Subject: Re: SNOW - hold forever?
Isn't that a big barrier for Oracle customers to switch to SF so it's unlikely SF will catch up with Oracle?

I work around the space, and I don't get a lot of opportunities where BRK investment discussion and technology overlap.

Snowflake has a number of academic papers floating around, a few of which are quite good (PDFs, view at your own risk)
* https://event.cwi.nl/lsde/pape... "The Snowflake Elastic Data Warehouse" (slides from talk accompany paper here, https://15721.courses.cs.cmu.e...)
* https://www.usenix.org/system/... "Building An Elastic Query Engine on Disaggregated Storage"

Migration is tedious but it's one time in nature. In this case you do it to reduce long term costs, which Snowflake does in ways the papers describe. Those papers are accessible but fairly technical, so to summarize it, they optimize storage efficiency (compression in "blob storage", or s3-like systems), adapting established data warehousing techniques to work with that storage architecture, decoupling storage from compute resources, and highly cost efficient compute utilization. To summarize further, they support extremely big data in generic cloud resources very efficiently.

The difficulty in chipping away at ORCL is the depth of integration into things like CRM, ERP, financials, ERM, and so on. I have the impression that the CRM/ERP/et al. data may reside in those system but gets replicated into SNOW and that SNOW runs a lot more data analysis oriented workloads, but I may be wrong. No need to scale up expensive Oracle systems to handle unrelated analytic workloads, and risk license noncompliance or audits, right?

I don't know why BRK has the SNOW position, out of all of their tech adjacent positions. It is a very high valuation and high risk (from disruption, many competitors can emerge in 10-15 years to grow into 20 P/S multiple. But I'd guess it came from the insurance operations, and from analyzing the future market for data, the structural cost advantage, and stickiness or cost to moving data out once they acquired a customer.