$Snowflake(SNOW.US)

SNOW's business is strong, but the financial translation is difficult.

Institutions' favorite companies have one characteristic: the business logic can be directly translated into a financial model.

The monetization path for NVIDIA is clear: GPU shortage → Data center revenue soars → High gross margin → EPS revisions upward.

Microsoft's narrative isn't as strong as NVIDIA's, but it's still solid. Azure growth → M365 Copilot price increases/penetration → backlog growth → stable upward EPS revisions.

Now let's look at Snowflake's narrative chain: Enterprise Data Cloud → Data Governance → AI Access Layer → workload increase → consumption growth → revenue recognition → margin / FCF.

Too many intermediate steps. Customers using Snowflake doesn't mean they'll spend more immediately; AI accessing Snowflake also doesn't guarantee Snowflake captures the most value.

So institutions ask: how much does AI increase Snowflake's usage? Is this usage a one-time migration or a sustained consumption? Will customers optimize query costs, thereby reducing consumption? Will Databricks / Microsoft Fabric / BigQuery take away the new workload?

This makes it hard for SNOW to form a single, straightforward buy thesis like NVDA.

SNOW isn't a traditional SaaS company that primarily charges per seat; it's more of a pay-as-you-go model.

When customer data volume and query volume increase, revenue can grow quickly.

But there's a downside: if customers optimize costs, reduce queries, or compress workloads, revenue will also be affected. Institutions are wary of this model because, unlike subscription SaaS, it's not easy to predict NRR, ARR, and renewal rates.

SNOW often becomes a quarterly execution play:

Strong product revenue guidance this quarter, stock price rises;

Management says customer optimization continues, stock price falls;

AI workload has potential, stock price rises;

Consumption doesn't accelerate significantly, stock price falls.