By: Aidan Madigan-Curtis and Ryan Gibson

The United States is facing the perfect storm: A trillion dollars of data center/AI infrastructure that must be built for us to win the AI race, and another two to four trillion spent on roughly 1.6M in new residential housing units under construction — per year — in the next decade. At the same time, more than 40% of the construction workforce is retiring over the same time period, while other sources of construction labor, including skilled migrants, are facing disruption and restructuring.

Slow construction means three things:

1.  Less supply / onshore manufacturing
2.  A loss of AI competitiveness
3.  A perpetuation of our housing affordability crisis

There is zero chance that the United States achieves its manufacturing, AI, and residential housing stock/cost goals, without radical productivity improvements — the kind you can only get through a truly revolutionary form of technology. 

Enter: Bedrock Robotics. Bedrock develops advanced autonomy for specialized heavy machinery, starting first with construction equipment like excavators. Each machine is a 100-ton vision for our future.

Great Investments Feel Obvious In Retrospect, But Start Non-Consensus

At first blush, it feels obvious. Autonomous heavy equipment should follow autonomous driving: If society craves a robot chauffeur to the tune of over a hundred billion dollars invested, why wouldn’t it make sense to automate heavy, dangerous equipment that operates in the harshest of environments (think: 120 degree weather highway construction in Texas, infrastructure in Alaska, or gigafactory pad-leveling in Nevada), especially given utilization-sensitivity and skilled labor shortages have a profound impact on construction ROI? Sweltering or frigid, our landmass under development in this country is fraught, and construction is actually the #1 cause of workplace mortality in the U.S.

Far From a No-Brainer

If it is so clear that autonomous heavy equipment was a smart bet, why hasn’t it been done? The answer: True autonomy has been brutally hard to achieve. 

In fact, it has been almost two decades since the concept of a self-driving car entered the Zeitgeist. Kids graduating from college in the late 2000s questioned whether our children would even need drivers’ licenses! Yet, it’s almost 2026, more than 100 billion invested, and only one company (of many) has successfully begun scaling their autonomous driving business. Why? Because it’s SO HARD to get the technology right.

Finally, Waymo cracked the code, in no small part due to the approach Boris Sofman and Kevin Peterson (Bedrock Co-Founders) had taken throughout their long careers building autonomous systems: Leveraging machine-learning approaches where the system could self-learn — for example using experience to learn to interpret sparse, far away data using a real-time feedback loop that would fine tune its performance in real-time (without human involvement). They were two of the key engineering leaders at Waymo who finally deployed safe and reliable self-driving for urban streets.

Qualified To Do The Impossible

Bringing autonomous technology to hard-to-penetrate industries is in Boris and Kevin’s blood.

When most of us were learning how to use the internet and discovering the early joys of smartphones, Boris and Kevin were building “Crusher,” their DARPA FCS PhD project at Carnegie Mellon.

Boris and Kevin each founded, led and sold a different company in the robotics space between the DARPA Grand Challenge and joining Waymo. After working together on delivering scalable autonomy to city streets, they decided their next act was going to be as co-founders, applying the lessons from the first hundred million miles driven at Waymo to a large, adjacent and inevitable market. The question was which market? 

Most people in the AV space have experienced and architected failures, not successes. Boris and Kevin are the rare exceptions: scale-tested leaders who have committed their careers to advancing the frontier technologies in robotics such as multi-domain autonomy and reinforcement learning.   

We met Boris and Kevin through parallel tracks, Kevin had a long standing relationship as a portfolio founder of Eclipse’s, working closely with Seth Winterroth. Boris and I (Aidan) met early in his "what's next” phase, and my experience building and scaling Samsara from the company’s earliest days provided endless fodder for thought-partnership and exploration. At Eclipse, we knew immediately we wanted to build with them, and committed early on to finding an idea with the perfect intersection of founder authenticity, a massive opportunity, our conviction, and experience as a firm. This stage of collaboration is extremely special but familiar for both Ryan and I, as we’ve had the privilege of partnering at the inception stages with some incredible founders over the years. It didn't take too many curiosity-led whiteboard sessions together to unearth the obvious fit to apply the learnings and breakthroughs in on-road autonomy, to the dull, dirty, and dangerous work in our most critical industries.

The Ultimate Challenge Meets The Ultimate Reward: Deciding To Focus On Construction

Over a naturally evolving months-long partnership we all put pencil to paper to down-select which industry was the best first start. What would be Bedrock’s equivalent of “Books” for Amazon? We carefully researched and considered the dynamics that would make the ideal industry; margin/cost structure (how much was labor vs. inputs vs. overhead), go-to-market (repeatable?), size-of-market, buyer dynamics (stuck in pilot hell?). We know from experience just how critical picking the right first market and buyer is. Although Bedrock will shape all critical industries over time (like agriculture, mining, etc.), we found a love for construction due to its unique combination of sheer size, critical labor shortage, and cost structure we could affect with autonomy. 

