Eric Wu, who built and ran Opendoor before stepping away in 2022, has had his new company, NavigateAI, out of stealth since May. The startup is building AI copilots that give construction workers real-time, hands-free guidance through smartphones and Meta's AI glasses. It officially launched in late May with $25 million in seed funding at a $225 million post-money valuation led by Elad Gil, with participation from Khosla Ventures, Fifth Wall, Lennar, Tishman Speyer, Helix Electric, and angels including Tony Xu, Apoorva Mehta, and Brian Armstrong.
The product lets a worker point a camera at what they are building and ask whether it is installed correctly, whether torque is right, or whether it meets code. NavigateAI can pull up building specs, manufacturer manuals, and company policy in real time. Wu says the hands-free experience is measurably superior, and the company is working with Meta to get the glasses safety-certified for environments where protective eyewear is required. The startup is also partnering with AIM, a Meta-backed fiber installation trade school, to reach workers before they reach a job site.
The labor shortage is severe. Associated Builders and Contractors has said roughly 349,000 additional workers would be needed this year just to keep pace with demand. Data center projects have grown from roughly 750 workers at peak to 4,000 or 5,000 workers apiece, with Meta’s Hyperion campus reportedly requiring about 5,000 workers and OpenAI’s Stargate site involving 6,400. Kelly has said 90% of data center operators name staffing shortages as a critical constraint. NavigateAI’s pricing has shifted from token-plus-margin to a share of value created, such as capturing roughly 20% of a $20,000 savings if a builder reduces a home’s all-in cost from $300,000 to $280,000.
What remains unknown is how quickly experienced workers will adopt the technology. Wu says adoption curves split sharply between younger workers, who embrace the product, and 30-year journeymen, who trust their own instincts and are not lining up to strap a computer to their face. The longer-term data play—every job completed generates labeled egocentric video—is also still unproven at scale.