Snowflake launched production AI capabilities at BUILD London 2026, including Cortex AI Functions, general availability of Notebooks, an Online Feature Store, and Agent Evaluations tooling. AWS, Google Cloud and Azure responded with near-identical agent orchestration features within days.
The platforms now offer the same stack: managed model hosting, feature stores, evaluation frameworks, and agent development environments. This convergence mirrors the 2010-2015 database wars, when cloud vendors bundled analytics tools to lock in customers.
Global financial institutions are testing these systems in production. HSBC partnered with Paris-based Mistral AI to deploy models across operations. Wells Fargo launched Google Agentspace internally. BNP Paribas signed a Mistral AI integration deal. Citigroup built Stylus, an internal AI platform using cloud primitives.
Banking adoption signals enterprise readiness beyond the U.S. market. Financial services requirements—regulatory compliance across jurisdictions, audit trails, security controls—exceed most industries. If these platforms handle multinational bank workloads, they can likely support other sectors.
Snowflake's Online Feature Store targets real-time inference, a capability AWS and Google already offer. The Agent Evaluations framework addresses the production gap: most enterprise AI projects fail because companies lack testing infrastructure for non-deterministic systems.
Competition is driving rapid feature parity. Tools announced this quarter will be standard by Q3 2026. Vendors are betting on ecosystem lock-in rather than technical differentiation. Companies standardizing on Snowflake's Cortex or AWS Bedrock face rising switching costs as they build workflows on vendor-specific APIs.
The winner will likely emerge from enterprise deployment volume, making early banking contracts strategically critical. Google and Microsoft are leveraging existing global enterprise relationships to drive AI platform adoption against cloud-native competitors.