My Experience Migrating to dbt Fusion

Fusion is the future. That was one of the biggest takeaways from Coalesce 2025. It’s clear that dbt is investing heavily in Fusion—and that won’t be slowing down anytime soon. When dbt Fusion was first announced in May 2025, I was excited. It promised a far better developer experience and the potential to save both time and money. At the time, though, there were plenty of bugs and I couldn’t get it working with my projects. After Coalesce, I decided to give it another try on two active dbt projects. Below, I outline the steps I took, the features I’m enjoying, and the areas where Fusion can still improve.

Upgrading the First Project

The first project I upgraded was one we set up in July for a client. They’re on dbt Cloud with Snowflake, and I had been developing locally in VS Code using the Cloud CLI. Since the project was relatively new and small, I expected the Fusion migration to be simple—and thankfully, it was.

The only pre-work I did was uninstalling Altimate AI’s dbt Power User extension. I’m a big fan of that extension, so one thing I was eager to compare was whether dbt’s official extension could match or exceed its functionality.

After installing dbt’s extension, I used the built-in walkthrough to upgrade to Fusion and to debug, parse, and compile the project. I did run into a small hiccup: I expected the extension to use the dbt_cloud.yml file, but instead it used the profiles.yml, so I had to ensure the connection info lived there. This may have been user error, but it was ultimately minor.

Once setup succeeded, the Problems tab displayed all existing issues in the project—mostly deprecation warnings. I installed and used dbt-autofix, which handled the bulk of changes automatically. For the most part, this just meant adding arguments and config keys to my data_tests. After merging a PR with those updates, I started digging into the new features.

I was already familiar with lineage and previewing query results from the Power User extension. Fusion’s version includes the ability to view lineage by resource type or materialization, which is a nice touch. The Query Results tab is also clean, though I do wish you could expand columns or click into long values to see them in full.

Testing Fusion’s Parsing Capabilities

The feature I was most excited to test was Fusion’s parsing engine. As a developer, there’s nothing more frustrating than running a query only to get a database error because of a typo or a non-existent column. Fusion eliminates most of that pain.

If you reference a column that doesn’t exist, Fusion surfaces an inline error immediately. Hovering over it shows a helpful message:

This means you catch errors before previewing results or running/building the model—saving time and compute.

One limitation: if multiple errors exist, Fusion only surfaces the first one. Once that is fixed, the next appears. Ideally, all errors would display at once, so I’m hoping this gets addressed in a future release.

I also tested the new rename functionality for columns and models. It worked well, though YAML files do not update automatically yet. This is a known limitation and support is planned.

What’s Working Well—and What Still Feels Buggy

Overall, the most impactful new feature is the live error detection. It meaningfully improves the developer experience. Many other features overlap with what already existed in the Power User extension, so they felt less novel to me.

That said, the dbt extension has been somewhat buggy. At times everything works great; at other times, key features disappear. For example, when demoing Fusion to my team, the “Preview CTE” option simply didn’t appear. Lately, the extension has been more stable, so I’m optimistic most issues have been ironed out.

Upgrading a More Complex Project

The second project I upgraded is a few years old and relies heavily on Fivetran packages. This project had far more warnings, since each package has its own Fusion compatibility concerns. Regardless of whether you adopt Fusion, if you use packages heavily, it’s important to keep them up to date so they remain compatible with the latest version of Core or Fusion.

Final Thoughts

I’m excited about where Fusion is heading. It’s great to see dbt investing in improving the developer experience—especially given that dbt Studio has never been the most developer-friendly environment.

However, conversations I had at Coalesce suggested that adoption has been slow. I’m curious to see how usage evolves over the next few months and what share of the community ends up on Core versus Fusion.

It’s also worth noting that Fusion does not yet support features like the semantic layer. Depending on the complexity of your project, you may not be able to upgrade right away—or you may run into a long list of warnings and errors if you do.

Fusion is promising, fast-improving, and in many cases already usable—but it’s not fully mature yet. Still, I’m optimistic. The future of dbt development looks much brighter with Fusion in the picture.