Microsoft ($MSFT) – Azure VP of Data Arun Ulag Interviews with Citi – September 07, 2024
Fabric & GenAI:
This interview got into the weeds a bit on Azure Fabric’s approach to data amid the GenAI explosion. It’s a somewhat slept-on Microsoft product, so I’m glad we got a brief interview on it.
As countless players race to build the biggest and best model, a data foundation is a prerequisite to winning. Models are only as good as the quantity of relevant data they have access to train on. That’s where Azure’s broad suite of database products, query language options and integrations come in handy. “Fabric” is its overarching platform to tie together all of its data products. Fabric is inherently open-sourced, which creates a more cohesive environment for minimizing data silos and lockage. And shockingly, there’s a copilot for Fabric to ensure customers have an easier time opening their arms to data modernization and GenAI transformation. Again, the two ideas go hand-in-hand.
To deepen the value of this structure, Microsoft unleashes its data across its wildly broad suite of products. It’s not just organized and secure… it’s actually usable in Excel, Teams, LinkedIn etc. That reality has helped Microsoft successfully create intelligent copilots across several individual products. Again, these copilots need data. Enter Fabric.
Ulag offered an analogy of other platforms forcing customers to buy pieces of a car and assemble them on their own. He joked about Chief Information Officers becoming Chief Integration Officers, as they seemingly spend their time ensuring fragmented systems work well together. By housing everything a customer needs in one place (with endless scalability thanks to Azure’s infrastructure), Azure offers the full car, in one package and in a fully managed fashion. Developers don’t need to paste anything together.
Fabric Impact & Innovation:
In practice, Fabric greatly reduces costs by driving vendor consolidation and broader interoperability. Furthermore, in this case, vendor consolidation removes the need to “create pools of isolated compute” across several point solutions. This means compute spend can be allocated to whichever Fabric use case a customer wants. This minimizes wasted budget. For one customer, this meant cutting data spend from $165 million annually to $45 million. Indirectly, Fabric’s ability to help fetch and conjoin all needed datasets has a way of making companies smarter, more efficient and, in turn, more profitable too.
Ulag is especially excited about Fabric’s developments in real-time intelligence. In his point of view, most firms still work with “batch mode” data on at least a 24-hour delay. This blocks the ability to act on real-time information. Fabric unlocks this ability by processing data as it’s created.
