Conference Dispatch · MS Build 2026

MS Build Day 1

Starting with the keynote, and throughout the breakouts I've been to today, one thing is clear: in 2026, Microsoft is investing heavily in two themes — and every session I walked into was building on one or both.

Where Microsoft Build · Seattle When June 2, 2026 Filed from the conference floor

Theme 01

Data consolidation across the Azure and Microsoft 365 platforms — one context layer instead of a dozen silos.

Theme 02

A firm foundation for true edge AI compute — intelligence moving onto the device, out of the cloud round-trip.

One fabric, not four products

On the data side — the announcement of Microsoft IQ tells the story. Four pillars combining into a single context layer that gives agents unified access to enterprise knowledge, business data, and the live web:

This isn't four new products. It's one fabric designed to eliminate the data silos that have made enterprise AI implementations crazy and convoluted. Some in Fabric, some in Entra, some in SQL — and always the problem of either consolidating it all onto a single place or medium, or running disparate ETL processes to pull it together.

If your agents can't see across your systems, they can't reason across your systems. Microsoft is solving that at the platform level.

Intelligence moves to where the work happens

On the edge — Satya opened with Windows and never left it. New on-device models (Aion Instruct and Aion Plan) running locally without cloud round-trips. The Surface RTX Spark Dev Box running three local models simultaneously while the developer keeps working. A trillion-parameter model running on a desktop with the DGX Station. And Project Solara — an entire device ecosystem designed from scratch for agents, not adapted from phones or tablets.

Jensen Huang joined from Taipei and said something that stuck with me:

"Past CPUs were designed for humans. Humans are more patient than agents." — Jensen Huang, via link from Taipei

I believe we're seeing personal compute evolve — from being centered on how much resource it takes for the human to do their job, to specifications that let the agents operate more efficiently, which assists the human in doing their job. Silicon redesigned around latency and performance targets that humans never needed.

The through-line

There will be plenty of things I write about in the coming weeks from this trip. But the overarching theme will still be these two: the platform is consolidating its data story and pushing intelligence to the edge. Every session I walked into today was building on one or both.

Day 2 tomorrow.

← hills & valleys / home