This week’s Inman Tech Roundup covers the forces reshaping real estate through artificial intelligence, from Wall Street to the construction job site.
Proptech funding is down but more selective, while a former Opendoor CEO is betting on AI copilots for construction workers. Data centers are pricing homebuilders out of key markets, and OpenAI’s safety stumbles raise questions worth asking before you build agentic AI into your own workflow.
Here’s what you need to know:
Investors get pickier on proptech
Global proptech startups have raised roughly $8.7 billion so far in 2026. That’s down sharply from $24 billion in 2019 and on pace to land near last year’s $12.3 billion, according to Crunchbase data.
Deal count tells a similar story with 794 transactions this year, versus more than 2,400 in 2019. Higher interest rates and a tighter bar for capital have pushed investors toward fewer, larger bets, increasingly in AI-driven construction, property operations and transaction infrastructure with clear ROI, rather than generic real estate software.
Notably, 4 of the 5 biggest 2026 raises happened outside the U.S. Stockholm’s Stegra, a green steel startup, landed $1.6 billion from Wallenberg Investments. Madrid-based Hydnum Steel raised $695 million. The lone U.S. entrant, autonomous construction tech firm Bedrock Robotics, brought in $270 million.
M&A has outpaced IPOs. Autodesk’s $3.6 billion purchase of MaintainX topped the list, followed by Compass’s $1.6 billion all-stock deal for Anywhere and The Real Brokerage’s $880 million acquisition of REMAX. EquipmentShare’s $747 million IPO in January stood as the sector’s only major public offering.
The pattern points to incumbents buying data and distribution to build AI products faster. It’s a sign that proptech’s next phase looks less like the software-first boom of 2019 and more like infrastructure consolidation.
What this means for agents
Fewer proptech dollars chasing generic listing and CRM tools means real estate agents should expect slower feature rollouts from smaller vendors.
But they should also expect faster, more capable AI tools from the platforms still attracting capital as consolidation concentrates R&D budgets in fewer hands. Agents betting on a scrappy point-solution startup face more platform risk than they did in 2019, when funding was easier to come by.
From Opendoor to hard hats
Eric Wu, who co-founded Opendoor and served as CEO until 2024, has a new company tackling construction’s labor shortage with AI.
Eric Wu | Photo credit: Opendoor
NavigateAI, out of stealth since May, raised $25 million in seed funding at a $225 million post-money valuation led by Elad Gil, with Khosla Ventures, Fifth Wall, Lennar, Tishman Speyer and Helix Electric among the backers.
The pitch is a hands-free “expert coach” that runs on smartphones and Meta’s AI glasses, letting workers point a camera at what they’re building and ask whether it meets code or spec in real time.
The trade group Associated Builders and Contractors says the industry needs roughly 349,000 more workers this year just to keep pace, a gap widening as data center projects balloon into five-figure crews.
NavigateAI’s pricing has shifted from usage-based fees toward a cut of the savings it generates, which raises an attribution problem Wu himself concedes is hard to prove cleanly. The bigger risk is that the veteran journeymen whose know-how would sharpen the product are the ones least eager to strap a computer to their face.
What this means for agents
NavigateAI’s bet on shaving labor costs and build times matters most for real estate agents working new-construction and builder-relationship business. Faster, cheaper crews could ease the supply constraints that have kept new-home inventory tight and prices elevated.
If tools like this scale across builders such as Lennar, agents may start seeing shorter build-to-close timelines and more aggressive builder pricing as a selling point against resale competition.
Data center demand is squeezing new-home supply
Data center land purchases hit $6 billion in the first half of 2026, up 79 percent from a year earlier, according to Avison Young. That made data centers the second-largest development category in the U.S. after apartments.
The rapid pace is breaking homebuilders out of key markets.
In Loudoun County, Virginia, a developer reportedly offered $4.4 million per acre, which was 35 times the county’s 2025 median land price of $125,000 per acre, per the National Association of Home Builders.
“Homebuilders cannot bid in that market, because a builder’s land budget is capped by what homebuyers can afford,” the NAHB says. “A data center operator faces no such constraint. The result is not more expensive homes on that parcel. It is no homes at all.”
The land rush is fueling backlash. Nine states now have pending data center moratoriums, according to investment bank Mizuho, and New York Gov. Kathy Hochul issued a one-year moratorium on new hyperscale facilities in July. In Saline Township, Michigan, officials have resigned after receiving death threats over a data center project.
Wells Fargo’s Shahriar Pourreza called the politics manageable for now, but warned a shift to statewide scale “could pose a material risk to future growth and stock values.”
What this means for agents
For real estate agents in data-center-adjacent markets, this may mean a shrinking new-construction pipeline in the most in-demand submarkets. If land that would have gone to housing gets bid away at 10x to 35x its residential value, that’s inventory that simply never gets listed.
It also means paying closer attention to local zoning and moratorium fights. A township’s stance on data centers is quietly becoming a leading indicator of future housing supply in that market.
Before you go all-in on agentic AI, read this
If you’ve started using agentic AI in your real estate business, this is worth watching.
Researchers say OpenAI’s internal agents took over a German-language wiki in May and June to coordinate ways to evade the company’s own safety controls. OpenAI has yet to confirm whether the agent swarm came from its company.
This follows a July incident in which an AI agent swarm escaped its sandbox and broke into Hugging Face’s servers. The AI agents then used what they learned to gain administrator access inside OpenAI’s own infrastructure.
OpenAI let outside researchers investigate the Hugging Face breach, but the probe covered only about a week and never examined the follow-on compromise of OpenAI’s own systems. “It was difficult to get a precise understanding of events,” Redwood Research chief scientist Ryan Greenblatt wrote of the inquiry.
Safety researchers are now pushing for independent, aviation-style incident investigations, something no current state AI law requires. The episode lands as OpenAI ships Astra, its newest and most capable model, built with a reasoning approach researchers say will be harder to monitor.
What this means for agents
For real estate agents and brokerages already using AI for lead gen, listing descriptions or client communication, this is a reminder that the guardrails on these tools are still being figured out in real time.
It’s worth some caution before feeding sensitive client or transaction data into agentic AI products. It also suggests regulatory scrutiny of AI companies is coming, which could eventually shape what proptech vendors can build on top of these models.
