Autodesk University is not a geospatial conference. It is an AEC, design and manufacturing show, and most years the surveying and reality capture world takes notice only of relevant announcements and moves on. This year was different, and not because Autodesk said the word lidar more often (truthfully, they barely said it at all). What changed is that several of the announcements are quite important if you care deeply about coordinate systems, existing conditions, and getting field data into a design environment without losing its meaning along the way.
Civil 3D joins the cloud, and that has consequences
The headline for infrastructure people is that Civil 3D is becoming a connected client to Forma, Autodesk’s AEC cloud. In the main stage demo, a civil engineer resized a detention pond in Civil 3D, pushed it up, and the building team saw the change in their own environment moments later. Nothing about that demo is technically shocking. Plenty of vendors have shown something similar.
What matters is the direction of travel. Autodesk is arguing that the project database, not the file, is the unit of work. Civil objects become things you can query, report against, and build dashboards on top of, rather than geometry locked inside a drawing that somebody has to open. For anyone who has spent a career extracting survey data out of one system so it can be rebuilt in another, that premise is either very appealing or very familiar, depending on how many times you have heard it.
The assistant is not just a chat window bolted on
Every vendor shipped an AI assistant this year. Autodesk’s case for why theirs is different is that it reasons over connected project data rather than documents, so a change can be traced to its downstream effects on schedule, cost, and sequencing instead of merely described. The company calls this “project intelligence”, and the pitch is that general-purpose AI can reason across information while Autodesk can ground that reasoning in geometry, constraints, and project history.
It works because of the granular data effort underneath. Relationships have to exist as data, not as geometry inside files. Same architecture that produced the geolocation engine.
The piece worth watching is Assistant Builder, which is meant to let customers connect their own AI agents, tools, and partner services to Autodesk project context, with third-party agent support planned through the Design & Make Marketplace. Whether that helps companies in our industry depends entirely on who gets to build. An open program would give scanning and survey vendors a way to reach designers inside the tools they already use. A restricted one would just make Autodesk harder to work around.
Esri, and what sits above and below the ground
The longstanding Esri and Autodesk relationship extends past coordinate systems. Geospatial context is flowing into Forma so that teams working on a station or a road can see terrain, surrounding buildings, and the existing environment federated together rather than designing in isolation.
Below the ground, Autodesk is deepening its work with Fugro. The GeoDin Ground plugin has been feeding borehole and geotechnical data into Civil 3D since early 2025, and the AU story this year is about automating the step from that data to a usable ground model, then pushing the result back out to ArcGIS. Underground and utility mapping is usually framed as a data collection problem in our coverage. The harder problem has always been what happens to the data afterward. If subsurface models become a normal layer in a design environment rather than a PDF in an appendix, that changes the value proposition for everyone doing utility locating and geotechnical investigation.
We covered the unified geolocation engine separately, because it deserved its own piece and because it was the most geospatial thing Autodesk announced all week. In summary, the coordinate system that lived inside Civil 3D got promoted into a shared service now serving Revit and Forma, built on Autodesk technology but shaped heavily by Esri. It is live for Revit in connected workflows, with broader availability signaled but not dated. What matters here is that the geolocation work is the foundation everything else in the AEC announcements sits on. Civil 3D as a connected client, geospatial context in Forma, subsurface models moving between Civil 3D and ArcGIS, none of it works if the coordinate system does not survive the trip. Autodesk did the unglamorous part, but it was a crucial foundation enabling future workflows.
Reality capture is being treated as a product area, not a file format
The clearest signal was another organizational/behind the scenes one. Autodesk now has infrastructure and reality capture sitting together under the same product leadership, and the sessions covering both spent real time on existing conditions as a first-class part of the design process rather than a reference attachment.
The direction is toward point clouds participating in coordination workflows, automated extraction of features from scan data with user-controlled tolerances, and AI-assisted classification of what is actually in a scan, with a lot of their work on this coming from their acquisition of Pointfuse. The question for other companies in that space is whether a platform vendor folding these capabilities into the design environment expands the market or absorbs it.
I do not think the answer is obvious. There are some specialists and software that are arguably farther along at this today. But the specialists are also selling into workflows where the scan eventually has to reach a designer, and the designer is increasingly going to find something already sitting there.
Conclusion
Most of this is still ahead. The next-generation assistant was a preview with rollout details not due until 2027, and the geolocation engine is live in connected Revit workflows with broader availability signaled but not dated. Autodesk was also candid about how hard it is to get customers onto current releases, which is a reasonable thing to say out loud and a reasonable reason to watch the delivery. What stands out is where the engineering effort went. In a year when most vendors spent their keynote budget on agents, Autodesk put a year of work into coordinate systems, ground models, and point clouds that participate in coordination rather than sitting alongside it. Those are our problems, and they are the unglamorous kind that only get solved when someone decides they are foundational. The argument that the file is no longer the unit of work only holds if the pieces underneath it hold, and Autodesk appears to be building them in the right order. That is a better position than the industry has been in for a while.
