GizmoData
GizmoData offers Hardcharger, an open-source SQL server built on DuckDB 2.0 that delivers warehouse-grade analytics at a fraction of traditional cloud data warehouse costs. The server runs over plain HTTP with Apache Arrow results, includes built-in access control, SSO, and a web UI, and processes a 1 TB TPC-H-derived workload in 42.4 seconds for about $0.09 of AWS on-demand compute. GizmoData Cloud provides a fully managed deployment option on AWS, Google Cloud, and Azure starting at $29/month.
- Data & Analytics
- Enterprise Software
- Developer Tools
- Software Only
- Artificial Intelligence
- AI Agents
Funding
Founders
Founder details are not available yet.
Product
Problem
Traditional cloud data warehouses like Snowflake, Databricks, and BigQuery impose high compute costs that scale with query volume, making analytics expensive for organizations processing large datasets. These platforms also create vendor lock-in through proprietary storage formats and require moving data outside a company's own infrastructure, raising security and compliance concerns.
Solution
GizmoData provides Hardcharger, an open-source server for DuckDB 2.0 that delivers warehouse-grade analytics at up to 90% lower cost than traditional platforms. The server runs SQL over plain HTTP with Apache Arrow results, processing a 1 TB TPC-H-derived workload in 42.4 seconds for approximately $0.09 of AWS on-demand compute on a single machine. Customers can self-host Hardcharger in their own cloud environment, keeping data and keys within their infrastructure, or opt for a fully managed GizmoData Cloud service on AWS, Google Cloud, and Azure. The platform includes built-in access control, SSO, mTLS, row policies, and a web UI in the binary, with every feature included in the free Apache-2.0 build. Hardcharger supports concurrent multi-agent AI workloads, allowing multiple LLM agents to share a single lightweight SQL server, and integrates with open table formats like Parquet, Iceberg, Delta Lake, and DuckLake.
Target Audience
Primary customers are data engineering and analytics teams at mid-to-large enterprises seeking to reduce cloud data warehouse costs while maintaining performance, as well as AI/ML teams needing shared SQL infrastructure for multi-agent workloads.
Features
- Powered by DuckDB 2.0, processing a 1 TB TPC-H-derived workload in 42.4 seconds on a single Graviton5 machine, about $0.09 of AWS on-demand compute per run
- Open HTTP protocol (Chargewire) with Apache Arrow columnar results, streaming data transfers, and drivers for Python, Go, Java, and JavaScript
- Built-in security including users, roles, row policies, column masking, SSO (OIDC/JWKS), mTLS with client certificate identity, and audit logging
- all in the free Apache-2.0 build
- Session management with --max-sessions caps and --session-idle-timeout eviction that only resets on actual user SQL activity including result streaming, preventing idle connections from consuming memory
- Horizontal scaling via Kubernetes with round-robin session load balancing across nodes sharing a DuckLake catalog, with each query running entirely on one node
- Asynchronous result pipeline with lz4_frame or zstd compression on a pool of threads, large result partitioning for parallel reads, and parallel uploads committed as one transaction
- Native builds for Linux, macOS, Windows, iOS, Docker, and Kubernetes, plus a desktop app for macOS and Windows with graphical dashboard and system tray operation
- Open-source Apache-2.0 core with no proprietary storage, using open table formats such as Parquet, Iceberg, Delta Lake, and DuckLake to prevent vendor lock-in