Real Estate and Construction: AI trends to focus on - Spatial AI shifts from lab to infrastructure

What changed this week

AMD agreed to buy Fei-Fei Li's World Labs for $8.2 billion, signaling that spatial AI — the ability to understand and generate 3D environments — is now a core infrastructure play, not a research sideshow. For real estate and construction, this means the models that digitize buildings, sites, and cities are about to get faster, cheaper, and more deeply embedded in hardware.

EliseAI raised $350 million at a $4 billion valuation, confirming that AI for leasing, maintenance, and resident workflows is consolidating into large, well-funded platforms. Property owners who haven't set rules around data portability, fair-housing testing, and human escalation now face a market where a few dominant systems will define how apartments get rented and managed.

On the physical side, Dyna-2.1 shipped a semi-humanoid robot with workflow-level AI, Runway introduced Praxis-1 for video-trained robot control, and Destro AI focused on shared plans for human-robot coordination. Construction and logistics teams are seeing autonomous equipment move from pilot to product, but the week's releases all emphasized safety handoffs, not full autonomy — a signal that job-site integration remains the hard part.

What it means for you

Your AI strategy now has a physical dimension you can't ignore. Grid capacity, data center siting, and on-site power are shaping what gets built and where. When hyperscale campuses and AI inference clusters compete for the same constrained electrical infrastructure you need for a new development, project feasibility depends on energy planning as much as entitlements. Start asking your engineering teams about peak load, backup, and phasing early in the design cycle.

Leasing and property management AI is becoming a platform decision, not a point-solution choice. EliseAI's scale means more owners will be pushed toward integrated systems that handle inquiries, tours, renewals, and maintenance requests inside one vendor's stack. Before signing, require contractual commitments on data extraction, audit trails for fair-housing compliance, and documented human escalation paths. If a vendor can't show you how resident outcomes improve when automation increases, walk away.

Robotics and spatial AI are converging on your job sites and buildings. Runway's Praxis-1 uses video training to teach robots manipulation tasks. Destro AI builds shared plans so humans and machines don't collide. These tools will reach you through equipment manufacturers and specialty contractors. Your role is to demand measurement accuracy specs, connectivity requirements, and clear handoff protocols before any autonomous system enters a live site. The technology works in controlled settings; your sites are not controlled settings.

Document-heavy workflows — leases, permits, contracts, specs — are getting faster. Cohere's Parse 5 improves enterprise document extraction. OpenAI's GPT-6.1 Sol lowers the cost of running large language models. Combined, these make it practical to query thousands of pages of project records, lease agreements, or maintenance logs in seconds. The catch: you still need a human to verify source quality and sign off on decisions that carry financial or legal weight.

What to focus on next week

• Audit one leasing or property management workflow for AI dependency. Identify where automated decisions touch residents — screening, renewal pricing, maintenance prioritization — and confirm you have auditable logs and a human override for each step.
• Ask your project teams to map site energy constraints against any AI tool that requires on-premises compute, cameras, or autonomous equipment. If a vendor can't tell you the power draw and connectivity needs, delay the pilot.
• Test a document extraction tool on a real contract or permit set. Pick 50 pages you actually use, run them through a modern parser, and measure whether accuracy and time savings justify the cost before scaling to your full document library.
• Request a data portability clause in any renewal or new contract with a property-tech vendor. If the vendor processes resident data, leases, or maintenance records, you need a written guarantee that you can export your data in a usable format without penalty.
• Watch one robotics or spatial AI demo with your safety lead present. Have them ask the vendor three questions: what happens when connectivity drops, how are near-miss events logged, and who is liable when the system fails.

For the full list of stories that informed this week's analysis, see all Real Estate and Construction AI news.