AI build follows cloud story pattern: data centers/compute first, then foundation models, then applications. Trillions in data center spend implies $45 trillion in software value over next several years.

First-gen cloud/SaaS is oversold but not obsolete. Separation coming between AI beneficiaries and non-transitioners. ServiceNow, Aviva interesting; Cloudflare, AppLovin, Snowflake may be overvalued; Databricks, ClickHouse more interesting for next cycle.

Next wave of AI-native application companies will emerge beyond foundation models. Foundation models (OpenAI, Anthropic, Google) are Wave 1, but won't dominate all software. New companies being built now, with some transitional companies possibly IPOing by end of 2026, and AI-native applications bursting on scene in 2027.