The GizmoData Dispatch — June 2026 Edition

🦆 Two months of shipping, squeezed into one dispatch.

Jun 12, 2026

We skipped the May edition — not because nothing happened, but because everything happened. Between two major releases of GizmoSQL, a ground-up rewrite of our Power BI connector, two brand-new drivers for DuckDB’s new Quack protocol, a conference, a meetup, a webinar, a complete overhaul of gizmodata.com (you can now buy and trial GizmoSQL Enterprise entirely self-serve), and the boldest offer we’ve ever made — cut your warehouse bill in half, guaranteed, or you don’t pay — the newsletter had to wait its turn in the queue.

Speaking of queues — we shipped one of those too. More below. 👇

🆕 New here? What is GizmoSQL?

GizmoSQL is a high-performance SQL server built on DuckDB and Apache Arrow Flight SQL. Think of it as “DuckDB as a service”: DuckDB’s blazing-fast analytics, accessible over the network from any language, BI tool, or notebook — with enterprise-grade security, authentication, and observability on top. Run it in the cloud, on-prem, on your laptop, or (yes, really) on your iPhone.

💸 The Half-Bill Guarantee: Cut Your Warehouse Bill in Half — or You Don’t Pay

We put the boldest promise we’ve ever made right on the homepage, and we mean every word of it:

We’ll cut your data warehouse compute bill by at least 50% — guaranteed in writing — or the migration costs you nothing.

Here’s how it works:

We baseline your bill. Together we document your current Snowflake / Databricks / BigQuery compute run-rate for the in-scope workloads — in writing, so there’s no moving target.
We migrate you in 30 days. A done-with-you migration sprint, run by our engineers. You keep your tools: GizmoSQL speaks standard SQL plus Arrow Flight SQL, ADBC, and JDBC, and works with dbt, Iceberg, DuckLake, and S3/GCS/Azure — so your dashboards, notebooks, and queries come along. Keep your data in your own cloud, or let us host it.
Your bill drops 50%+. We measure the new run-rate against the baseline. If it isn’t at least 50% lower, you don’t pay for the sprint — and we keep tuning for free until it is.

Why are we so confident? Because the math is stacked in your favor: warehouses bill marked-up credits/DBUs/slots for always-on, overprovisioned clusters, while GizmoSQL runs the same analytic SQL on right-sized VMs with DuckDB + Arrow — pennies of raw compute per query, open formats, no lock-in tax. GizmoSQL Enterprise itself is typically under 2% of the bill it replaces.

Not ready for a migration? Start with a free Bill Teardown — we’ll analyze your current warehouse spend and show you exactly where the savings are.

👉 gizmodata.com/cut-your-warehouse-bill

📊 Power BI Connector v2.0: Rebuilt on ADBC

The headline release of June: we’ve rebuilt the GizmoSQL Power BI Connector from the ground up — replacing the ODBC backend with ADBC over Arrow Flight SQL.

Why does that matter? The v1.x connector funneled Arrow data through ODBC’s decades-old row-based API — every column converted, every row marshaled. v2.0 speaks Arrow end-to-end using the newer Power Query ADBC extensibility APIs, driving the Apache Arrow Flight SQL ADBC driver directly:

Native Arrow data path — no more ODBC row-at-a-time conversion between the server and Power BI
Full query folding in both Import and DirectQuery modes — filters, aggregates, date/time functions, casts, and text predicates all fold into native DuckDB SQL (right-click any step → “View Native Query” to see it)
Cross-table joins fold in DirectQuery — all tables from one connection share a single data-source identity, so joins execute server-side instead of in the mashup engine
OAuth browser SSO — sign in to Power BI with your IdP (via GizmoSQL Enterprise’s server-side OAuth), alongside the existing username/password and bearer-token options
One-click install — the MSI sets up everything Power BI Desktop needs

Grab it from the GitHub releases page (requires GizmoSQL server v1.23.0+).

💡 One note for v1.x users: table relationships are no longer auto-imported (a limitation of hand-rolled navigators in Power Query). Power BI’s built-in Autodetect picks them right up — the connector declares primary keys to make sure of it.

