Microsoft Construct has spent the final a number of years changing into an important calendar occasion for builders engaged on the Microsoft platform. The 2026 version, operating June 2–3 at Fort Mason Heart in San Francisco, seems to hold that momentum additional, with AI sitting on the heart of virtually all the pieces on the agenda.
Whether or not you are attending in particular person or watching on-line totally free, the session catalog has over 90 entries overlaying all the pieces from agentic AI to mannequin coaching. Registration for on-line entry prices nothing, which suggests the total two days of content material, together with the opening keynote from CEO Satya Nadella, are open to anybody with an web connection.
Why Construct 2026 is essential to AI’s future
Final 12 months’s convention set a excessive bar. Construct 2025 produced greater than 50 bulletins underneath the theme of AI brokers, together with a brand new GitHub Copilot coding agent that might autonomously repair bugs, write assessments, and open pull requests with out developer prompting. That shift, from AI instruments that help to AI techniques that act, is the thread operating via virtually each session this 12 months.
Microsoft has moved rapidly since then. Copilot Studio has gained autonomous agent capabilities, and the AutoGen multi-agent framework has matured right into a platform builders are utilizing in manufacturing. Construct 2026 is the place a lot of that work is predicted to be demonstrated publicly for the primary time, with the session catalog organized round six predominant themes: developer instruments and frameworks, cloud platform and knowledge, mannequin coaching, brokers and apps, accountable AI, and Home windows.
The convention can be notable for its format. This 12 months’s version is likely one of the extra intimate Microsoft Construct occasions in latest reminiscence, with round 2,500 in-person attendees anticipated, down from the bigger crowds of earlier years. Microsoft and GitHub management have described that as a deliberate option to preserve the deal with technical depth reasonably than broad-audience product launches.
Azure is central to the image as properly. The Azure AI platform has expanded considerably over the previous 12 months, including companies round retrieval-augmented technology, fine-tuning, and agent orchestration. Periods this 12 months are anticipated to maneuver previous introductory ideas and tackle the tougher query of what it truly prices to ship AI-powered purposes reliably at scale.
For builders who wish to perceive the place AI-assisted software program improvement is heading, Construct 2026 provides a transparent view. You will get an in depth have a look at what Microsoft’s engineering groups are constructing, alongside sincere takes from practitioners on what’s working in manufacturing proper now.
10 free Microsoft Construct periods you need to attend this 12 months
All keynotes and most breakout periods at Construct 2026 can be found on-line without charge, so you may observe the convention with no journey price range. For every session under, on-line streaming or recorded entry has been confirmed via Microsoft’s revealed session catalog.
These are the ten periods value blocking out in your calendar, whether or not you are targeted on AI brokers, GitHub Copilot updates, native mannequin inference, or understanding how Microsoft is considering the subsequent section of developer AI.
You may view Microsoft’s online session catalog for a whole checklist of keynotes and occasions.
Opening keynote: Creating new alternative for builders within the period of AI
The opening session is the place Microsoft units the course for all the convention. Satya Nadella is headlining, and primarily based on the session description, the main focus is on what Microsoft calls “creating new alternative for builders throughout our platforms on this period of AI.”
Count on dwell product bulletins and demos from Azure and Home windows engineering groups, with OpenAI seemingly represented given how central that partnership is to Microsoft’s present course. That is the session to look at if you would like the massive image earlier than diving into the technical tracks. It tends to be dense with information, so plan to look at it dwell and return to the recording to catch something value noting intimately.
BRK233: Software program defensibility within the period of AI coding
One of many extra philosophically pointed periods within the catalog, this one is geared toward builders enthusiastic about the place their work is heading as AI brokers grow to be able to producing code, writing assessments, and deploying purposes autonomously. The query of what human builders uniquely contribute in that atmosphere is urgent for lots of groups proper now.
The session is more likely to body that problem as a possibility, with sensible steering on the sorts of higher-order work AI can’t simply replicate. For engineering leads and product groups, it ought to floor helpful enthusiastic about the place to focus improvement effort in an AI-first atmosphere.
BRK260: Construct native AI experiences that harness the GPU, NPU and CPU on each Home windows PC
Microsoft’s on-device AI story has grow to be significantly extra particular up to now 12 months.
This session covers Home windows AI APIs, now increasing past the NPU to help GPU and CPU, alongside Foundry Native for operating open-source fashions instantly on a Home windows machine. There’s additionally new tooling for VS Code that can assist you optimize and put together fashions for on-device deployment, and Home windows ML now helps internet apps via WebNN.
In case you’re constructing purposes that must run AI fashions domestically reasonably than calling cloud APIs, that is the session to prioritize. On-device inference issues most the place knowledge privateness and low latency are non-negotiable, notably in disconnected or air-gapped deployments.
BRK261: Construct and ship sooner with a developer-optimized expertise on Home windows
This session sits on the intersection of Home windows platform improvement and AI tooling, overlaying how Microsoft’s developer expertise enhancements over the previous 12 months match collectively. It is more likely to tackle the enhancements to the inner-loop improvement expertise on Home windows, together with how Copilot integrates into on a regular basis workflows for builders constructing cloud-native and AI-powered purposes.
