By Sean Young, Director of AECO, Geospatial and AI Solutions Industry Marketing, NVIDIA
AI as infrastructure
Mention electricity and most people picture a lightbulb. But it's a strange thing to summon when you stop and think about what electricity actually does.
It runs the building you're sitting in, the vehicle in your driveway, the water in your tap, the data center storing every photo on your phone and the equipment on every active jobsite in the world. Calling electricity "the thing that powers lightbulbs" isn't wrong, but it badly undersells what electricity is.
AI is getting the same treatment right now. When someone says AI, most picture a chatbot writing an email. And yeah, that's a real thing AI does. It's also a lightbulb-sized way of thinking about something that could reshape the world.
My view is based on what I see every day: an entire industry being rebuilt around a new kind of utility. AI is not a feature being added to the software you already use. It's a layer the rest of the world will run on, the same way electricity quietly became the layer everything runs on a century ago.
Generative AI: the “lightbulb” moment
The release of ChatGPT in late 2022 was a genuine inflection point. For the first time, AI stopped being something engineers talked about at conferences and started being something your aunt asked you about at Thanksgiving.
A lot of early opinions compared it to the arrival of the internet. I understand why that framing took hold, but it's the wrong size. Saying generative AI is an advancement in communication is like saying electricity was an advancement in late-night reading. Technically true, and almost entirely beside the point.
The chatbot was the lightbulb. It was the first thing AI did that felt magical to a general audience, and it deserves credit for that. But the real “magic” in AECO has been going on for a lot longer than generative AI.
The foundation
For years, the AECO industry has been using AI to process enormous volumes of data and find patterns no human could surface in a reasonable timeframe. Computational simulation. Clash detection. Reality capture. Generative design. Geospatial analysis at planetary scale. Digital twins that hold a living model of a physical asset and update as conditions change.
But if the foundation was perception AI, and the present is generative AI, what comes next?
AI that shapes physical reality
Everything in the AEC industry is physics. Concrete cures on a schedule. A crane swings with a load that has weight and momentum. A beam deflects under a calculable force. A worker walking across a jobsite occupies space at a measurable speed. This is an industry that runs on the behavior of real things in real conditions.
That is where physical AI comes into play.
Picture a jobsite camera tracking a crane moving a heavy payload across the site. The system knows the weight of the load, the speed of the swing, and the position of every worker in the frame. If a worker walks into the predicted path of that payload, the system doesn't have to wait for someone to notice. It can sound an alarm, alert the operator, or trigger the kill switch on its own.
That's agentic AI: physical understanding plus the authority to act on it.
The same logic applies in the office. A design engineer today spends an enormous amount of time learning software, clicking through menus, and shepherding a model from one tool to the next. Instead, imagine giving AI a set of design parameters and letting it work. It generates the geometry. Runs the structural simulation. Reads the results. Adjusts the model. Runs the simulation again. Explores three or four design options in parallel rather than one at a time. By the time the engineer comes back from lunch, the system has narrowed the field to the options that meet every threshold.
The engineer’s job just got a whole lot faster and easier.
AECO and AI: a symbiotic relationship
The AI doing all of this needs somewhere to live. These fifty-billion-dollar campuses—with cooling requirements, power requirements, and timelines that no traditional data center playbook can handle—don’t build themselves. They are designed by engineers and built by construction firms.
The industry building the factories is the same industry the factories transform. The tokens coming out of those facilities are the same tokens that power the physical AI on the jobsite, the agentic AI in the design office, and the simulations that test buildings before a single piece of steel is cut.
What comes next
The companies that figure this out first are going to look very different from the ones that don't because they’ll already have rebuilt their work around what software can do now. Every engineer will operate alongside what amounts to an infinite team of agents, running simulations, exploring options, handling the work that used to eat the calendar.
That shift is going to reach the jobsite, too. Robotics informed by onsite cameras and reality capture. Prefabrication and modular construction done inside factories where quality, safety, and schedule can actually be controlled. Geospatial AI running on satellites overhead, analyzing conditions on the ground in close to real time. None of this is speculative. All of it is being built right now, by people in this industry.
I have two kids in university studying engineering, and the question they ask me at dinner is the same one a lot of people in AEC are asking themselves. Is there still going to be a job for me on the other side of this?
My honest answer is this: There's going to be more work, but the work is going to look different.
Someone has to design the AI factories. Someone has to build them. Someone has to figure out how to bring this technology into a workflow without breaking what already works. What happens next gets built by the people who know how to build.
Hear more from Sean Young and leaders from Microsoft, Snowflake and Trimble in the “Beyond the Hype” AI webinar recording.
About the author
Sean Young is Director of AECO, Geospatial, and AI Solutions Industry Marketing at NVIDIA, where he leads industry marketing strategy focused on helping customers and partners accelerate their AI solutions. Sean has 25 years of 3D visualization and simulation experience across AEC, automotive, and manufacturing. Previously, he led Omniverse sales at NVIDIA, AEC business development at HP, and 3ds Max product management at Autodesk.
