As GenAI applications transition from chat interfaces to autonomous agents, they introduce significant infrastructure challenges, especially for enterprises aiming to progress beyond proof-of-concept stages. At AWS re:Invent 2025, Cockroach Labs presented a session focused on building resilient GenAI applications using intelligent memory, emphasizing the importance of integrating context, tools, and memory at scale. The session highlighted that traditional enterprise systems, with 70% still operating on legacy infrastructure, struggle to meet the demands of modern GenAI applications, which require capabilities like Retrieval-Augmented Generation, agent interactions, and memory to provide advanced user experiences. A live demo showcased a banking assistant using CockroachDB, demonstrating its resilience and ability to continue operating seamlessly despite simulated cloud region failures. CockroachDB's distributed SQL database design, compatible with PostgreSQL, offers horizontal scaling, always-on availability, data locality, and distributed vector indexing, enabling GenAI applications to scale globally and maintain consistent performance. The session underscored the necessity of robust infrastructure for deploying GenAI applications, with CockroachDB providing a reliable foundation for such systems.