Power BI Alternative: Keep Your Data, Skip the Cloud Round Trip

The Question Worth Asking First

You have a database. It already contains every number anyone is going to ask about, it is already backed up, already secured, already audited, and already sitting on hardware you control.

The modern BI answer to that is: copy it into our cloud on a schedule, and we will send you pictures of it back.

That is worth saying plainly, because once you write it down it stops sounding normal. To get charts out of data you already had, you end up running an extract, a gateway, a refresh schedule, a semantic model hosted somewhere else, and a per-person licence bill that grows every time someone new wants to look.

Power BI is a genuinely good product. This page is not about whether it works. It is about what that architecture costs you, and whether you actually need it.

What Power BI Does With Your Data

Two modes, and the trade-off between them is the whole story.

Import mode — the default, and the one that performs well — copies your data into a semantic model hosted in Microsoft's cloud. It refreshes on a schedule: up to 8 times a day on Pro, up to 48 on Premium Per User or Fabric capacity. So your entire dataset leaves your building on a timer, repeatedly, forever — and between refreshes the dashboard everyone trusts is showing yesterday's numbers.

DirectQuery — leaves the data where it is and queries it live. It also comes with documented performance costs and modelling restrictions, which is why most real deployments end up back on Import.

And if your data is on-premises, neither mode reaches it on its own. You install the On-premises Data Gateway: a piece of infrastructure whose entire job is to pump your data out to a vendor so the vendor can show it back to you. You maintain it, you patch it, you get paged when it falls over.

None of this is a scandal. It is just a lot of moving parts, and every one of them exists because the analytics live somewhere other than the data.

"But Power BI Has an On-Premises Version"

It does, and it is worth being straight about it rather than pretending otherwise: Power BI Report Server runs in your own datacentre and serves both paginated RDL and interactive PBIX reports.

Here is why people who want on-premises usually still end up in the cloud.

Microsoft documents it as a subset. Power BI Report Server does not have dataflows, deployment pipelines, goals, Copilot, AI visuals, streaming, or real-time dashboards. It does not connect to OneLake, Lakehouse, or Warehouse. The comparison page exists on Microsoft Learn and it is not a close-run thing — the on-premises edition receives a fraction of what ships to the service, on a much slower cadence.

It is gated behind the expensive licences. Power BI Report Server is available to customers on SQL Server 2025 Enterprise and Standard editions; for SQL Server 2022 and earlier, the usage rights apply only to Enterprise edition core licences with active Software Assurance, and the right expires when that coverage does. It is also available through Premium capacity. In other words: the self-hosted option is not the cheap option. It is a benefit attached to the premium tier.

So the honest answer to "why doesn't everyone just self-host Power BI?" is not that self-hosting is a bad idea. It is that the self-hosted edition is deliberately the lesser product, and it costs more to qualify for it.

Then There Is the Bill

Power BI Pro went from $10 to $14 per user per month — a 40% increase. Premium Per User went from $20 to $22.

The Premium P SKUs were retired. Microsoft announced it in March 2024; new purchases stopped on 1 July 2024, and existing customers were moved onto Fabric F SKUs by 1 January 2025. If you had built your capacity planning around P1, that plan was rewritten for you.

And the shape of the pricing is the part that compounds. You pay per person who looks. A report that becomes popular is a report that costs more. Which means somebody in your organisation is, right now, deciding not to give a colleague access to a number because of what the seat costs — and that is a strange way to run an analytics programme.

What DataPallas Does Instead

DataPallas runs on your infrastructure, next to your database. There is no copy, no gateway, no refresh window, and no per-viewer licence. It is modern, open source, self-hosted and fairly priced, connecting to any JDBC database, on Windows, Linux, or macOS.

Your Data Never Leaves

This is the architectural difference and everything else follows from it. Queries run against your database, from a server you control, inside your own network. Nothing is extracted to a third party, so there is no gateway to maintain, no refresh schedule to tune, and no conversation with your compliance team about where customer records are being processed.

For teams under GDPR, HIPAA, financial regulation, or plain data-sovereignty rules, that is not a feature — it is the difference between a project that is approved and one that is not.

Licensed by Who Builds, Not by Who Looks

DataPallas is open source with fair commercial pricing, and the thing it is priced around is the people who build and administer reports — not the people who read them. Viewers are simply not a licensed tier.

Which means the awkward question disappears. Nobody has to weigh up whether a particular colleague is worth a seat, and a report that catches on across the company does not turn into a bigger invoice.

It Is Not Only Dashboards

This is the part that surprises people evaluating Power BI alternatives: the dashboard is a fraction of what most organisations actually need from BI.

• Pixel-perfect documents — invoices, statements, payslips, regulatory returns — built from HTML and CSS templates, with paged output, running headers and footers, and a hot-reload preview. Also XSL-FO for strict print work, DOCX with merge fields, and Excel with formulas. This is the workload Power BI has always handled awkwardly.
• Bursting and distribution — one report split by customer, employee, or branch and delivered by email, FTP/SFTP, cloud storage, or portal, on a schedule, with quality checks before anything goes out.
• A document portal — where each recipient sees only their own documents, with notifications, your branding, and online payments.

Dashboards and Embedded Analytics, Without a Capacity

Interactive dashboards, plus data tables, charts, and pivot tables that drop into your own application with a script tag — React, Vue, Angular, or plain HTML.

Embedding analytics in a product you sell is the case where Power BI's economics get genuinely painful. Here it is a web component.

A Semantic Layer You Can Read

Cubes define dimensions, measures, joins, segments, and hierarchies once, and every consumer — reports, dashboards, AI chat, embeds — resolves to the same definition. AI drafts it from your live schema.

And here is a difference that is easy to miss until you have lived with it. You configure reports and models through a UI, exactly as you would expect — with AI beside any field that wants code, and invisible otherwise. But the Cube DSL and everything alongside it are written to disk as plain text.

A .pbix is a packaged container: you open Power BI Desktop to find out what is inside, and comparing two of them means opening both. Here the definition is a file in your repository. Review a change in a pull request, see precisely what moved, put it back if it was a mistake. You get the editor and the audit trail, not one or the other.

Scale Without Copying to Someone Else

When analytical load would genuinely hurt your production database, OLTP-to-OLAP sync via CDC replicates into DuckDB or ClickHouse — on your own hardware, in seconds rather than on an eight-times-a-day timer. You get the performance of a copy without the copy leaving the building.

AI Without Handing Over the Dataset

Chat2DB turns plain English into SQL, and the AI crew covers reporting, ETL, portals, and web. Asking questions through a cube means the model works from your reviewed definitions instead of guessing at raw columns — and you choose the provider.

The Honest Bit

Power BI's visual authoring in Desktop is excellent, and DAX is a genuinely powerful modelling language with an enormous community behind it. If you have people who are fluent in it, that fluency is real and it is worth something.

If your organisation is already all-in on Microsoft 365 and Entra, the integration story is seamless in a way that a separate platform will not match — single sign-on, Teams embedding, Excel connectivity.

What this page argues is narrower: that architecture is the right trade when your data is already in Microsoft's cloud. When your data is in your own database, on your own hardware, the round trip is a cost you are paying for no structural reason.

Where to Start

1. Write down what actually leaves your network today, and how often. The gateway, the extract, the refresh schedule.
2. Count your Power BI seats and multiply by the new price. Include the people who were not given access because of it.
3. Install DataPallas next to your database and build one dashboard against live data, with no extract.
4. Then take a report Power BI has always struggled with — an invoice run, a statement, anything paginated and bursted — and build that too.
5. Compare on the two things that matter: what it costs per additional viewer, and where your data sat while you did it.