For builders who spend most of their time on Home windows machines, this one provides sensible worth even when autonomous AI brokers aren’t but a part of your day-to-day work. The sooner suggestions loops being added to the platform are value monitoring no matter what you are constructing.
BRK222: The sincere practitioner’s tackle agentic AI on Kubernetes
The title does many of the work right here. That is positioned as a no-hype actuality examine from engineers who’ve shipped agentic AI techniques in manufacturing on Kubernetes, which makes it stand out in a catalog that may typically lean towards polished demos over subject expertise.
You will seemingly hear concerning the particular failure modes of multi-agent techniques at scale and the gaps between what product bulletins promise and what manufacturing deployments truly ship. In case you’re already constructing agentic techniques or planning to, that is in all probability essentially the most instantly helpful session in all the monitor.
BRK207: GitHub Copilot in Visible Studio: brokers that debug, profile, and take a look at
This demo-heavy session covers what Copilot brokers inside Visible Studio can do past producing code. You will see brokers root-cause bugs utilizing dwell runtime habits and pinpoint efficiency bottlenecks with focused repair suggestions. There’s additionally a thread on constructing take a look at protection to catch regressions earlier than they ship, with all of it aimed particularly at enterprise builders working in C#, .NET, and C++.
In case your crew makes use of Visible Studio for manufacturing work, that is in all probability essentially the most instantly relevant session on the convention. The deal with diagnostics and code high quality reasonably than simply code technology is what makes it value attending over the various different Copilot-adjacent periods within the catalog.
BRK202: Azure DevOps meets GitHub, the trail to AI-powered SDLC
Many improvement groups at the moment run each Azure DevOps and GitHub, managing two techniques that overlap significantly however do not absolutely unify. This session lays out Microsoft’s present integration story and the course it is heading, with AI-powered software program improvement lifecycle administration because the connecting thread.
The AI angle right here goes past GitHub Copilot. It additionally covers how hybrid patterns connecting GitHub with Azure Boards and Azure Pipelines can allow what Microsoft calls Agentic DevOps, with a first-person account from Microsoft’s personal engineering groups who’ve already adopted this strategy.
For engineering managers planning a tooling consolidation or migration, this session provides a transparent image of what is truly out there now.
DEM322: Smaller, sooner, smarter: distilling brokers with fine-tuning
Inference prices are a rising concern for groups operating AI purposes at scale, and this session addresses one of many extra sensible methods for managing them. Mannequin distillation, coaching a smaller mannequin to duplicate the outputs of a bigger one, enables you to hit latency and price targets with out sacrificing an excessive amount of on high quality.
The session covers distilling frontier fashions into purpose-built brokers, a sample that is gaining traction in manufacturing environments the place operating massive fashions on each request is not economically viable. Count on concrete examples of the place distilled fashions carry out properly and the place they fall brief, which is precisely the nuance that is lacking from most AI price discussions.
BRK234: Delivery customized fashions at scale from fine-tuning to inference
Wonderful-tuning a mannequin on Azure is comparatively simple. Getting it into manufacturing, maintaining it performant as utilization grows, and managing the operational overhead is significantly much less so. This session covers the end-to-end workflow, together with the components that the majority product demos skip over.
You will seemingly hear concerning the infrastructure selections that have an effect on price and reliability at scale, and the monitoring approaches that floor issues earlier than customers encounter them. The handoff between coaching and operations groups additionally will get protection, which is likely one of the extra sensible subjects that manufacturing groups truly care about.
DEM364: Simplify app dev with cloud-native PostgreSQL in Azure HorizonDB
The final session on this checklist is likely one of the extra technically particular, but it surely addresses an issue a rising variety of groups are hitting: AI-driven purposes are sprawling throughout separate companies for vector search, fashions, and retrieval pipelines, and the complexity is changing into a legal responsibility. This session introduces Azure HorizonDB, which embeds AI and search instantly right into a cloud-native PostgreSQL database.
You will see how one can run hybrid vector queries and name managed AI fashions instantly from SQL, prototyping agentic workflows with out stitching collectively a separate stack of companies. For builders who wish to preserve their structure tighter and ship sooner, it is a session value watching.
What to anticipate from Microsoft Construct 2026?
Construct 2026 takes place at some extent the place the preliminary wave of AI bulletins has settled into one thing extra operational. The questions this 12 months are much less about whether or not AI belongs within the improvement stack and extra about making it work reliably at manufacturing scale with out blowing the price range. That shift reveals up clearly within the session titles, which lean towards manufacturing considerations and sincere tradeoffs reasonably than introductory overviews.
The San Francisco venue can be value noting. Fort Mason is a departure from the Seattle conference areas which have hosted most up-to-date editions. It seems like Microsoft has used the change in venue to justify a tighter occasion with a smaller participant capability. Fewer attendees means fewer broad-audience periods and extra depth within the technical tracks, which ought to make the content material extra helpful for builders who’ve discovered earlier editions too surface-level.
On-line attendance stays free, which makes Construct accessible to builders who aren’t in San Francisco. Keynotes and recorded breakout periods can be found via build.microsoft.com, and lots of periods run concurrently on-line and in particular person. If you wish to monitor how AI improvement is evolving on the Microsoft platform, that is the clearest two-day view out there.