
As enterprises race to deploy AI agents at scale, much of the public conversation fixates on GPUs and other AI chips and large language models. Yet, according to G2 Krishnamurthi, VP and GM of databases at AWS, the real bottleneck to turning agentic AI into tangible business value often lies one layer deeper: the database.
In this episode of Targeting AI, co-hosts Shaun Sutner and Esther Shittu sit down with Krishnamurthi, who brings nearly eight years at Meta - most recently as head of AI infrastructure - and 16 years at Microsoft, to his current post leading non-relational databases at AWS.
The conversation explores why AWS has built a sprawling portfolio of specialized database engines rather than betting on a single general-purpose system, and how that philosophy is being tested and extended as AI agents become a dominant workload. Krishnamurthi explains how data governance, accessibility, and proximity to where data already resides determine whether AI agents can act on trustworthy information.
In this episode, we discuss:
- The database layer, not just GPUs or models, is the most underappreciated piece of the AI infrastructure stack, since it determines whether agents can access accurate, well-governed data.
- AI agents need data that is fast, simple to access, and located close to where it already lives - minimizing friction between storage and intelligent action.
- Relational databases remain the most widely adopted category, with open source engines such as MySQL and Postgres powering many modern applications.
- AWS built DynamoDB to meet internet-scale businesses' need for elastic, horizontally scalable database infrastructure.
- DocumentDB emerged because builders adopting JSON-based document models needed a database service that natively understood that data structure.
- Just as software engineers choose different data structures for different problems, organizations need different databases tailored to different application scenarios.
- AWS is layering AI-specific capabilities, such as vector indexing and MCP server access, onto its existing databases while preserving consistent governance across all of them.
To further explore topics about AWS, databases and AI agents, check out TechTarget AI, and please subscribe to our newsletter to keep up to date on the most important AI news.
To watch video clips from our podcast, subscribe to our YouTube channel, @EyeonTech.
References:
AWS latest to introduce context layer for agentic AI
Otros episodios de "Targeting AI"



No te pierdas ningún episodio de “Targeting AI”. Síguelo en la aplicación gratuita de GetPodcast.