🦆 Quack, Quack: DuckDB’s New Client-Server Protocol (and Our New Drivers)

In May, DuckDB Labs announced Quack — a native client-server protocol for DuckDB, now built into DuckDB core as of v1.5.3. Some asked: isn’t remote DuckDB your whole thing? We say: welcome to the party — a rising tide lifts all ducks. 🛶

And the DuckDB team said something that made our month. From the acknowledgements of their launch post:

“We also want to thank Philip Moore from GizmoSQL / GizmoData, who has blazed this trail for us already and shown that client-server DuckDB is a very worthwhile thing.”

Thank you right back — being recognized by the DuckDB team as the project that proved out client-server DuckDB means a great deal to us.

We shared our full take on the blog: “Welcoming Quack: DuckDB’s New Client-Server Protocol — and Where GizmoSQL Goes From Here”. Short version: Quack validates the remote-DuckDB model we’ve been championing, and GizmoSQL’s Arrow Flight SQL architecture — with enterprise auth, instrumentation, and ecosystem integrations — is complementary, not competing.

But we didn’t stop at a blog post. One day later, we shipped the first third-party drivers for the Quack protocol:

🔌 quack-jdbc — a JDBC driver for quack:// (Maven Central · GitHub)

Connect any JVM tool — DBeaver, Spark, dbt, HikariCP — to a remote DuckDB server with a jdbc:quack://host:9494 URL
Full DatabaseMetaData support for IDE integration
Streaming cursors with typed primitive vectors, bulk-load via APPEND_REQUEST
Already attracting outside contributors — v0.2.0 includes a community-contributed pluggable HTTP transport

🏹 adbc-driver-quack — a Go ADBC driver with Python wheels (PyPI · GitHub)

Zero-copy Apache Arrow RecordBatches straight from a Quack server
Streaming reads, Decimal128, nested types, full GetObjects metadata
Bulk ingest with create/append/replace modes
Per-platform wheels for Linux (x64 + arm64), macOS, and Windows

Both are experimental/alpha — Quack itself is young — but if you want to talk to remote DuckDB from Java or Python today, we’ve got you covered.

🏢 GizmoSQL: The Enterprise Way to Run DuckDB

Quack’s arrival makes one thing official: DuckDB-as-a-server is now mainstream. Which raises the question we get asked most — “so why GizmoSQL?”

Because running DuckDB as a server and running DuckDB as enterprise infrastructure are two different jobs. GizmoSQL is the enterprise-grade DuckDB server: a secure, concurrent, horizontally-scalable server that exposes DuckDB over Apache Arrow Flight SQL — speaking Flight SQL, ADBC, and JDBC so your existing BI, ETL, and notebook tools connect immediately. On top of that engine, Enterprise adds the layer that production actually demands:

SSO/OAuth authentication against your IdP
Fine-grained, per-catalog permissions (now with IAM-style wildcards)
Full session instrumentation and auditing — every session, statement, and outcome, queryable in SQL
Automatic query profiling and statement queuing (more on both below 👇)
KILL SESSION, admin controls, and an LTS channel for change-averse environments
Runs on Linux, macOS, and Windows — plus Docker and Kubernetes, where it scales horizontally behind a shared DuckLake + PostgreSQL catalog

We’ve put this case on the site in depth — see Running DuckDB as a server and the new Enterprise features page. The short version: if DuckDB is becoming your analytics engine, GizmoSQL is how you run it like a database platform — with the security, governance, and observability your team already expects from one.

And the next two sections are Exhibits A and B.

🚦 New in GizmoSQL Enterprise: Statement Queuing

This one came straight from a customer request — and it’s a big one.

Statement Queuing (GizmoSQL Enterprise v1.27.0+) adds admission control to your GizmoSQL server. Instead of letting a burst of concurrent queries fight over memory and CPU, the server admits statements up to a configurable concurrency limit and queues the rest — in strict FIFO order, so no statement starves.

