
We got reminded that AWS is the closest proxy for SNOW’s performance. And well, AWS’ guide could have been a warning for me. Somehow I thought Snowflake would be able to squeeze out some unexpected surprise.
Actually, they did. But not a positive one. Guiding down 2 quarters in a row is pretty meh and indicates that they have very little visibility these days.
The metrics hat I disliked the most (besides the down-guidance) were total customer additions and the Cloud Data Metrics which slowed down across the board, but the Data Market place listings slowed down significantly. No one asked about these on the call. Too bad IMO, the call was pretty much only about optimization/usage and AI - I would have loved some color on the other metrics, too.
Long-term, I think Snowflake is still a great company with a great product and lots of tailwinds (AI, cloud transformation, data gravity, platforming different verticals, etc.).
Short-term, I don’t see so much upside with SNOW. Its proud valuation is currently not reflecting the trend in the fundamentals - actually, vice versa: its current fundamentals don’t properly reflect the high valuation. Of course, with consumption businesses, we can always get quick and unexpected upside again, but I feel like I better don’t count on it for the remainder of this year.
Bottom-line: Long-term thesis intact, short-term upside likely limited.
I won’t sell out of my position, but I am going to trim it to better reflect my current conviction.
What are your thoughts about this quarter? What do you do with your SNOW allocations - that is if anyone still has any ?
Thanks for reading and looking forward to your thoughts.
@LisaOnCloouud9 on Twitter. Long SNOW (but now: shorter than before )