A theory that cloud vendors will focus on the lowest layer of the infrastructure stack and startups will take over the software layer.

Erik Bernhardsson

A theory that cloud vendors (AWS, Azure, GCP) will increasingly retreat to the lowest infrastructure layer — leasing raw compute capacity — while pure-software startups take over the higher layers with databases, runtimes, and developer tooling. The Redshift vs. Snowflake story illustrates the dynamic: despite AWS's scale advantages, Snowflake grew to $100B+ by building a better developer experience and running across all clouds. Financial analysis suggests AWS may actually benefit from this arrangement, effectively outsourcing R&D and sales costs to startups while still collecting infrastructure revenue. Predictions for 2030 include most engineers never interacting directly with cloud vendors, database markets dominated by cloud-agnostic vendors, Kubernetes fading like Hadoop, and dramatically improved resource utilization through serverless runtimes.

Storm in the stratosphere: how the cloud will be reshuffled