Runtime-tunable — adjust queue limits live with SET GLOBAL, no restart required
Fully observable — queued and cancelled statements flow into Enterprise instrumentation, so you can see exactly what waited, for how long
Plays nice with session management — KILL SESSION promptly cancels a session’s queued statements too
Zero overhead when off — a lock-free fast path keeps the hot path clean

If you’re running mixed or bursty workloads on a right-sized instance, this is the difference between graceful degradation and an OOM. It’s exactly the kind of feature that comes out of operating GizmoSQL with real customers in production — if you have a feature you need, talk to us. We ship fast. 😉

🔬 Also New in Enterprise: Automatic Query Profiling

Ever wished you could see the query plan for that slow query after it ran — without asking the user to re-run it with EXPLAIN ANALYZE? Now you can.

Automatic Query Profiling (GizmoSQL Enterprise v1.27.0+) captures DuckDB’s native per-query profile — the full operator tree with timings, cardinalities, and memory usage — into Enterprise instrumentation, automatically, for every query:

Opt-in at any scope — set it server-wide (--capture-query-profile), per-session, or live across the whole server with SET GLOBAL (admin only)
Two capture levels — standard (per-operator profile, negligible overhead on real workloads) or detailed (per-expression timing too)
Stored as DuckDB’s native profiling JSON in the sql_executions table, queryable through the execution_details view — and the foundation for a graphical query-plan viewer we have on the roadmap

Together with statement queuing’s new observability (queued statements, wait times, and cancellations all land in instrumentation too), this is the data backbone for an Oracle-style SQL-monitor view of everything in flight on your server.

🚀 GizmoSQL Server: v1.22 → v1.29 in Six Weeks

The server had a very productive spring. Highlights:

🟢 Introducing the LTS Channel (v1.25.0) GizmoSQL now ships on two parallel tracks: Stable (latest DuckDB) and LTS (pinned to DuckDB’s Long-Term-Support line). Same GizmoSQL features and API on both — they differ only in the bundled DuckDB version. Available everywhere: DockerHub, Homebrew, and GitHub releases. Docs here.

🐍 The official gizmosql Python library pip install gizmosql and launch a managed GizmoSQL server from Python — it auto-downloads the right binary, supports stable/LTS channels, and works beautifully as a pytest fixture or context manager. Perfect for integration tests and notebooks. Docs · PyPI.

Everything else, rapid-fire:

PostgreSQL instrumentation backend (Enterprise, v1.28) — write instrumentation to PostgreSQL, plus catalog-level logging and a cluster_id for fleet-wide observability
IAM-style wildcard catalog access (Enterprise, v1.27) — --catalog-access now supports AWS-style wildcard patterns for catalog-level permissions
Inline license keys (v1.29) — pass your Enterprise license via --license-key / GIZMOSQL_LICENSE_KEY, no file mounting required (great for K8s)
New gizmosql_settings() table function, --memory-limit flag, --session-log-level, gRPC keepalives, and a --storage-version option for DuckDB file-format compatibility
Hardened Docker images — multi-stage builds with attestations, clean Docker Scout reports
DuckDB v1.5.3 under the hood, and a new Quick Start guide to get from zero to queries in minutes

🛒 gizmodata.com: Buy, Trial, and Manage GizmoSQL Enterprise — Self-Serve

We rebuilt the website around one idea: you shouldn’t need a sales call to get started (though we’re happy to take one!).

💳 Buy online — GizmoSQL Enterprise licenses are now purchasable directly on gizmodata.com via Stripe checkout, with your license key minted and emailed automatically the moment payment clears
🆓 30-day free trial — instant, automated Enterprise trial licenses, signed with the same key as production licenses, so what you test is what you get
🔧 Self-service management — manage or cancel your subscription anytime through the Stripe Customer Portal
🔥 Hot June offer — there’s a limited-quantity promo code live on the site right now, with a real-time counter showing how many remain. When they’re gone, they’re gone: gizmodata.com/cut-your-warehouse-bill

While we were at it: a new dark-first design, a dedicated GizmoSQL Enterprise features page, and a dedicated home for the Half-Bill Guarantee (see above ☝️).

