MotherDuck pricing changes in 2026: the $25 plan is gone, where to host DuckDB now

If you went to renew a MotherDuck plan in the last few weeks and noticed the $25 per month tier was missing, you are not imagining it. Somewhere between December 2025 and February 2026 MotherDuck rebuilt its pricing page. The Lite plan was restructured rather than removed: the $25 per month minimum was replaced by a free starter tier with included usage and pay-as-you-go overages on top, so the Lite offering now starts at $0 per month and bills by consumption above the included envelope. The Business plan, meanwhile, went from $100 per month to $250 per month. The new free tier is real, but it is heavily capped: 3 users, 10 GB storage, 10 compute hours per month, community support only.

No blog post, no email, no announcement. The pricing page just changed. This post is the rundown of what changed, when, who it affects, and where to host DuckDB if the new tiering does not fit your team or your budget.

If you just want the punchline: Layerbase Cloud hosts plain DuckDB behind a PostgreSQL wire protocol. Free dev tier, no compute-unit math, no proprietary SQL extensions, no Pulse-to-Giga instance tier juggling. Skip to where to go from here if that is what you came for.

Contents

• What MotherDuck used to cost
• What MotherDuck costs now
• The new free tier in practice
• Who gets squeezed
• Why this happened
• Where to host DuckDB if MotherDuck is too expensive
• What you do not give up by leaving MotherDuck
• Migration notes
• Where to go from here

What MotherDuck used to cost

The pricing timeline, captured from MotherDuck's own pricing page archive:

The two big shifts:

1. The $25 plan is gone. A small team that wanted a paid tier with a real SLA used to pay $25 per month. The cheapest paid tier is now $250 per month, a 10x jump from the price floor.
2. The Business plan jumped 2.5x. From $100 per month to $250 per month, with no compute or storage included for the extra $150 (compute is metered on top, starting at $0.60 per hour on Pulse and going up to $36 per hour on Giga).

Storage prices did come down (from $0.08/GB/month to $0.04/GB/month), so heavy-storage workloads recover some of the difference. Compute-heavy workloads see the full price increase.

What MotherDuck costs now

The current plan structure, taken from motherduck.com/product/pricing:

Lite (Free)

• Up to 3 internal active users and 2 service accounts
• 10 GB storage included
• 10 hours of Pulse-tier compute per month
• Community support only

Business ($250/month per organisation, plus usage)

• Up to 10 internal active users, unlimited service accounts
• Access to all 5 instance types (Pulse, Standard, Pro, Mega, Giga) with read-scaling replicas
• 90-day snapshot retention
• Query history and "expert support"
• 99.9% availability SLA

Enterprise (custom)

• Unlimited internal active users and service accounts
• Fixed-cost capacity pricing
• AWS PrivateLink and HIPAA BAA included
• Same feature set as Business otherwise

Compute pricing on top of the platform fee runs from $0.60 per hour (Pulse) to $36.00 per hour (Giga), billed per second. Storage is $0.04 per GB per month across all tiers.

The new free tier in practice

The free tier exists, but the practical limits are:

• 3 users. That is not enough for most teams that would otherwise be a paying customer. A four-person data team has to upgrade to Business at $250/month or stay below the headcount.
• 10 GB of storage. Workable for a hobby project. Tight for any production workload, even small ones. The CSV-and-Parquet datasets that make DuckDB shine are often larger than 10 GB compressed.
• 10 compute hours per month on Pulse only. Pulse is the smallest instance type. Heavier workloads either run slowly or you blow through the 10-hour budget in days. There is no overage option, you simply stop computing.
• Community support. Fine for hobby use, hard to defend on a production system.

The free tier is best understood as a developer trial rather than a small-team home. If you are evaluating MotherDuck for a project that will eventually pay, the free tier shows you what it does. If you were running a real workload on the old $25 Lite plan and looking to keep the lights on for $25 to $50 per month, the gap to the new $250 Business plan is large.

Who gets squeezed

In rough order of impact:

1. Small data teams that were on the old $25 Lite plan. They have to choose between dropping to a free tier that cannot host their workload or jumping 10x to the Business plan.
2. Solo data engineers and analysts running side projects. $25 per month for managed DuckDB was a reasonable hobby budget. $250 per month is not.
3. Early-stage startups using MotherDuck as their warehouse. $250 per month as a platform fee (before compute and storage) plus per-hour compute charges adds up faster than the old pricing did. A workload that cost $50 to $100 per month in 2024 likely costs $300 to $500 per month under the new pricing.
4. Anyone weighing MotherDuck against open-source DuckDB plus an S3 bucket. The economic case for managed shrinks as the price floor rises.

Larger customers buying capacity-priced Enterprise contracts are probably unaffected or even better off (the storage price came down). The squeeze is at the small end.

Why this happened

There was no public statement from MotherDuck, so the rest of this section is informed speculation rather than reporting.

The broader market context is the venture-capital pivot from growth-mode pricing to sustainability-mode pricing. A lot of cloud data platforms (Snowflake, Databricks, Confluent, and similar) priced for growth in 2022 to 2024 and have been quietly raising prices in 2025 and 2026 as runway calculations got tighter. MotherDuck appears to be following that pattern: cap the free tier, move the price floor up, target larger customers who can absorb $250 per month without thinking about it.

This is a legitimate business decision. It is also a decision that pushes a lot of MotherDuck's early adopters out of the product, because the people who happily paid $25 per month for managed DuckDB are not the same people who happily pay $250 per month.

Where to host DuckDB if MotherDuck is too expensive

The DuckDB hosting market is small, but it exists.

