Snowflake CFO at GS conference

I was drawn to webcast from a SA headline quoting the CFO that Snowflake has enough GPUs, and wouldn’t be buying more until he had the revenue to support it. (The CFO is talking about on improving operating income, but it is unclear to me where this GPU cost is hitting. The “R&D” GPUs are expensed (or depreciated) and they have enough. Clear CFO emphasis here. (So income won’t further deteriorate for R&D.) But other customer-specific build-out GPUs are in COGS. And since the revenue is consumption-based, and part of multi-year deals for large enterprises, it seems like they have to carry a lot of it as inventory, and let it bleed out as RPO is recognized so as to not adversely hit current gross margin. This seems like a knife edge that is tricky to balance on … given how expensive and supply-constrained GPUs have been … and how aggressively they model consumption revenue.)

CFO is saying they “have enough GPUS” for the US but they are some supply constraints in other markets–but adding any more non-US GPUs won’t materially impact COGS.

From Goldman Sachs Communacopia + Technology Conference on Thursday: Link to webcast: Snowflake - 1685612

SA article: Snowflake sees traditional large enterprises as top customers; not investing more in GPUs | Seeking Alpha

He also fielded a query on the role of new initiatives such as Copilot, Cortex and Document AI, which are not in the guidance. Scarpelli said that when people ask what would Cortex and Iceberg, and other things, would do for the company next year, he says, “I need to see a few quarters of history before I can put that into our guidance.”

CFO Scarpelli is doing a bit of an introduction of new CEO Sridhar Ramaswamy. Emphasizing that they are working harder than ever, CEO is data driven, kind of more of an engineering or analytical type brain than a sales type. He’s making the culture more customer focused–trying to get the right products to market faster. By right, he means ones the customer wants.

On go-to-market, they’re a lot of metrics about how productive the sales engagement is. Focuses compensation on getting new logos. Shrinking the sales cycle.

Use of AI internally in Snowflake: data migration, data cleansing, quality assurance test scripts … very promising savings that we will show up in margin improvement (some day). Also want gen-AI to improve customer support, but not there yet.

Also on AI, the data that sits in their warehouse has tremendous (potential) value for example in training domain-specific LLMs, but no one knows exactly what the monetary value will be, and when it will hit.

My question for this webcast was: is the glass half full or half empty? You could interpret almost any comment the CFO made about current focus as an indictment of previous strategy/execution, or that Snowflake is on the cusp of something big. My other takeaway is on AI, and it’s the same that most everyone on this board talks about, AI hasn’t really been realized yet. It’s the “early innings.” All of Snowflake’s customers are working on ways to it incorporate AI either as productivity gains in their own operations, or to enhance their product offerings, but these aren’t real yet. Everyone seems to believe it’ll happen, and Snowflake will benefit from it when it does, as will many other companies, but it isn’t there yet.

I remain LONG Snowflake. I’ve been constantly waging internal debates about whether to exit or not. My conclusion remains the same–this isn’t a Saul stock–but I do see it for me as a long term hold and it will take off, I just don’t know when. The large enterprises, for example banks, that are signing 3 year deals and migrating from legacy systems, aren’t going back. And the AI applications will hit.

From Feb 24 press release announcing CEO change:

Since joining Snowflake in May 2023 in connection with the company’s acquisition of Neeva, the world’s first private AI powered search engine, Ramaswamy has been spearheading Snowflake’s AI strategy. He led the launch of Snowflake Cortex, Snowflake’s new fully managed service that makes AI simple and secure for all users to quickly drive business value. Prior to joining Snowflake, Ramaswamy co-founded Neeva in 2019.

rgds,
Bill

PS. I’m not a CFO type. Sorry for that COGS/Depreciation gobbledygook on GPUs up top. If some real CFO wants to weigh in and fix it, I’d appreciate it.

42 Likes