🏆 Benchmark Corner: #2 on ClickBench

GizmoSQL is currently the #2 hot-run result on ClickBench among comparable single-node systems on c-class hardware — ClickHouse’s own benchmark suite of 43 analytical queries. Not bad for “just DuckDB with a network stack.” 😏

We also forked ClickHouse’s CostBench — an open benchmark for comparing cloud data warehouse performance per dollar, because speed without a price tag is only half the story. GizmoSQL performs over 4x more cost-efficiently for the 1 Billion row benchmark than the next highest competitor - on a single node.

🦫 GizmoSQL Is Now an Official Built-In DBeaver Driver

This one deserves its own headline. As of DBeaver 26.1.0, GizmoSQL ships as an official driver in DBeaver’s built-in driver list — no manual JDBC driver download, no classpath wrangling, no custom URL templates. Open DBeaver, pick GizmoSQL from the driver list, enter your host, and you’re connected over Arrow Flight SQL.

DBeaver is one of the most widely used database IDEs in the world, so this is a big deal for first-touch experience: the distance between “I heard about GizmoSQL” and “I’m querying it” just dropped to about thirty seconds.

🌍 Community & Ecosystem

📱 Hoyt Emerson took GizmoSQL-on-iOS for a spin. Hoyt’s The Full Data Stack published “Running DuckDB on an iPhone” — a hands-on walkthrough of running a full GizmoSQL server on a phone, querying it remotely from a laptop, and watching the logs scroll by in real time. His take: “DuckDB is a technology that can run analytics on a phone with ease.” (The app is $0.99 on the App Store — cheaper than the coffee you’ll drink while benchmarking it.)

🗺️ qgizmosql keeps rolling. Our QGIS plugin for GizmoSQL is live on the official QGIS plugin repository — stream spatial layers from a remote GizmoSQL server straight into QGIS. v0.4.x brought catalog support, live integration tests, and our first community PRs.

🔧 Driver maintenance across the board: the JDBC driver picked up a smarter Connection.isValid() (v1.6.3), and the Python ADBC driver fixed INSERT/UPDATE/DELETE ... RETURNING result sets (v1.1.7).

🎤 Where We’ve Been: May Events Recap

Introducing GizmoSQL for Power BI Users (May 12, online) — Gustavo Leo of Fabricent and Philip Moore walked through connecting Power BI to GizmoSQL live, from install to interactive DirectQuery dashboards. Here’s the video:

IoT Tech Expo North America (May 18–19, San Jose McEnery Convention Center) — We exhibited at TechEx alongside 8,000+ professionals and 250+ speakers covering IoT, Edge AI, and digital twins — talking GizmoSQL, GizmoEdge, and DuckDB-powered analytics at the edge all day, both days.

MotherDuck + DuckDB May Meetup (May 21, San Francisco) — Philip presented “GizmoSQL: DuckDB, Anywhere You Want It” at the SF meetup hosted by MotherDuck and Greybeam, alongside talks from Jacob Matson and Kyle Cheung. Great crowd, great questions, great ducks. Here are the slides.

💰 Pricing Reminder

GizmoSQL (open core) is free — Apache-licensed and ready for production. GizmoSQL Enterprise — OAuth/SSO, instrumentation, statement queuing, query profiling, catalog permissions, and support — is now priced at $300 per vCPU per year (with a $1,500/year floor) — typically under 2% of the warehouse bill it replaces. Start a 30-day trial or buy online today, or go all-in with the Half-Bill Guarantee migration. Prefer invoicing and a human? Talk to sales.

🔭 Looking Ahead

GizmoData Cloud — our fully managed GizmoSQL service — is in private preview with active customers, and there’s now a “coming soon” section on the site. Want early access? Reply to this article.
More Quack protocol coverage as it matures in DuckDB core
Continued CostBench buildout — transparent performance-per-dollar comparisons across the warehouse landscape

Enjoying the Dispatch? Subscribe, share it with a data friend, and find everything we build at gizmodata.com and github.com/gizmodata.

Until next month — keep your queries fast and your bills small. 🦆

GizmoData

Discussion about this post

Ready for more?