The most watched Fabric content this week is not a feature demo. It is Guy in a Cube answering "when should you start using Microsoft Fabric", and it did 22,000 views in seven days. Nobody is asking what Fabric is any more. They are asking when.
So here is the answer I give clients, before any of the nuance:
Move when your Power BI estate starts fighting its own workarounds. Stay put while Pro or Premium Per User is quietly doing the job.
Everything else is detail on top of that sentence. But the detail is where the money is, so let's do the detail.
The signs you are ready
You do not need all of these. Two or three is usually enough.
- Your refresh schedule is a negotiation. Datasets queuing behind each other, refreshes timed to avoid other refreshes, someone keeping a spreadsheet of when things run. That spreadsheet is a capacity plan. You are already running a platform, just without the platform.
- Data leaves source systems through side doors. Exports landing in SharePoint, CSVs emailed on the first of the month, an Access database someone is embarrassed about. Fabric's whole pitch is one place where data lands once and everything reads from it.
- The same data is modelled twice. Two teams, two semantic models, two versions of "revenue". A lakehouse with shared models on top is the actual fix, and that is a Fabric-shaped fix.
- Your data people want more than Power Query. The moment someone asks for SQL endpoints, notebooks or dbt, you are bolting engineering tools onto a BI licence. Fabric has them in the box, and as of this month dbt jobs are properly supported, with Lakehouse support in preview.
- You are paying three vendors to do one job. A pipeline tool, a warehouse, and Power BI on top. Sometimes that stack is right. Often it is three invoices for one flow of data.
The signs you are not
- Pro is coping. Models fit, refreshes finish, nobody is blocked. Fabric would give you headroom you do not need yet, billed monthly.
- Nobody would own it. A capacity is a thing you run, not a thing you buy. Someone has to watch consumption, set the guardrails and own the bill. If no one owns your workspaces today, Fabric multiplies that problem rather than solving it.
- You want it for one feature. One team wants notebooks? There are cheaper ways to buy notebooks.
What changed in August 2026
Three things this month moved the maths, which is partly why I am writing this now.
First, Fabric Runtime 2.0 went GA, with Microsoft saying it becomes the default from late September. The engine under Fabric's data engineering side just had its big generational upgrade, which is the kind of thing you want to land before you arrive, not while you are mid-migration.
Second, dbt support stopped being a workaround. That removes the loudest objection analytics engineers had.
Third, Microsoft shipped three separate cost-control features in a single week: capacity monitoring events, a proper guide to tracing warehouse consumption, and custom SQL pools to cap compute. I read that as an admission that unpredictable bills have been the real adoption blocker, and as evidence they are actually fixing it.
The security caveats around Direct Lake that made cautious IT teams wait are also being closed off, one by one. The "wait for it to mature" argument is running out of things to point at.
The sensible way in
Not a migration project. Nobody needs a migration project.
Buy a small F SKU and stand it up next to what you already have. Pick one workload that is currently painful, a slow dataflow or an oversized model, and move just that. Run it for a month with the Capacity Metrics app open. You will learn more about your real capacity needs from four weeks of one workload than from any sizing calculator, and pay-as-you-go capacities can be paused while you are not testing.
Then decide with numbers instead of a sales deck. Your reports, your workspaces and your licences carry on as they are throughout. Nothing about trying Fabric requires you to bet the estate on it.
What I would not do
Sign up for a big-bang migration because a renewal date is looming. Renewal pressure is how estates end up half-migrated, with the old stack still running because nobody had time to move the awkward last third. The awkward last third is the whole job.
If you want a second pair of eyes on whether your estate is actually ready, that is literally a thing I sell: a fixed-price Power BI health check that tells you what state you are in and what I would move first, if anything.
And if Pro is coping? Keep your money. The moment you catch yourself building a workaround for a workaround, that is the moment.