Power BI got you in the door cheap.
Then the toll started.
Golden is the AI-native, open BI platform. Your data, your models, your tools, no Microsoft tax. Here's an honest look at where Power BI is taking its customers, and where we're going instead.
Built by the team that led product at Tableau and Amplitude. We've watched this movie before.
Three things changed under Power BI in 18 months
Microsoft raised prices for the first time in about a decade, started pushing Premium customers onto Fabric capacity, and drew a hard line on the semantic layer. Each one opened a gap. Here they are.
Up 40% on Pro
Pro went from $10 to $14 a seat. PPU from $20 to $24. No grandfathering, applied at renewal.
The consumption on-ramp
Premium capacity is being retired. The replacement bills by capacity, by viewer, and now by storage.
"The answer is no"
Microsoft said publicly it will never properly support third-party semantic models, then pulled the Databricks option.
What Power BI costs you, and what Golden does instead
Six things, the ones that actually shape your bill and your day. No 40-row feature checklist.
| The angle | Power BI | What it costs you | Golden |
|---|---|---|---|
| Pricing | Pro up 40% to $14, PPU up 20% to $24 (April 2025). No grandfathering. | The "cheap" pick quietly got more expensive, mid-contract. | Predictable, value-based pricing. No surprise renewal jumps. |
| The Fabric toll | Premium retired. Below F64, every viewer needs a paid Pro seat. OneLake storage billed on top. | Capacity units, plus per-viewer seats, plus storage, plus overage. | One clear price. No toll to let a teammate read a report. |
| Openness | Won't support third-party semantic models. Databricks metric-view option removed. | Your metrics have to be rebuilt in DAX, inside Microsoft's model. | Reads the models and warehouses you already have. Bring your own LLM key. |
| Who can build | Building anything real means learning DAX. Everyone else waits. | Business users file a ticket and wait days for an answer. | Ask in plain language. A slider of autonomy from assist to auto. |
| Where it runs | Windows-only authoring. Weak on non-Microsoft and multi-cloud data. | You pay in workarounds if you're not all-in on Microsoft. | Runs anywhere. Warehouse-native across clouds. |
| The AI | Copilot needs Fabric or Premium, works best on pre-modeled data. | AI sits on top of the old dashboard workflow, gated behind upgrades. | AI-native across Discover, Prep, Analyze, Communicate. Not a chat box stapled on. |
Power BI is turning into an on-ramp for Fabric
The licensing tells the story better than any pitch could. Follow the arrows and they all point at more Fabric consumption.
- Premium got retired. Microsoft stopped selling P-SKU capacity to new customers in July 2024. Non-EA renewals ended February 2025. EA customers move to Fabric by January 2028 at the latest.
- A per-viewer tax below F64. Any Fabric capacity under F64 drops free viewer access. Every person who wants to look at a report needs their own $14 Pro seat.
- Free viewing is going away. The old EM rights that let all users read reports are being removed. Those customers face a real cost jump.
- A new storage meter. Fabric bills OneLake storage separately, around $0.023 per GB per month. A line item Premium customers never had.
You budgeted for a flat Premium fee. Now you're modeling capacity units, counting per-viewer Pro seats, forecasting OneLake storage, and watching for overage.
With Golden, the whole spreadsheet is one line. You don't pay extra for someone to see a chart.
Free is never free
Pro comes bundled with Microsoft 365 E5, so it looks free. A free puppy looks free too, right up until the vet bills, the food, and the chewed-up couch.
Microsoft gives away the authoring because that's the cheap part. The money shows up later, in the places that never make the sticker.
- Every report someone runs burns Fabric capacity that you pay for by the unit.
- Below F64, viewing isn't free. Each person who opens a report needs a paid Pro seat.
- Your data in OneLake is metered by the gigabyte, every month.
- The more you build, the more Azure consumption you drive underneath it all.
Because the seat was the bait. The consumption is the business. A free authoring license is the cheapest way to get your data, your reports, and your people inside a cloud bill that only goes up.
Golden doesn't need a loss leader. We're not trying to pull you into a cloud meter, so we can just charge a fair price for the product.
Microsoft said the quiet part out loud
In April 2026 a member of the Microsoft Fabric team published a post on whether Power BI would ever properly support third-party semantic models. The answer, in writing:
What it means in practice
If you want your people in Power BI, your business logic has to live in Microsoft's model, in Microsoft's language, on Microsoft's terms. A few weeks after that post, Microsoft removed the option that let Power BI query Databricks metric views. It was previewed, then pulled. Reports that used it stopped working.
