Microsoft Fabric or BigQuery: which one fits a small European data team?
Fabric rents you an engine with a floor price; BigQuery meters what you actually scan. Here is the honest cost, residency and skills comparison for a five to twenty person team.
Every comparison I read treats this like a feature checklist. Rows and rows of ticks and crosses, both platforms mostly tied, and a conclusion that says "it depends on your needs" without saying what that actually means for a team your size. It is a genuinely close call for a big enterprise. For a five to twenty person team it usually is not, and the reason has nothing to do with features.
Which one is actually cheaper for a small team?
Usually BigQuery, because of how the two platforms are metered, not because either one is inherently pricier.
Fabric rents you an engine. The smallest capacity, an F2, lists at roughly €0.38 an hour in West Europe, which is about €278 a month left running around the clock, or about €165 with a one year reservation (Agilytic's breakdown, Pexon's F-SKU table, both checked August 2026; I go through the full math, including the pause schedule that gets a small team closer to €100 a month, in the F2 pricing post). That bill exists whether the capacity ran one query or ten thousand. It is a floor, not a meter.
BigQuery meters what you actually scan. The first tebibyte of on-demand queries each month is free, and beyond that the EU multi-region charges $6.25 per TiB, converting to roughly €5.44 at $1 ≈ €0.87 (Google's own BigQuery pricing page, checked 21 September 2026; the ECB's free API was unreachable from this environment, so the rate comes from the Federal Reserve's H.10 foreign exchange release for 18 September 2026 instead). Storage runs about €0.017 per GB a month active, half that once a table has sat untouched for 90 days. A quiet month costs close to nothing beyond storage. A busy month costs more, but it costs more because you did more.
There is a matching floor on the BigQuery side too, and it is worth naming so nobody gets surprised by it: slot reservations (the Editions tier, for teams that outgrow on-demand) have a 100 slot minimum. At EU Standard edition rates that is about €3.83 an hour, north of €2,700 a month run continuously. That is a real floor, just one almost no team your size should ever reach, because on-demand querying has no minimum at all. Stay on it until a specific, measured problem tells you otherwise.
What does "EU region" actually buy you on each platform?
Different things, and the difference matters more than either vendor's marketing suggests. I wrote the long version in the residency-versus-sovereignty post; here is the platform-specific part.
Pick a Fabric capacity in a Multi-Geo enabled EU region and your OneLake storage and compute sit in that specific region, though your tenant's own metadata stays in its home region regardless (Microsoft's multi-geo docs). You are choosing one country: West Europe means the Netherlands, North Europe means Ireland.
BigQuery's "EU" location is a multi-region, not a single country: your table
data can sit in Belgium or the Netherlands, and Google guarantees it stays
inside that EU boundary without guaranteeing you which of the two
(Google's regional locations documentation).
For most companies that is a distinction without a difference. It starts to
matter if a client contract or a sector regulator names a specific country,
not just "the EU", in which case you want a single BigQuery region like
europe-west1 (Belgium) or europe-west4 (Netherlands) instead of the
multi-region bucket, and the honest sovereignty question, who can legally be
compelled to hand the data over, is identical on both platforms until you pay
for a genuinely EU-owned provider.
Which one fits the tools your team already has?
This decides more comparisons than price does. If your company already runs on Microsoft 365 and Power BI is already how people read numbers, Fabric removes a real integration tax: Power BI, the warehouse, the lakehouse and the pipelines share one workspace and one copy of data in OneLake, so nobody copies exports between tools. If your team already lives in GCP, or wants SQL-first tooling that stays out of the way, BigQuery pairs naturally with Dataform for transformation and Looker Studio for a free, if less polished, BI layer, with paid Looker available when you outgrow it.
The honest disqualifier: switching either direction to chase a features list costs more in retraining and rebuilt pipelines than either platform ever saves you in list price. Pick based on what your team can run well on day one.
What about the AI layer?
Both platforms now sell you a governed way to build on top of the data, and both are genuinely usable at small scale. Fabric's Data Science workspace and Copilot features sit on the same OneLake data your BI already uses. BigQuery ML lets you train and run models with SQL directly against warehouse tables, and Vertex AI takes over from there for anything heavier. Neither wins outright here; the governance work underneath, clean layers, access control, lineage, matters more than which vendor's AI menu you pick, which is exactly what I cover in AI-ready data work.
The comparison, in one table
| | Microsoft Fabric | BigQuery | |---|---|---| | Pricing model | Rented capacity, wall-clock billed | Metered per query scanned | | Cost floor | ~€165 to €278/month (F2) | None on-demand; ~€2,700/month if you buy reservations | | Free tier | None for production capacity | First 1 TiB queried/month | | EU residency granularity | One chosen region (country-level) | EU multi-region (Belgium or Netherlands) or a single region | | Native BI | Power BI, built in | Looker Studio (free) or Looker (paid), separate | | Best fit | Already a Microsoft shop, steady daily use | Already on GCP, spiky or unpredictable use |
So which should you actually buy?
If your usage is steady, all day, every working day, and your team already lives in Power BI, Fabric's floor stops being a downside and starts being a predictable line item. If your usage is spiky, exploratory, or you are not yet sure how much querying you will actually do, BigQuery's metered model means you find out at low cost before committing to anything. Most five to twenty person teams I meet are still in that second position, which is exactly why the free tier and the absence of a floor tends to win the argument before features ever come up.
Send me your query patterns and your team's current stack and I will tell you, honestly, which floor you would actually be paying for. Ask any question you like in the free 30 minutes; you leave with an answer or a clear next step.