About this formationWhat used to be a category about abstracting away servers has quietly split into two distinct businesses: platforms that manage infrastructure for humans, like Brainboard's visual IaC design or Omnistrate's docker-to-SaaS conversion, and platforms that manage infrastructure for AI agents themselves, like Ravion's agent-native AWS provisioning and Cedana's live migration for GPU workloads mid-inference. The batch progression makes the shift visible: the Winter 2022 cohort (Jamsocket, Brainboard, Strong Compute) was still solving classic developer-experience problems, while by Winter 2023, Cakework and infra.new were explicitly building for autonomous or semi-autonomous deployment. This formation's 3.10x growth sits almost exactly alongside AI Compute Infrastructure's 3.07x, suggesting the two are effectively growing together rather than as separate trends, even though cloud computing as a category predates the current AI wave by two decades. The more surprising gap is against Developer Tools, growing at just 0.48x, which implies that generalized coding productivity tooling is stalling out precisely as the underlying compute layer it depends on accelerates. Cedana's claim of 80% cost savings through live workload migration only makes sense as a business now that GPU scarcity and inference costs, not storage or CPU cycles, are the dominant expense, which is a direct consequence of the compute demands unlocked by large model deployment since 2023. Companies like Beam and Cakework, both explicitly selling serverless infrastructure for AI inference and agents rather than general web apps, are the clearest evidence that "cloud computing" in 2022 and "cloud computing" now describe two different customers entirely.
Cambrian Signal, “Cloud Computing,” publication yc-2026-10-02-semantic-v8, snapshot October 2, 2026.
cambrian.jafar.com/formation/cloud-computing