CapitalG is thrilled to co-lead a Series B investment in Bedrock Robotics, the autonomy platform for construction equipment. Bedrock is building towards fully autonomous fleets as the new job site standard.
The Physical Economy’s Bottleneck
The most important innovations of the next decade won’t happen with bits behind a screen, but with atoms in the physical world.
Construction is a massive industry, representing nearly 15% of global GDP. That translates into a whopping $10 trillion growing at a 5-6% rate over the next 5 years. To support these projects, heavy equipment OEMs like Caterpillar drive over $40B in annual sales, supporting an active equipment base representing $500 billion of embedded spend. This hardware powers a U.S. labor pool valued at $650 billion annually.
However, the demand for infrastructure is outstripping our ability to build it:
Data Centers: Analysts forecast $1.7 trillion in global data center spend by 2030, compounding recently at a 60% annual growth rate
The Power Grid: Aging transmission lines (averaging 40 years old) are buckling under peak loads, with a five-year backlog for new capacity
The Manufacturing Rebound: U.S. manufacturing construction spend has doubled since 2022, fueled by the CHIPS Act and a global push to "re-shore" critical supply chains and renewable energy production
The industry is hitting a breaking point. To meet demand, the construction industry needs more than 800,000 new skilled workers over the next two years, and the gap is expected to widen as more than 40% of the workforce approaches retirement over the next decade. As such, demand for vehicles outpaces available skilled operators as fewer people are entering skilled construction trades, leading to underutilized equipment and costly operational gaps. Beyond productivity, there is a human cost: Over 100K annual injuries involve off-highway vehicles in the U.S. alone, costing the construction industry $11 billion per year, plus the obvious human toll.
The gap between our global ambitions in the physical world and our capacity to build them has never been wider – but that dynamic is now shifting.
Advances in AI perception, edge compute, and data-driven learning have crossed a threshold where autonomy (think Waymo for construction) can be deployed continuously, learn from live production environments, and deliver unit-economic value for construction jobsites. Construction operators are actively seeking solutions that improve safety for their team, unlock step-function productivity gains, and enable work to continue when people cannot be on site.
This is the moment Bedrock was built for.
Bedrock Robotics: Autonomy for the Built World
Bedrock builds the software and hardware that allow heavy machinery to operate without human drivers. Bedrock specializes in "always-on" autonomy for industrial environments like construction and mining sites. From the outset, the team approached autonomy as a system-level challenge, integrating perception, planning, control, and safety into a cohesive platform.
Construction environments pose some of the most demanding challenges in autonomy: constantly changing conditions, tight coupling between software and the physical world, and little tolerance for failure. Bedrock is building an adaptable autonomy engine designed to master diverse workflows and equipment types. By creating a system that learns and generalizes across different operating contexts, we are moving industrial automation past simple, repetitive tasks and toward the high-maturity, high-uptime standards of modern on-road autonomous vehicles.
Today, Bedrock’s platform is already operating at meaningful scale, deployed across fleets with leading operators such as Sundt Construction, Zachry Construction, and Champion Site Prep, running in live production environments. What differentiates Bedrock is not only the progress achieved to date, but the depth of its vertically integrated autonomy stack:
Universal Hardware Integration: A "plug-and-play" hardware kit that connects directly to a machine’s existing hydraulics and electronics. It allows companies to transform their existing fleet into an AI-operated autonomous fleet without needing to buy new equipment.
Adaptive Machine Intelligence: Bedrock’s system learns from every movement, allowing machines to navigate changing worksites and handle unpredictable obstacles. By training on full operations rather than just simple "if/then" rules, Bedrock’s equipment moves with the fluid, natural precision of an experienced human operator.
Operational Control Software: A digital command center where managers can turn project plans or CAD files into specific tasks. The software will coordinate everything from a single repetitive job to complex, multi-machine workflows, ensuring the fleet executes the high-level project goals safely and efficiently.
Together, these layers form a cohesive autonomy platform designed not just to automate individual tasks, but to serve as a scalable operating layer for physical-world systems.
The Operating Layer for Construction
In conversation after conversation, customers consistently confirmed that Bedrock’s autonomy and machine intelligence are delivering immediate, tangible impact across their operations. Customers highlighted meaningful improvements in safety, equipment utilization rates and operational flexibility.
That operational flexibility is key: Autonomy expands what is possible. It enables work to continue in environments that are remote, safety-critical, or constrained by staffing and scheduling, transforming idle or underutilized assets into dependable capacity. Bedrock’s success is driving a market shift: The ecosystem, from operators to OEM partners, is moving beyond experimental autonomy and embracing these proven systems within their operations.
Bedrock’s ambition is broad and extends well beyond individual machines. Bedrock is starting with the excavator because it is the most complex "anchor" on any site. Once mastered, the platform will coordinate entire workflows. For a customer, this means a single site manager could oversee an autonomous fleet where an excavator automatically clears a trench, loads a waiting haul truck, and signals a dozer to backfill – all synchronized by Bedrock to ensure the site moves safely and at peak efficiency without constant manual intervention.
Founded by Builders of Real-World Autonomy
Bedrock was founded by Boris Sofman (CEO), Kevin Peterson (CTO), Ajay Gummalla (VP Engineering), and Tom Eliaz (VP Engineering), leaders who have built and deployed autonomy in some of the world’s most demanding real-world environments. Boris and Kevin previously led autonomous trucking efforts at Waymo, bringing safety-critical systems into large-scale operations. As Bedrock took shape, the team brought on Laurent Hautefeuille (COO), another seasoned operating leader with deep experience from Uber Freight, who helped build the business from $0 to >$5 billion in annual revenue.
The incredible Bedrock founding and early engineering team are world-leaders in frontier autonomy given their leadership roles at Waymo, Caterpillar, Uber, and beyond. This remarkable team exudes founder-market fit given their experience building and operating frontier autonomous systems safely, reliably, and at scale in live environments.
What has always stood out to us about this team is not only the strength of their technical backgrounds, but also their thoughtful approach to building real products. The Bedrock team brings a clear understanding of what it takes to deliver autonomous systems that can operate day after day in the field. This level of operational execution is rare in autonomy and essential for physical-world systems, where safety, reliability, and uptime are foundational.
Bedrock is not simply advancing autonomy as a technology; it is laying the groundwork for durable infrastructure that can transform how physical work gets done. We are excited to partner with Boris, Kevin, Laurent, Ajay, Tom, and the entire Bedrock team as they bring AI and autonomy into the physical world.