Power BI is also absent from the Open Semantic Interchange, the industry effort to make semantic definitions portable. Snowflake and Databricks are in. Microsoft is not.
What practitioners are saying
The walls matter more than the "no"
The "no" is the symptom. The real cost is a platform built so that leaving is expensive and living outside Microsoft is second-class. Analysts have a name for the pattern. Directions on Microsoft called it "the Fabric takeover."
- Speed is the bait. Direct Lake, the fast path, only works when your data sits in OneLake. Performance pulls your data into Microsoft's lake and keeps it there.
- Outside Microsoft, you get the slow lane. Non-Microsoft sources fall back to DirectQuery or unsupported connector workarounds. First-class treatment assumes Azure.
- Your logic gets trapped in DAX. Business rules live inside the Power BI model, so moving off means rebuilding all of it by hand.
- Every new feature deepens the well. Copilot, Direct Lake, governance, each one needs more of your estate inside Fabric to work.
That's what hostage looks like. Not a locked door, a set of walls that get taller every release.
Golden's take
Openness is the product, not a talking point. Golden reads the models and warehouses you already have. We don't make you rebuild your metrics in a proprietary language to see a chart. We use all the major models, let you bring your own LLM key, and we don't train on your data.
Easy to consume. Hard to build in.
The gap between reading a Power BI report and building one is called DAX, and it's where the daily frustration lives.
The Power BI day
A marketer wants to know why last week dipped. She doesn't know DAX, so she files a request and waits three days for an analyst to get to the ticket. By the time the answer lands, the moment's gone. Multiply that across every follow-up question in the company.
The Golden day
She asks in plain language and gets an answer now. The analyst still owns governance and the deep work, but the slider of autonomy lets Golden do more of the driving when it should. Humans plus AI, not a ticket queue. Think a cursor for data, and a Canva for data.
You're always migrating, and always testing
A platform that keeps moving isn't free to live on. In about two years, Microsoft has deprecated, retired, or scheduled the removal of roughly a dozen Power BI capabilities. Every one lands as a project on your plate.
- Premium P-SKUs, retired and pushed to Fabric capacity.
- Real-time streaming, being wound down through 2027.
- The legacy Excel and CSV import path; models built that way stop refreshing in mid-2026, then reports fail to open.
- Scorecard hierarchies and the heatmap view.
- R and Python visual embedding.
- Q&A, the old natural-language tool, folded into Copilot.
- The Databricks metric-view connector option.
Each one means re-pointing connectors, rebuilding models, re-testing every report that touched them, and retraining the people who used them. On your time, on your budget.
The flip side of things getting pulled is things that never quite arrive. Ship early, stay in preview, let customers find the bugs.
- Two to three years in preview is normal before general availability (composite models and managed VNet gateways went this route).
- Datamarts and Desktop developer mode each spent well over a year in preview.
- AutoML in Fabric is still preview, and not supported for production use.
- Basic source control has been asked for by the community for over a decade.
Preview means you're the guinea pig, building on features that can change or break, with no promised date for when they'll be safe to lean on.
How Golden handles this
We ship at AI speed, and we ship things that work. When we put something in your hands, it's meant to be used, not survived. Fewer forced migrations, fewer "preview" asterisks on the features you're trying to build a business on.
Look at three years, not the sticker
The headline seat price is the hook. The real number shows up when you add the parts Power BI doesn't put on the label.
Per-seat licenses (now higher) + Fabric capacity + per-viewer Pro seats below F64 + OneLake storage + analyst time to build and maintain in DAX.
Rebuilding metrics in DAX. Waiting on the analyst queue. Workarounds for non-Microsoft data. Overage when consumption spikes.
One clear price. No per-viewer toll. No rebuild tax. More people answering their own questions, so your analysts do the work that matters.
Questions we actually get
We're a Microsoft shop. Does Golden even fit?
Do we have to rebuild all our models to switch?
What about governance? Doesn't AI-native mean less control?
Can we bring our own LLM?
Won't we just keep re-doing work every time Microsoft changes something?
Is Golden really cheaper?
Bring one question. We'll show you the day-and-night difference.
Pick a report your team keeps rebuilding in Power BI. Ask Golden the same question in plain language. Watch what happens.
Pricing, licensing, and product details reflect publicly reported Microsoft changes as of mid-2026 and are subject to change. Quotes are drawn from public sources including Chris Webb's BI blog and its comment threads. This page is a comparison of publicly available information and reflects Golden Analytics' point of view.