Snowflake Summit 2023
It should go without saying, but nothing below is intended as investment advice of any kind. All the references below are from publicly disclosed news at Snowflake Summit, sources are in the footnotes.
Snowflake Summit occurred this past week. Among other announcements, Snowflake announced key features focused on a new partnership with Nvidia1, LLMs and machine learning in Snowflake via Snowpark Container Services2, and updates to the Native Application Framework3.
I had the privilege of presenting Cybersyn in the keynote. You can watch the Cybersyn section or watch the entire keynote here.
While there is no shortage of summaries on announcements4, I wanted point out some of the less glamorous announcements that may be useful to alternative data practitioners and data-as-a-service companies alike.
Pay with Capacity5: Snowflake announced that clients can pay for Marketplace Listings using their Snowflake capacity contracts. Customers can use their Snowflake budget to buy the vetted (from a security standpoint) datasets and applications available on the Marketplace. From a vendor perspective, this is a very interesting bundling proposition.
Custom Billing Events6: Custom Events allows a Native Application to bill users based on a custom event or custom data in a native application (which can be a Snowflake UDF or Streamlit dashboard, for example). For example, if you build a Snowflake UDF that properly formats a data field, you could charge based on the number of rows that UDF processes. Clearly, this allows for creative monetization of Snowflake functionality.
Data Governance and Privacy7: Snowflake announced the Private Preview of Aggregation Constraints and Projection Constraints. Users can be granted access to a raw data table but limited what columns they can query (projection constraint) or be mandated a minimum aggregation for any result (aggregation constraint). This is a large step forward for the Snowflake data clean room and a natural extension of Snowflake’s fine-grained permission system.
Budgets and Utilization8: Snowflake now allows you to set spending alerts and also calculate cost run-rates by project. This should be quite helpful in isolating projects (say data trials) and ensuring an alternative data group sees ROI.
Snowflake SQL: Snowflake SQL added a few new verbs, most notably MAX_BY, MIN_BY9, and GROUP BY ALL10. At the risk of going into the true minutiae, these verbs are particularly useful for anyone that works with point-in-time data. The Window-subquery pattern that otherwise achieves the MAX_BY, MIN_BY result is among the most common alternative data queries (so common, I used to use this as the first SQL interview question).
Snowflake Performance Index11: This is essentially a single summary of the speed/performance improvements introduced into the Snowflake engine, as tested on the average Snowflake customers’ actual workloads. If anyone has listened to Snowflake earnings calls, you will frequently hear a discussion about performance improvements reducing credit consumptions. This index quantifies this effect.
One somewhat unpolished but effective blog post I enjoyed is here:
I have not found a good written blog post or source summarizing this, watch the Snowflake recorded keynote for this announcement.