Fivetran + dbt Labs is announcing the general availability of dbt v2 and dbt State, delivering new levels of speed and cost optimization, along with introducing Fivetran Context Layer, new dbt Wizard experiences, dbt Charts, and an open lakehouse vision for greater flexibility across storage and compute.
According to the company, Fivetran + dbt Labs is advancing its vision for Open Data Infrastructure: a vendor-neutral, interoperable architecture that lets organizations independently choose and evolve their storage, compute, data movement, transformation, and visualization technologies at every layer. This enables AI systems to work across platforms using data and context the enterprise owns, not a single vendor.
"Every model our customers have built, every test they've written, every metric they've defined already captures the context AI agents need to do meaningful work," said Anjan Kundavaram, chief product officer, Fivetran + dbt Labs. "What we're delivering now is the open infrastructure to put that context to work across systems, while giving organizations the freedom to choose how their data is stored, moved, transformed and used as AI evolves."
dbt v2 is a full Rust rewrite of the dbt engine built for the scale that teams run at today and for how agents write SQL. It parses a 10,000-model project up to 10x faster than v1 and gives teams and their agents accurate real-time feedback, surfacing errors, column checks, and lineage before anything runs. With this release, the two engine era of Core and Fusion ends.
Now, dbt is one engine with two versions: dbt Core v1, the python implementation, is dbt v1. Fusion, the Rust implementation, has become dbt v2. Both versions remain Apache 2.0-licensed and security-supported.
dbt State determines what has changed by checking warehouse metadata and model SQL, then builds, skips, clones or defers each run accordingly. This simplifies orchestration and allows engineers to iterate faster without complex development rituals, while reducing unnecessary warehouse compute.
Fivetran's Managed Data Lake Service, already generally available, organizes, structures and maintains data as managed Apache Iceberg tables in customers' own cloud storage.
Lake Compute, now in Private Beta, is a single-node SQL engine built on DuckDB and runs dbt models directly against Apache Iceberg tables, built and priced specifically for transformation, not general-purpose compute. Together, the two give teams the flexibility to run each workload on whichever engine fits best, optimizing cost without re-platforming, the company said.
Fivetran Context Layer (Private Beta) unifies the data and metadata needed to give LLMs and AI agents relevant context, building on dbt’s structured context and adding unstructured knowledge, such as docs and Slack threads. This service uses Agents Schema, an open source standard, that centralizes context in a structured, extensible format directly in the data warehouse. The context is accessible to teams via preferred MCP or AI tools, including generally available integrations through AI marketplaces including Anthropic and a plugin in ChatGPT.
dbt Wizard in the dbt platform (Public Preview) is built to close that gap. It's natively connected to your project, knows which tool to call, pulls the right context automatically, and proactively validates changes before they ship.
Wizard is also expanding beyond the dbt platform with Wizard CLI (Public Beta), bringing the project-grounded agent directly into the terminal, and Wizard Desktop (Private Beta), a dedicated local workspace for longer, more complex work.
Wizard Explore Mode (Public Preview) brings conversational analytics to business users, enabling them to ask questions in plain language and get answers grounded in the same dbt project the data team maintains. When an answer falls short, those questions can also surface what the data team should improve next.
dbt Charts (Public Beta) brings governed BI alongside the models it depends on. Instead of governance living in a separate, closed tool, they are defined as YAML and version-controlled alongside the dbt models they reference, creating a shared, declarative format that both humans and AI agents can read, write, and review.
For more information about this news, visit www.getdbt.com.