SNOW - Why I think the market is wrong

Today, after Snowflake reported 22Q4 earnings, the stock has dropped by 22.08% to $206.25 in AH as I’m writing this post. It used to drop to below $190. I raised considerable cash by trimming my outsized DDOG position and added a lot of SNOW at about $193. I won’t post the financial numbers, which could be easily found in the press release, but I will explain briefly why I think the market is wrong about Snowflake this time.

To summarize what in the earning call that the market is not satisfied with:

  1. Snowflake is rolling out a huge 10-20% improvement to their system, which they expect to negatively impact the FY23 full year revenue by $96.7 million. The actual number they projected was even larger at $160.2 million, but the management expected that they could make up $65.5 million from increased workloads.
  2. The management said the DBNRR in FY23 will be definitely below 170% and should be above 150% for quite a while, primarily due to the revenue impact from system improvements and the law of large numbers.
  3. The management mentioned a slower-than-expected consumption in January, primarily due to less-than-usual human-driven queries.

Before talking about the positive side of the earning call, let’s pause here a little bit. Isn’t this very similar to the situation of Datadog in 20Q2? At that time, because of businesses cutting IT budgets, the customers of Datadog did a lot of optimization work to reduce the total number of machines they need for the cloud workloads. What’s the same between that optimization and Snowflake’s current one is that both optimizations are one-off and will only affect the QoQ revenue growth of the quarter when the optimization happens. And of course, both optimizations are beneficial for their customers, though Datadog’s was driven by its customs and Snowflake’s was driven by Snowflake.

At the same time, despite the one-off revenue impact which is reflected in the Q1 guidance, what does Snowflake’s growth look like?

  1. RPO grew $796 million in a single quarter which was a 44% sequential increase!
  2. Total number of customers with TTM revenue > $1 million grew 24.32% sequentially to 184 and grew 139% YoY!
  3. And in the quarter, Snowflake added 14 new Fortune 500 customers and 21 Global 2000 customers.

Do you know any other companies delivering such strong growth in both large customers and contract values? Think about these numbers twice! As Bear and other board members shared before, it typically takes Snowflake half to even one year to collect meaningful revenue from new customers. How much revenue can these new customer drive after a year? That’s going to be huge!

The optimization will make YoY comparison tough for the next quarter and maybe for the entire FY23. But it won’t impact Snowflake’s hyper growth in the long run. And I’m sure that, after a one-off dip in Q1, the QoQ revenue growth will quickly recover in Q2. After that, Snowflake will just do as well as before if it does not do even better.

Cheers,
Luffy

36 Likes

Before talking about the positive side of the earning call, let’s pause here a little bit. Isn’t this very similar to the situation of Datadog in 20Q2? At that time, because of businesses cutting IT budgets, the customers of Datadog did a lot of optimization work to reduce the total number of machines they need for the cloud workloads. What’s the same between that optimization and Snowflake’s current one is that both optimizations are one-off and will only affect the QoQ revenue growth of the quarter when the optimization happens. And of course, both optimizations are beneficial for their customers, though Datadog’s was driven by its customs and Snowflake’s was driven by Snowflake.

In my opinion, this is a great observation and one I’m considering myself. These optimization costs might create a one-time step down in SNOW’s revenue but don’t necessarily change the slope of the overall trend. However, we now need to weigh management’s new 6 month lag in workload transfers with the existing 9-12 customer ramping lag. That does complicate things a bit.

In the DDOG example, that step down created headline growth headwinds until the slope came back around to the first affected quarter (87%, 68%, 61%, 56%, 51%, 67%). Kudos to the board for identifying that very early with DDOG. However, it’s fair to point out was also a period where DDOG’s stock languished while waiting for the trend to become proven and obvious to everyone else. I’m wondering just how much of a headline drag that might create for SNOW over the next year, particularly against the surprise Q3 we just had.

Each of us individually must decide what to do here, but I thought that was worth pointing out. Still mulling it over.

29 Likes

The difference IMO is that Datadog’s ding in revenue was from the onset of a global pandemic where their customers were doing the “optimizing”.

In Snowflake’s case it is efficiency the company is creating and passing on to customers. So how do we know they don’t continue to optimize further? Longterm, Snowflake will still probably grow much faster for much longer than anyone expects, but it could be slightly more lumpy than previously thought. It’s a bit of uncertainty.

3 Likes

I don’t see why for Snowflake passing the savings to the customer here is a problem.

  1. The cloud game has always been reducing the cost per byte for customers but making up in scale and new products.

AWS has reduced the price on its products 107 times since 2006

https://aws.amazon.com/blogs/aws-cost-management/amazon-ec2-…

  1. Snowflake is considered very expensive for large workloads, which has been a deciding factor on why large organizations choose Google BigQuery over Snowflake. I have posted before that GCP BigQuery is their biggest competitor in data warehouse (Databricks can work hand in hand with any data warehouse) so there is potential for them to capture large customers by moving the cost needle.

References:

Reducing revenue per customer AND growing total revenue wasn’t a problem for AWS. I don’t see it being a problem in the future for Snowflake either. The problem for investors, however, will be in the execution of their product roadmap and sales, not in the action of reducing cost itself.

25 Likes

In Snowflake’s case it is efficiency the company is creating and passing on to customers. So how do we know they don’t continue to optimize further? Longterm, Snowflake will still probably grow much faster for much longer than anyone expects, but it could be slightly more lumpy than previously thought. It’s a bit of uncertainty.

Agreed, and it’s an uncertainty that wasn’t there before. Yes, they are passing this on to customers with the expectation customers will shift their current spending mix toward the higher margins of increased workloads. Based on management’s comments, they have plenty of past instances to back this up. However, it’s entirely possible some customers will NOT increase workloads since they are already maxed out (or close to it). Those customers would likely spend those savings in a way that doesn’t benefit SNOW. In the meantime, not only is there no guarantee SNOW gets a 1:1 dollar switch with this move, but it’s also going to take multiple quarters to absorb the resulting revenue hit.

While DDOG’s step back was caused by unforeseen external factors, SNOW’s is a calculated adjustment to strategically trade some of its short term “is” for a longer term “could be”. In fact, I’d consider it the most significant operational adjustment we’ve seen from one of our companies this quarter. The question is exactly how that might change the thesis or timeline. Yes, knee jerk reactions are bad. However, so is failing to talk through underlying changes to our businesses when it’s obvious one has occurred. In my opinion, this one deserves more examination than ZS or MNDY simply missing a metric. Maybe it’s a big deal, maybe not. I’m still TBD on it myself.

28 Likes