With construction within our crosshairs, I (Ryan) was able to draw on my journey founding a construction-tech company which sold productivity software to contractors across North America - the kind of experience that leaves your brain packed with how the industry works, and your phone full of project managers, innovation teams, developers and builders who actually pick up. We did the work together to have dozens of blunt conversations with the industry and on‑site walk‑arounds, equipped with hardhats and safety boots. We uncovered valuable data regarding the diversity of work on projects, the practical labor shortages businesses face, and the massive economic benefit real construction projects would gain from autonomy if we could compress schedules, improve work reliability, and safety. Working obsessively with the industry from day 0 let us take an analytical approach to mapping scopes of work, project types, machine utilization and cost structures to lock-in on the central figure to begin the autonomous march on: the excavator.

Excavators are like the black belts of construction machinery. Four actuators. Multiple axes. Digging, trenching, grading, moving earth with a giant steel hand. They’re essential and unforgiving. Job sites are dynamic, GPS is fickle, and the payloads are lethal if mishandled. An excavator has to see dust clouds, dangling rebar, and trench walls—then coordinate hydraulic muscle with centimeter precision. The sensor fusion, control theory, and real‑time planning load are an order of magnitude above most robotics projects. But the prize is enormous: 24/7 operation, fewer injuries, and throughput jumps that turn multi‑year projects into single‑season realities. A problem worthy of the most experienced robotics and autonomy team on earth to solve.

So Bedrock started there. Because with the cutting edge of autonomy, when you can solve for an excavator, you can solve for almost anything with hydraulics and treads. That’s how Bedrock’s platform is built—generalizable, retrofit-first, sensor-agnostic, deployable across the $100B+ installed base of modern machines already in the wild.

Quick and Dirty

The inception-stage groundwork to go after a monster, well-validated problem has paid off. It's a common belief that the construction industry doesn't move quickly, but Bedrock’s industry pull and quality of partnerships have broken the mold. We know our critical industries move quickly when they have an existential problem on their hands, otherwise, they can’t afford to step away from intensive day-to-day operations to entertain far-out ideas. Bedrock has clearly struck a chord, signing incredible partnerships in under 12 months since inception, representing over $5B of earthwork collectively, with Sundt Construction (Bedrock’s first partner), Zachry Construction, Champion Site Prep, and Capitol Aggregates. When successful, Bedrock will help hospitals, roadways, manufacturing facilities, energy projects, and datacenters built by these incredible companies move faster and safer.

But when a $13T opportunity comes lining up at your doorstep in droves, it's time to DELIVER. Bedrock is fully leveraging the $100B of tuition paid for and supply chain built by the on-road autonomy giants who came before. Ajay Gummala (Bedrock founding team) put it perfectly: “Even we have been surprised at how fast we’ve been able to move - it was only four or five months from the day we hired the first employee to the day the first machine was up and running, with all sensors integrated and data ingested. Then it was another two months to actually demonstrate the machine doing a full task autonomously by machine learning model.”

Why This Is Bigger Than Earthmoving

Autonomy for heavy equipment isn’t a feature upgrade; it’s a new computing layer for the physical world. Once Bedrock’s intelligence becomes the default spec on yellow iron, the company graduates from “product” to platform—the control layer for how we shape land, pour concrete, lay pipe, and mine raw materials. That impact is measured in trillions, not billions. 

Beyond just the sheer economic impact, the stakes are nation‑scale.

• If we can’t clear pads, we stall the AI compute build‑out.
• If we can terraform our earth for our transportation systems, logistics incurs a society-wide-toll to growth.
• If we can’t trench utilities or dig foundations, the housing backlog widens.
• If we can’t mine safely here, we stay hostage to foreign supply chains.

Humanity’s new super‑power is teaching steel to think. The excavator is chapter one. Loaders, dozers, and graders are queued up. Agriculture and mining follow. Every time Bedrock adds a machine class, its autonomous brain gains new data—widening the moat and shrinking the payback period.

We have a chance to turn a looming productivity crisis into a renaissance of Human capacity. To do it from day 0 alongside this exceptional team of humans, Boris, Kevin, Ajay, Tom, and the whole Bedrock family, has also been our unique privilege. That’s why we invested. That’s why we’re still on site at sunrise.

Let’s Build.
Aidan and Ryan

Read more on today's news.