Agentic Dev Reshapes Databases

The software development lifecycle is undergoing a radical transformation. Large language models (LLMs) have unlocked agentic frameworks capable of analyzing requirements, writing code, executing tests, and deploying services at unprecedented speeds. This shift, detailed in a recent Databricks blog post, is fundamentally altering database requirements. As AI is Changing Software Development Lifecycles, the traditional approach is giving way to rapid evolution.

Companies working on this

StartupHub profiles of the companies this article names, with funding and a one-liner from our database.

Databricks

$190.0B

A unified data analytics and AI platform built on the lakehouse architecture.

DigitalOcean Holdings, Inc.

$1.1B

A cloud infrastructure provider offering scalable and developer-friendly computing services.

Tiger Data

$1.1B

Managed PostgreSQL database service designed for AI agents.

PhoenixAI

$90M

AI-native database for regulated industries.

Evolutionary Development

Agentic software development mirrors an evolutionary algorithm: generate, iterate, and evaluate. Applications can now be modified and redeployed in minutes, a stark contrast to slower, linear pre-LLM cycles. This rapid iteration requires databases that can branch instantly and cost-effectively, much like code repositories using Git.

Lakebase, Databricks's third-generation database architecture, addresses this with an O(1) metadata copy-on-write branching mechanism. This allows near-zero cost branching of database states alongside code, with compute costs only incurred during experimentation.

Cost Sensitivity and Elasticity

As the marginal cost of software development plummets, the value of individual applications decreases. Many agent-generated services are small, ephemeral, or serve highly bursty workloads. Infrastructure must support this at minimal marginal cost, eliminating high baseline price floors common in traditional databases.

The serverless, elastic nature of Lakebase directly tackles this. By decoupling compute from storage, it scales database compute based on load in sub-second time and scales down to zero when idle, eliminating cost floors.

Seamless Growth

Agent-driven development creates a vast number of small, ephemeral databases. The critical challenge is that developers, and the agents themselves, cannot predict which will become production systems. Database architecture must support seamless, elastic growth from tiny instances to massive capacity without manual re-platforming.

Lakebase is designed to handle this evolution natively, making instant scaling from near-zero to massive capacity a fundamental requirement.

Open Ecosystems are Essential

AI agents, trained on vast public code and documentation, operate most effectively within familiar open-source ecosystems. Databases like Postgres are deeply embedded in this training data, allowing agents to generate schemas, queries, and integrations reliably. Proprietary databases suffer from a lack of context.

This openness must extend beyond query interfaces to the storage layer. Lakebase stores data in standard, open Postgres page formats directly in the data lake, enabling native interaction without proprietary compute engine bottlenecks.

The shift is already underway; Databricks's Lakebase service sees AI agents creating roughly four times more databases than human users.

Databases designed for experimentation, openness, and elasticity are poised to lead the agentic era.