Layerbase Cloud

Plain Apache DuckDB behind a PostgreSQL wire protocol. The free dev tier is free in perpetuity (no expiry, no card), with no per-organisation user cap, and the storage envelope is generous enough for real workloads rather than evaluation. Paid add-ons are itemised (keep-warm to disable scale-to-zero, larger storage, scheduled backups, region pinning); the engine itself stays free.

psql "postgresql://layerbase:password@your-host.cloud.layerbase.dev/duck1?sslmode=require&sslnegotiation=direct"

Because the protocol is PostgreSQL-wire, every existing pg client works without modification: psql, DBeaver, TablePlus, the pg npm package, Drizzle's Postgres adapter, Prisma's Postgres adapter, Python's psycopg2, the Go pgx driver. You point them at the connection string and queries run through DuckDB.

The trade-off versus MotherDuck: you do not get MotherDuck's hybrid execution (the cleverness of automatically splitting work between local and cloud DuckDB based on where the data lives). If hybrid execution is the reason you picked MotherDuck, Layerbase is not a substitute. If you just wanted managed DuckDB without proprietary extensions and a reasonable price floor, it fits.

Self-host on a small VM

DuckDB itself does not have a server, so "self-host" means running a Postgres-wire proxy like duckgres in front of a DuckDB file. You get a $5 to $10 per month Hetzner or DigitalOcean droplet, install DuckDB, run duckgres, expose the port. You are now responsible for TLS, backups, restarts, and disk pressure.

For a personal workload this is fine. For anything you depend on, the managed option saves you operational time worth more than the price difference.

SpinDB on your laptop

If your workload is a single-developer analytical workflow that happens to need DuckDB as a network service (not embedded), SpinDB installs and runs DuckDB locally with the same Postgres-wire protocol that Layerbase Cloud uses. Free, fast, and the queries run on whatever CPU and disk your laptop has rather than on a metered cloud instance.

pnpm add -g spindb
spindb create duck1 -e duckdb --start
spindb url duck1

For analysts who used MotherDuck mainly to avoid Python environment setup, SpinDB scratches the same itch without the platform fee.

Embedded DuckDB

DuckDB was designed to be embedded. If your workload is a single application or pipeline, pip install duckdb or pnpm add duckdb and skip the managed layer entirely. No server, no connection string, no monthly fee. This is the right call more often than people realise.

import duckdb from 'duckdb'

const db = new duckdb.Database(':memory:')
db.all('SELECT * FROM read_parquet(?)', 's3://bucket/data.parquet', (e, r) =>
  console.log(r),
)

For more on when each option fits, MotherDuck alternatives: managed DuckDB without lock-in goes through the trade-offs in detail.

What you do not give up by leaving MotherDuck

A few things people worry about that turn out not to be problems:

1. Your DuckDB files are portable. A .duckdb file is the storage format. Move the file, point a new DuckDB at it, you are back in business. There is no migration tool to run, no schema conversion, no export step.
2. CSV and Parquet reading works the same. SELECT * FROM read_csv_auto(...) and SELECT * FROM read_parquet(...) are vanilla DuckDB, not MotherDuck features. They work on any DuckDB.
3. The S3 extensions work the same. Reading from S3 with httpfs, IAM role-based auth, the works. All upstream DuckDB.
4. Your SQL mostly transfers. DuckDB SQL is a generous dialect. The parts of your queries that are not MotherDuck-specific (which is most of them, for most people) run unchanged.

The parts that do not transfer:

1. Hybrid execution. Queries that explicitly route between local and cloud DuckDB. Rewrite as either pure-local or pure-cloud.
2. MotherDuck-specific SQL extensions. A handful of statements like CREATE DATABASE FROM 'md:...' are MotherDuck-only.
3. Shared databases. MotherDuck's collaborative shared-database feature is proprietary.

For most workloads, the porting effort is minimal.

Migration notes

If you are moving off MotherDuck, the high-level steps are:

1. Export your databases. From MotherDuck, EXPORT DATABASE writes a directory of Parquet files plus a load script. The result is a self-contained, portable snapshot.

EXPORT DATABASE 'my_export' (FORMAT PARQUET);

2. Import on the new host. Against a Layerbase Cloud DuckDB or any other DuckDB, IMPORT DATABASE loads the directory.

IMPORT DATABASE 'my_export';

3. Audit your queries for the MotherDuck-specific statements listed in the previous section. Replace them with vanilla DuckDB or with calls into your application code.

4. Update connection strings. The Postgres-wire endpoint that Layerbase Cloud and SpinDB use looks like a regular Postgres URL. Most ORMs and clients need no other changes.

5. Validate row counts and a few representative queries before you cut over.

The whole thing is faster than most people expect, because DuckDB's storage format is a single file you can move around.

Where to go from here

• Spin up a free DuckDB right now: Layerbase Cloud, DuckDB engine. No credit card, no expiry, no per-organisation user cap.
• Compare options for managed DuckDB: MotherDuck alternatives: managed DuckDB without lock-in.
• Learn DuckDB end-to-end: Getting started with DuckDB.
• Compare DuckDB against ClickHouse for production analytics: ClickHouse vs DuckDB.
• Run DuckDB locally without Docker: SpinDB.

MotherDuck is still a fine product if hybrid execution is the reason you picked it. The team is real, the technology works, and for large customers who can absorb $250 per month as a platform fee the new pricing is reasonable. For small teams, solo developers, side projects, and early-stage startups who were comfortable with the old $25 floor, the math has changed enough that "managed DuckDB on Layerbase Cloud" or "DuckDB embedded in your app" is now the better answer for most workloads.

Create a free DuckDB on Layerbase Cloud and have a working endpoint in under a minute.