Snowflake's Instacart protestations hint at challenges for poster child of the data cloud
If customers can slash bills by 'optimizing,' what does that mean for revenue?
Opinion Snowflake should have been enjoying positive results at the end of last month. Revenue for the second for the quarter was $674.0 million, a 36 percent leap on the same period last year, albeit with an operating loss of $285.4 million, up from $207.7 million on Q2 2022.
But financial news from elsewhere emerged to spoil the party. The cloud-based data warehouse firm was forced to publish an explanation after reports suggested stellar client Instacart, which is currently launching its own Initial Public Offering, had slashed usage of Snowflake's software.
While Snowflake made a good point – Instacart's SEC S-1 note said its Snowflake bills had fallen, not usage per se – the problem with issuing clarifications of this nature is that they can raise more questions than they answer.
Snowflake rival Databricks – they both now pitch themselves as executing structured analytics in a data warehouse, and querying unstructured data in data lakes all on the same platform – was quick to jump on the opportunity, taking to social media to suggest Instacart was moving workloads onto Databricks.
Databricks CEO Ali Ghodsi had already posted on LinkedIn how the client was migrating, and Instacart revealed that the company had "modernized their architecture that was originally architected with Snowflake to be a modern Data Lakehouse to cut costs and enable more complex machine learning."
The original blog post has since been removed by Instacart, although it has been helpfully archived here.
In Snowflake's repost, it pointed to a presentation it claimed showed "How Instacart Optimized Snowflake Costs by 50 percent".
"Snowflake has partnered closely with Instacart to scale up to meet the company's massive demand growth, and then to optimize for efficiency. Optimizations are undertaken on a workload-by-workload basis, and have been extremely successful," its clarification blog post explained.
The problem is that showing how customers are able to "optimize" their Snowflake workloads can be read as admitting the company's deployment and charging models quickly led to unexpected and exorbitant bills.
The Register reported in 2021 that, according to insiders, companies found unpredictability of the costs incurred by Snowflake's pay-as-you-go model could result in a big shock to users trying to understand their current and future expenditures.
So prevalent was the perception that Snowflake could lead to uncontrolled costs that Teradata – which cloud-native proponents saw as an antiquated on-prem data warehouse provider – based its relaunch in part on its ability to control costs.
In 2021, CEO Steve McMillan told The Register: "If we look at cloud-native competitors, the way that they address performance at scale is to spin up new compute and spin up new storage. We make sure that you get the absolute best and predictable results, from both a query perspective and a cost perspective, that customers want to have in the cloud. We give a lot of control over the performance, functionality and cost, organisations experience in the cloud."
- Microsoft Fabric promises to tear into the enterprise analytics patchwork
- IBM bets $4.6 billion that cloud bills and IT sprawl will be enduring problems
- Get ready, Snowflakes: Azure AI is coming for you with one click
- Amazon CEO says AWS staff now spending 'much of their time' optimizing customers' clouds
In 2022, Snowflake said its approach to performance improvements and efficiencies had yielded a 20 percent reduction in the average cost of warehouse queries for customers over the last three years, seemingly aware of some criticism.
The company became something of a poster child with the movement to big data in the cloud, at least in the eyes of investors who valued Snowflake at an eye-watering $120 billion following its 2020 IPO.
The lingering question about the Instacart episode is on how much of this valuation is predicated on the kind of fees the likes of Instacart was prepared to pay – it was already a customer in 2020. Another way of putting it is that the way investors value Snowflake could be adjusted to the optimization, which seemingly allows big customers to slash their bills.
The process may have already started. Following its Q4 2022 results, the company saw 30 percent wiped off its value in after-hours trading after it lowered guidance on revenue forecasts. This was despite Snowflake revenue for Q4 FY22 ended 31 January hitting $383.8 million, effectively doubling in a year.
However you read the Instacart episode, it's hard not to conclude the sheen has gone from the one-time darling of the NY stock exchange which seemed to promise near-limitless growth. With Google and Microsoft beefing up their cloud data warehouse offerings, both tightly knitted to their own cloud infrastructure, Snowflake can rest assured there will be no easy dollar's revenue. ®