There's a particular signal an organisation sends when it gives a product its own dedicated conference: this is a platform now, not a feature. Microsoft sent that signal loudly at the end of March by holding the first-ever Fabric Community Conference — FabCon — at the MGM Grand in Las Vegas, three days, over 130 sessions, barely six months after Fabric went generally available. I didn't make it to Vegas, but I've spent the days since reading through the announcements, and the overwhelming impression is of a platform sprinting. The pace is genuinely impressive and, if I'm honest, a little alarming — and I want to talk about both, because the speed is the story as much as any single feature.

Let me cover what actually landed first, then get to the pace problem.

Real-Time Intelligence: the streaming story, consolidated

The headline workload announcement was Real-Time Intelligence, which pulls together and extends what used to be Synapse Real-Time Analytics and Data Activator into a single workload, with a Real-Time hub as the front door for high-volume, high-granularity event and streaming data. For someone who spent years building streaming and IoT platforms the hard way — stitching together ingestion, processing, and a place to land it all — this is a meaningful consolidation. Streaming has always been the awkward stepchild of the analytics world, bolted on rather than designed in. Fabric making real-time a first-class workload alongside the warehouse and the lakehouse is the right architectural instinct. Whether it's as seamless as the demo is the question every consolidation raises, but the direction is sound.

A Workload Development Kit: the ecosystem play

The announcement I think is strategically the most interesting, and got the least breathless coverage, is the Fabric Workload Development Kit (in preview). It lets software developers build their own workloads that run natively inside Fabric — appearing as first-class citizens alongside Microsoft's own — and publish and monetise them through the Azure Marketplace. Read what that actually is: Microsoft trying to turn Fabric from a product into a platform others build on. That's the move that, if it works, makes Fabric genuinely sticky — an ecosystem, not just a suite. It's also the move whose success won't be clear for a couple of years, because ecosystems are built slowly and can't be announced into existence. One to watch, quietly.

Copilot, generally available and everywhere

And of course, Copilot. Copilot in Fabric for the Power BI experience is now generally available, Copilot came to Dataflows Gen2, and the AI assistant is being woven through more of the platform. I've been publicly measured about the gap between Copilot demos and Copilot on a Tuesday with your real, messy model, and I hold that line. But "generally available" does change the conversation from "will this ever be real" to "where does this actually save time" — which is the more useful question, and the one I'd be testing right now with real workloads rather than taking on faith.

OneLake keeps opening up

Threaded through the announcements was a steady widening of OneLake — Fabric's single storage layer — and its ability to reach data that lives elsewhere without copying it. Mirroring, which brings a live replica of an external database into OneLake so you can analyse it without building an ETL pipeline, keeps expanding to more sources. It's unglamorous plumbing, and it's arguably the most important through-line: the more OneLake becomes the place your data simply is, regardless of where it originated, the more the rest of Fabric's story holds together. A platform is only as compelling as its gravity, and OneLake is where Fabric's gravity comes from.

The pace problem nobody's quite saying out loud

Here's the part that struck me most, and it's less about any feature than about the rhythm. Fabric is shipping so fast that the community — the consultants, the platform teams, the people who actually have to run this — is visibly struggling to keep up. Every month brings a substantial feature drop; now there's a whole conference of them. In the community threads the excitement is real, but so is a quieter note of overwhelm: how do you build durable practice on a platform that changes under your feet weekly? How do you write documentation that isn't stale by the next update? And — the question that comes up in every thread eventually — what is all this going to cost, when capacity consumption is already hard to predict and every new workload is another way to burn CU?

That tension is the real headline of the first FabCon. Microsoft is moving at a pace that's thrilling if you're watching and exhausting if you're responsible. When I wrote, the day Fabric went GA, that "available" is not the same as "battle-tested", this is what I meant, only more so: the platform is maturing at a rate that outstrips most organisations' ability to absorb it.

What I'd actually do

So my advice, six months into the GA era and one conference in, is the same discipline with more conviction. Don't try to adopt everything Fabric ships; you can't, and chasing every feature drop is a route to a permanently half-finished platform. Pick the workloads that solve your real problems, build durable practice around those, and let the rest mature before you touch it. Treat the firehose as a menu, not a mandate. The organisations that will win with Fabric aren't the ones adopting fastest — they're the ones adopting deliberately, building on the parts that have settled while watching the rest from a safe distance.

The first FabCon confirmed that Fabric is real, serious, and here to stay — a platform with its own conference and a genuine ecosystem ambition. It also confirmed that the hardest part of the Fabric era won't be the technology. It'll be the discipline to not drink from the firehose just because someone turned it on.