Neon

A serverless Postgres platform that separates storage and compute, with instant database branching for dev/preview environments and scale-to-zero pricing.

Visit Neon →

Some links may be affiliate links — we may earn a commission if you sign up, at no extra cost to you. It never affects our rankings. Learn more.

Neon is a managed serverless PostgreSQL platform that separates storage from compute to enable autoscaling, scale-to-zero, and copy-on-write database branching. It runs standard Postgres, so existing drivers, ORMs, and tooling work unchanged, while adding a developer workflow where branches can be created per pull request, environment, or CI run. Neon was acquired by Databricks in 2025 (announced May 2025) and is positioned for both application developers and AI-agent-driven workloads.

Key features

Copy-on-write database branching that spins up near-instant, isolated branches from the parent's data
Autoscaling of compute between a configured minimum and maximum number of compute units
Scale-to-zero: compute automatically suspends after a period of inactivity and is not billed while suspended
Architecture that separates storage from compute, allowing each to scale independently
Built-in connection pooling via PgBouncer for handling high concurrent connection counts
Point-in-time restore within a configurable history window, plus read replicas

Specs

Category	Managed database
Engine	Serverless PostgreSQL
Pricing model	usage-based
Free tier	Yes
Regions	multi-region
Self-hostable	No
Open source	Yes

Last reviewed 2026-07-16 — see our methodology. Specs, pricing and free tiers change often, so confirm current details on the provider before deciding.

Pros & cons

Our editorial assessment — not vendor-stated facts.

Pros

Standard Postgres compatibility means existing drivers, ORMs, and tooling work without modification
Branching makes ephemeral dev, test, and preview databases cheap and fast to create
Scale-to-zero and usage metering can lower cost for intermittent or spiky workloads
Fully managed, with no servers, storage provisioning, or OS to maintain
Configurable autoscaling ceilings per branch give a predictable upper bound on compute cost

Cons

Scale-to-zero adds cold-start latency (typically a few hundred milliseconds up to around a second) on the first query after an idle suspend
Restricted extension support: extensions that require persistent filesystem access or background worker processes are unavailable
No OS-level access: you cannot SSH to the host, reach the filesystem, or run sidecar/agent processes
The pooled (PgBouncer) connection uses transaction-mode pooling, which limits Postgres features that rely on persistent session state; a direct connection is available but does not pool
Usage-based pricing can be less predictable and less cost-effective for always-on, steady workloads than fixed provisioning

Pricing

Consumption-based, pay-as-you-go pricing metered hourly (compute in compute-unit-hours, storage in GB-months) with no monthly minimum. A permanent free tier is available with no credit card required, alongside paid Launch and Scale plans that raise usage limits and add features such as higher autoscaling ceilings, longer point-in-time restore history windows, private networking, and an uptime SLA. An enterprise/custom tier and an AI-agent-focused plan are also offered.

Who should use Neon?

Developers who want managed Postgres with elastic scaling and per-branch database environments for feature development, previews, and CI, as well as teams building serverless, multi-tenant, or AI-agent applications with variable or bursty traffic. It is less suited to always-on workloads with steady load, or to teams needing deep OS-level control and unrestricted extension support.

Frequently asked questions

Does Neon have a free tier?: Yes.
How much does Neon cost?: Consumption-based, pay-as-you-go pricing metered hourly (compute in compute-unit-hours, storage in GB-months) with no monthly minimum. A permanent free tier is available with no credit card required, alongside paid Launch and Scale plans that raise usage limits and add features such as higher autoscaling ceilings, longer point-in-time restore history windows, private networking, and an uptime SLA. An enterprise/custom tier and an AI-agent-focused plan are also offered.
Who should use Neon?: Developers who want managed Postgres with elastic scaling and per-branch database environments for feature development, previews, and CI, as well as teams building serverless, multi-tenant, or AI-agent applications with variable or bursty traffic. It is less suited to always-on workloads with steady load, or to teams needing deep OS-level control and unrestricted extension support.