Proptech spent a decade improving decisions. Search got faster. Underwriting got cleaner. Leasing moved online. Those gains matter, but they stop at the keyboard. Physical AI starts where traditional proptech stops.
Physical AI moves value creation into the building. Software starts issuing work, not reports. Robots and autonomy systems execute tasks. Sensors turn every hallway, boiler room, and loading dock into a data source. Models improve because the building produces training data every hour.
VCs should treat this as a category shift, not a feature upgrade. The profit pool moves from lead gen and workflow tools into operating margin. Proptech venture capital hit $16.7 billion in 2025, up 67.9% year over year, with AI-centered proptech companies growing at 42% annually.
What physical AI means in a real estate context
AI Generated @25V
In real estate, action shows up as:
Autonomous task execution
Cleaning, security patrol in private areas, inventory checks, package room runs, and basic inspections.
Automated sensing tied to work orders
Leaks, HVAC faults, elevator anomalies, slip hazards, access anomalies, and equipment heat signatures.
Closed-loop optimization
The system detects an issue, schedules the fix, confirms completion, and updates the model.
This changes what buyers pay for. Owners do not budget for “AI.” They budget for fewer truck rolls, lower overtime, faster turns, and fewer claims. One mid-sized multifamily portfolio that deployed smart sensors and automation saw a 20% drop in maintenance costs and a 15% reduction in energy expenses within a year.
So why does the timing look different from past proptech waves?
Earlier proptech waves rode software distribution.
A cloud-based workflow tool does not need a service organization. A robot fleet does. Uptime becomes a product requirement. Training and maintenance become part of the offer. The best companies will feel like operators with strong software.
This also changes capital needs. Hardware, installation, and service working capital push burn higher early. The payoff comes later through long contracts and expansion across sites.
Looking at how prior behavior shifts have diffused, a practical adoption timeline from here might look like this:
25V Research
Year 2026 to 2027
Narrow deployments in controlled settings.
Warehouses, parking, large multifamily, Class A office.
Single task autonomy wins first.
Cleaning robots, patrol robots in private areas, basic inspection routines, and automated fault detection that generates tickets.
Year 2028 to 2029
Fleet management becomes the differentiator.
One team manages many sites.
Owners standardize across a region once they trust uptime and service response.
Year 2030 to 2031
Deeper integration with building systems becomes normal.
Access control, cameras, HVAC, elevators, and work order platforms connect into one loop.
Robots expand into unit turn checklists and exterior inspections.
Year 2032 to 2034
Mid-market adoption grows as costs fall and service networks mature.
Physical AI becomes a standard line item in operating plans, similar to access control and life safety.
Robots at home become synonymous with extended senior care.
Over the last 12 months, I have spent considerable time looking at physical AI in APAC and how eastern and western approaches diverge. Here are a few of my takes.
The US versus Asia opportunity set
Asia tends to lead in manufacturing scale and deployment pressure.
Manufacturing capacity shortens iteration cycles.
Aging workforces and persistent labor gaps — Japan and South Korea lead this trend — push automation.
High-density property types reward automation. A single tower offers repeatable tasks and clear routing.
Asia near-term wedge markets
High-rise residential with centralized facilities teams.
Transit-linked retail and mixed-use with constant operations coverage.
Industrial parks and logistics hubs where autonomy already has acceptance.
The US tends to lead in software packaging and capital markets.
The US builds strong enterprise software and distribution.
Private markets supply growth capital. Public markets supply liquidity.
Real estate fragmentation creates both a pain point and an opportunity. Owners need vendors who remove complexity.
25V Research
US near-term wedge markets
Logistics and cold storage portfolios with clean ROI math.
Sunbelt multifamily with heavy turnover and staffing pressure.
Single-family rental portfolios that need cheap, frequent inspections across a distance.
Healthcare real estate where staffing and compliance drive monitoring demand.
Construction progress capture and safety monitoring on large projects.
Constraints differ by region.
Asia often faces tighter public space rules and data governance differences across countries. The US faces procurement friction, liability concerns, and messy integration with legacy building systems.
I spent time with a VC who specializes in robotics. These were a few of his questions that he is using to underwrite physical AI.
Does the product remove a measurable cost line?
Ask for the baseline: hours per week on the task, all‑in labor cost, vendor spend tied to the task, and claims or incident costs when it fails.Does the company own uptime?
Ask for real uptime, not lab performance: mean time to repair, swap logistics, remote monitoring coverage, and a staffing plan by geography.Does the solution fit operator workflows?
Ask where the work order lands, how access is granted, what happens when a building system fails, and how the night shift actually uses it.Does pricing match how owners budget?
Favor pricing tied to units, square footage, tasks completed, or avoided service calls, and avoid models that require a brand‑new budget category.
What are winners?
We both agreed that the winners will show three traits.
Outcome first positioning
They sell fewer incidents, faster turns, lower cost per task, and higher uptime.Service plus software execution
They built a real field playbook.
They treat maintenance as a core product.Data advantage tied to deployment
Their models improve because fleets run daily in many buildings.
They build a defensible loop.
Where do I land?
Proptech digitized the front office. Physical AI targets the back office. The near-term value is not novelty. The value is the operating margin.
Asia will build and ship more machines faster. The US will package fleets into scalable operating systems faster. Both paths produce winners. The best returns will flow to teams who treat deployment as the product and who price to a building owner’s P&L.
If you invest in proptech, start mapping your thesis around tasks, not dashboards. The next platform will not sit on a screen.
Proptech digitized the front office.
Physical AI targets the back office.
It will show up at the service elevator door, badge in, and get to work. Or show up at your doorstep to help you with your day-to-day operations at home.
Thanks for reading.
Maximillian Diez,
GP, Twenty Five Ventures
If this writing adds clarity or connection to your week, consider supporting the newsletter. Every contribution helps keep the work thoughtful, independent, and free for all readers.
