Most of the AI conversation has centered on chatbots, coding assistants, and language models. The Bay Area startups below are doing something different. Founded in the last two years, they are applying machine intelligence to construction sites, factory floors, freight networks, and physical robot systems, industries that have resisted automation for decades and where the stakes for getting it wrong are measured in tons of steel, delayed infrastructure, and missed energy targets. This is the Bay Area cohort that is taking AI off the screen and into the world.
Bedrock Robotics: Autonomous Construction, Retrofitted
America needs nearly 800,000 additional construction workers over the next two years and has 40 percent of its existing workforce approaching retirement. Bedrock Robotics, founded in San Francisco in 2024, is building the autonomous systems to fill that gap. CEO Boris Sofman led autonomous trucking at Waymo for five years before co-founding the company alongside CTO Kevin Peterson, and VP of Engineering Ajay Gummalla, both also former Waymo engineering leaders. The Bedrock Operator retrofits existing excavators with cameras, LiDAR, GPS, and onboard compute in a single shift, no permanent modifications required. The company emerged from stealth in July 2025 with $80 million and raised a $270 million Series B in February 2026 led by CapitalG and Nvidia’s NVentures, bringing total funding to $350 million at a $1.75 billion valuation.
Physical Intelligence: Foundation Models for Any Robot
Physical Intelligence was founded in San Francisco in 2024 by a team drawn from Google DeepMind, UC Berkeley, and Stanford. CEO Karol Hausman was a Staff Research Scientist at Google DeepMind and adjunct professor at Stanford. Co-founder and Chief Scientist Sergey Levine is an associate professor at UC Berkeley whose lab pioneered deep reinforcement learning for robotic manipulation. Co-founder Chelsea Finn is an associate professor at Stanford specializing in meta-learning and sim-to-real transfer. COO Lachy Groom previously led product at Stripe before becoming an angel investor. The company builds general-purpose AI models that teach robots to perform physical tasks, from folding laundry to assembling packaging to making espresso, across different hardware without task-specific programming. Levine has described the ambition as building the equivalent of ChatGPT, but for robots. The company raised $70 million in seed funding in March 2024, $400 million in Series A in November 2024, and $600 million in Series B in November 2025 led by CapitalG, bringing total funding to $1.1 billion at a $5.6 billion valuation. As of March 2026 it is in talks to raise approximately $1 billion at a valuation above $11 billion.
Korso: AI Agents for Manufacturing Operations
Manufacturers still process incoming requests for quotation manually, follow up on supplier quotes by phone, and coordinate production disruptions through email threads. Korso, a YC-backed company in the Bay Area founded in 2026, builds AI agents that handle this operational layer automatically. The company was co-founded by Daichi Hiraoka, Alex Liu, who studied computer science at Notre Dame, and Martin Pan. Its agents plug into existing ERP and CRM systems, process incoming RFQs, generate quotes, track supplier responses, and manage production coordination, with verification layers and full audit trails built in to meet the accountability standards enterprise manufacturing environments require.
Lanesurf: Voice AI for Freight
Covering a single freight load typically means calling 50 or more trucking companies one at a time, checking compliance, availability, and rates through a process that takes hours. Pratham Bansal and Sarthak Singh Chauhan, both IIT Delhi alumni, founded Lanesurf in San Francisco in 2025 to automate that process end to end. Bansal previously deployed AI systems for Fortune 500 enterprises at Booz and Company and built a high-speed inference application for compute-heavy machine learning models that was acquired by an Nvidia-backed Bay Area company. The Lanesurf platform runs parallel voice negotiations across dozens of carriers simultaneously, handles compliance checks, manages exceptions, and books shipments in minutes. In a live demonstration at the 2025 Future of Freight Festival, the system spoke to 96 trucking companies simultaneously and booked a shipment in under 10 minutes. The platform is in production daily for logistics companies managing over $10 billion in annual freight, and is backed by General Catalyst, Y Combinator, and Jawed Karim, co-founder of YouTube.
Tensr: Autonomous Robotic Factories
Eric Berndt, Adith Sundram, and C.K. Wolfe met as graduate robotics researchers at UC Berkeley’s AI Research lab, where they worked on DARPA contracts and led an autonomous IndyCar team that raced full-scale vehicles at 160 miles per hour. Berndt holds MS and BS degrees in electrical engineering and computer science from Berkeley and serves as CEO. Sundram holds MS and BS degrees in mechanical engineering from Berkeley and serves as CTO. Wolfe is completing a PhD in AI and robotics at Berkeley. Together they founded Tensr to build fully autonomous robotic factories, with the stated ambition of making hardware manufacturing as scalable as cloud infrastructure. The team also brings operational experience from Rockwell Automation, aerospace manufacturing, and FDA-regulated medical device production. Tensr is YC-backed and based in Berkeley.
Charge Robotics: Robots That Build Solar Farms
Banks Hunter and Max Justicz met at MIT, where both studied robotics and computer science, and founded Charge Robotics in 2021. Hunter is CEO. Justicz is CTO. The company’s Sunrise system automates the assembly and installation of solar tracking hardware and panel racking, the most labor-intensive portion of utility-scale solar construction, directly addressing a labor bottleneck that is slowing the energy transition. Its first commercial deployment with SOLV Energy completed in February 2024. Charge Robotics is backed by Founders Fund, Energy Impact Partners, Lux Capital, and Y Combinator.
