No. of Recommendations: 16
Didn't Berkshire just disinvest from it too?
Yes, Todd Combs purchased SNOW shares for his portfolio at Berkshire at the IPO price of $120/share. Held through ups and downs over several years. Sold all of it during Q2 for an average of $131.90 / share. A small "profit" but a bad investment outcome.
GIECO, under his watch, had hired the Snowflake guru Linda Apsley to be the CIO and refocus the company as a more data-centric competitor like Progressive, Capital One, etc. Apsley had run up big Snowflake bills at each of her past 3 jobs, starting at COF, Snowflake's biggest customer. She was let go from GEICO in July 2023 and GEICO has probably become a bit less enamored with Snowflake since that time.
Seems like something called "Databricks" now has GEICO's affections? Maybe that is a competitor to Snowflake, maybe not. It is all above my pay grade.
No. of Recommendations: 18
"Seems like something called "Databricks" now has GEICO's affections? Maybe that is a competitor to Snowflake, maybe not. It is all above my pay grade."
Databricks is up there with the largest private software companies. They started out commercializing the Apache Spark project. They are a big company, but can be summarized simply by saying they allow customers to perform analysis on streams of data. This is opposed to storing that data in a warehouse for later analysis. They compete with companies like Ververica (Apache Flink) and Confluent (Apache Kafka and Apache Flink), among others, but are generally considered the leader.
That might be useful for trivia for bad cocktail parties, but broadly speaking once you get to very large scales of data, data warehousing (or lakehousing) like what Snowflake offers, gets tough and expensive to store it and do post-hoc analysis later. So you might just analyze it without persisting, or perform some kind of filtering and/or aggregation before persisting it, to minimize what's persisted. You might also train machine learning models on the stream of data and use that as a form of persistence. And just looking at the data warehouse side, Apache Parquet (columnar file format) is becoming fashionable and allowing companies to roll solutions similar to Snowflake internally.
As a result Snowflake revenue isn't scaling with the generation / processing of data, and the stock is underperforming as those alternate approaches are more efficient and capture a greater share of spending.
I work the area, in a specialized subdomain, and don't work for any of the above companies. It's interesting to hear GEICO's moving in that direction.
