Open SourceEnterprise ReadyYour Cloud or On-Prem

A lower-cost alternative to Databricks.
Same SQL, terabytes in seconds.

Q: How much cheaper is Hardcharger than Databricks?

In our 1 TB workload derived from TPC-H, Hardcharger 2.0.0 completed the 22-query suite in 42.4 seconds for $0.09 of AWS compute. This is not a fresh head-to-head run against Databricks. Your total cost includes infrastructure, storage and any managed services; benchmark your own workloads.

Q: Can Hardcharger read my existing Delta Lake tables?

Yes. Hardcharger reads Delta, Iceberg, and Parquet directly on your object storage — no ingestion, no copies, no export step. Your lakehouse data stays exactly where it is.

Q: Can I keep my PySpark code when I move off Databricks?

Yes. The sqlframe-hardcharger adapter (built on SQLFrame) provides a PySpark-compatible DataFrame API over Hardcharger, and its activate mode keeps standard pyspark imports working unchanged — so existing PySpark jobs run against Hardcharger with little to no code change, and no Spark cluster underneath.

Q: Do I have to manage clusters?

No. Hardcharger is a single lightweight server process that starts in seconds on a VM, in Docker, or on Kubernetes. There is no cluster to size, warm up, or pay for while idle.

Q: Is Hardcharger enterprise-ready?

Hardcharger is Apache-2.0 with every feature included: access control, SSO, audit logging and the web UI. There are no editions, license keys or per-vCPU software fees. You pay for your infrastructure and any optional managed services or support.

Databricks is powerful, but DBUs and cluster spin-up quietly stack the bill. Hardcharger runs the same standard SQL at a fraction of the cost, in your own cloud or on-prem, reading Delta, Iceberg, and Parquet directly. No clusters to babysit, no DBU meter.

$0.09per TPC-H 1 TB run

< 43 sto query 1 TB

Apache-2.0all features included

View integrations

Apache-2.0. Every feature included. No license keys or per-vCPU software fees.

hardcharger-server (local development)

$ HARDCHARGER_ADMIN_PASSWORD=change-me hardcharger-server --database warehouse.duckdb
sql> SELECT sku, sum(qty) FROM sales GROUP BY sku;
reads Delta and Iceberg directly
no Hardcharger license fee; infrastructure billed separately

Derived from TPC-H: 22 queries at scale factor 1000 (1 TB)

Total seconds, lower is better. AWS r9gd.24xlarge (Graviton5, 96 vCPU, 742 GiB, 3 × 1.9 TB NVMe), AWS on-demand at $7.69/h. Measured October 20, 2026.

GizmoSQL 1.42.0 · DuckDB 1.5.677.0 s · $0.164
Hardcharger 2.0.0 · DuckDB 2.0.042.4 s · $0.090

Cost per full run: the run’s seconds at the machine’s AWS on-demand price; a cloud cost, not a TPC price/performance metric. Same machine type and data for every server.

This workload is derived from the TPC-H Benchmark and is not comparable to published TPC-H Benchmark results, as this implementation does not comply with all requirements of the TPC-H Benchmark. Deviations: only the 22 queries are run (no refresh functions, no power or throughput test), timed as the sum of per-query means over three runs, without an audit. TPC, TPC Benchmark and TPC-H are trademarks of the Transaction Processing Performance Council (TPC).

Read these with the numbers

Hardcharger uses DuckDB 2.0.0; the historical GizmoSQL baseline uses DuckDB 1.5.6. The performance difference includes changes in the engine and does not isolate the effect of the server protocol.
The Hardcharger client ran on the same machine over TLS (127.0.0.1). The answers are a few rows each, so the network adds little.
Cold first runs are included in the means (page cache dropped and server restarted between passes). The DuckDB 1.5.6 row is a historical baseline from an earlier run.
Query 11 uses its generator’s fixed fraction, not one scaled to the data size, so it returns no rows on every server.
Defaults only: 96 threads, DuckDB’s default memory limit, nothing tuned. A second full pass came within 2% of the first, and every Hardcharger query returned the expected row count.

Why teams move off Databricks

No DBU meter

Flat, predictable cost. You are not billed per DBU or per second of cluster time.

No clusters to babysit

A single lightweight process that starts in seconds. No spin-up wait, no idle-cluster charges.

Open formats, no lock-in

Reads Delta, Iceberg, and Parquet in place. Apache 2.0 core and standard SQL you can audit.

Using PySpark? We've got that covered.

Your Databricks jobs don’t have to be rewritten in SQL. The sqlframe-hardcharger adapter (built on SQLFrame) gives you a PySpark-compatible DataFrame API over Hardcharger — and its activate mode keeps your standard pyspark imports working as-is. Existing PySpark pipelines run against Hardcharger with little to no code change, and no Spark cluster underneath.

Get the PySpark adapter on PyPI →

python

$ pip install sqlframe-hardcharger
py> from sqlframe_hardcharger import HardchargerSession, functions as F
py> spark = HardchargerSession.builder.config(map={
    "hardcharger.uri": "hardcharger://localhost?tls=false",
    "hardcharger.username": "hardcharger", "hardcharger.password": "change-me"
 }).getOrCreate()
py> (spark.table("gold.orders")
        .groupBy("region").agg(F.sum("revenue")).show())
same PySpark DataFrame code — no Spark cluster

Up and running in three steps

Install Hardcharger, free

Native binaries, Docker, pip or Homebrew. Start with the install guide.

Point it at your data

DuckDB, Parquet, Iceberg, Delta, or DuckLake on your object storage. No ingestion, no copies.

Connect your tools

Connect Java applications over JDBC, notebooks over ADBC, and browsers over HTTP. Use dbt-hardcharger for dbt.

Hardcharger vs Databricks — FAQ

How much cheaper is Hardcharger than Databricks?