Bedrock Robotics, the San Francisco startup built by former Waymo engineers, closed a $270 million Series B on February 4, 2026, to bring autonomous construction equipment from supervised prototype to fully unmanned deployment. The round, jointly led by CapitalG (Google’s growth fund) and Valor Atreides AI Fund, pushes Bedrock’s total funding past $350 million and represents one of the largest bets yet that the same AI systems that learned to navigate city streets can learn to move earth.

The investor list reads like a coordinated thesis on physical AI: NVentures (NVIDIA’s venture arm), 8VC, Eclipse, Emergence Capital, Perry Creek Capital, Tishman Speyer, MIT, Georgian, and others. When a real estate developer, a semiconductor giant, and a university endowment all write the same check, they are not chasing software margins. They are pricing in a conviction that autonomy’s next frontier is measured in cubic yards, not click rates.

The Team That Built the Playbook

The founding team matters here more than usual. Boris Sofman, Bedrock’s CEO and cofounder, previously cofounded Anki, the consumer robotics company that shipped over 3.5 million devices and raised more than $200 million before shutting down in 2019. He then spent roughly five years at Waymo as Director of Engineering and Head of Trucking, leading the autonomous freight program through its expansion into major U.S. cities. Cofounder and CTO Kevin Peterson also came from Waymo. The remaining cofounders, Ajay Gummalla and Tom Eliaz, both VPs of Engineering, round out a leadership bench that has collectively shipped autonomous systems at commercial scale.

Sofman’s trajectory tells a particular story. He built consumer robots, watched that company die from funding starvation, then spent five years inside Google’s most ambitious physical AI project. Now he is applying everything he learned to construction, an industry where the labor economics practically beg for automation. “The construction industry is being asked to build more than it can deliver,” Sofman said in the company’s Series B announcement. “Contractors are pulled across competing priorities with the same limited workforce and equipment.”

An Industry Bleeding Workers

The numbers confirm what Sofman describes. According to the Associated General Contractors of America, 92% of construction firms report difficulty finding workers, with 45% citing labor shortages as the primary cause of project delays. The ITIF reported in January 2026 that the industry faces a shortage of roughly 439,000 workers, driven in large part by the explosion of data center construction. Deloitte’s 2026 Engineering and Construction Outlook projects the gap will widen to 499,000 unfilled positions this year.

Construction also remains the deadliest sector in American industry by total fatalities, with 1,075 worker deaths recorded in 2024 according to OSHA data, accounting for 19% of all U.S. workplace fatalities. Over 60% of construction accidents occur within a worker’s first year on the job. This is not an industry resisting technology out of preference. It is an industry running out of people.

The pattern connects directly to a broader constraint we examined in our analysis of how physical limitations are gating AI’s real progress. Software intelligence is abundant. The bottleneck is getting that intelligence into the physical world, into machines that dig, pour, and grade. Bedrock is positioning itself exactly at that bottleneck.

What They Have Actually Built

Bedrock’s approach is retrofit, not replacement. The company installs a hardware rack on top of existing excavator cabs, equipped with LiDAR, GPS, inertial measurement units, eight high definition cameras, and an onboard computer. The system works across multiple excavator models from 20 ton to 80 ton machines. In November 2025, Bedrock partnered with Sundt Construction on what the company calls the construction industry’s largest known supervised autonomy deployment: mass excavation on a 130 acre manufacturing facility site in Phoenix, Arizona. The autonomous systems moved over 65,000 cubic yards of material by loading human operated articulating dump trucks using the same workflow as manual operations.

The company is now targeting its first fully unmanned excavator deployments with customers in 2026. If that timeline holds, Bedrock will have gone from stealth to unmanned commercial operation in under two years.

The Legal and Political Minefield

The Waymo pedigree looks strong on a pitch deck. It looks less convincing when you consider that Waymo itself is navigating wrongful death litigation, regulatory scrutiny in multiple states, and an operating model that still requires significant human oversight in edge cases. The self driving playbook transfers, but so do the liabilities. An 80 ton excavator operating without a human in the cab presents a legal and political exposure that no amount of venture capital can engineer away. OSHA has no framework for unmanned heavy equipment on active construction sites. Who is responsible when one of these machines kills someone? That question does not have an answer yet.

The political dimension cuts in two directions. The current administration’s infrastructure spending, data center buildout, and manufacturing reshoring create enormous tailwinds for autonomous construction. More projects, more demand, more urgency. On the other hand, these are the same political forces pushing for American jobs and union labor. Autonomous excavators do not pay union dues. They do not vote. The companies deploying them will face the same political friction that has dogged autonomous vehicles, except the workers being displaced wear hardhats and carry union cards, not rideshare apps.

The pattern is one we have tracked before in detail. Venture capital funds the technology that displaces the workforce, then frames the displacement as solving a shortage. The shortage is real today. It will not be real forever. When autonomous machines become cheaper per cubic yard than human operators, the narrative will shift from “we cannot find workers” to “we do not need workers.” The gains flow to the capital owners who funded the machines. The losses flow to the operators who used to drive them. The financial logic follows a similar arc to what private equity has done in other industries, as we documented with Snout’s $110 million raise to address a veterinary crisis that consolidation itself created. First you create the conditions for a shortage. Then you fund the solution to the shortage you created. Then you capture the margins on both sides.

What happens to the displaced operators is the question nobody on the cap table has an incentive to answer. The construction worker who used to run a $500,000 excavator does not become a software engineer. He becomes a gig worker or an unemployed statistic. The gains are real. The PR for everyday Americans is terrible.

What This Means for Everyday People

For general contractors hemorrhaging money to delays and labor gaps, autonomous excavation could compress timelines and reduce the single largest variable cost on a job site. For consumers, faster construction means lower costs on housing, infrastructure, and commercial development. Those are real benefits that affect real people.

For construction workers, the picture is darker than the investor presentations suggest. Bedrock frames autonomy as solving a shortage, not replacing a workforce. That framing holds as long as the shortage persists. The moment it doesn’t, every financial incentive in the system points toward fewer humans and more machines. The rich get richer. The workers who built the country’s infrastructure get a pink slip and a LinkedIn notification suggesting they learn to code.

The construction industry accounts for roughly $1.4 trillion in annual U.S. spending. It is one of the least digitized sectors in the global economy. The talent pipeline is shrinking, the safety record is grim, and the demand curve, driven by data centers, manufacturing reshoring, and infrastructure legislation, points in only one direction. Bedrock is not the first company to promise autonomous construction. Built Robotics started autonomous excavator trials in 2017 and eventually narrowed its focus to solar farm installation. Caterpillar has been running semi autonomous field trials for years without reaching full autonomy in construction applications.

What separates Bedrock is the pedigree of the team, the speed of execution, and the size of the capital behind them. Whether that is enough to solve a problem that has humbled every prior attempt is the $350 million question. The dirt will tell.

For inquiries and analysis contact laterstack@proton.me

Montreal-based Vention closed a $110 million USD Series D on January 27, with backing from Nvidia’s NVentures venture arm. The round was led by Investissement Quebec, the Quebec government’s investment division, with participation from Desjardins Capital and returning investor Fidelity Investments Canada. Total funding now exceeds $300 million CAD.

Vention physical AI is not another chatbot story. This is AI that moves things in the real world — and the numbers say it is working. The company hit $100 million CAD in annual run rate in late December.

What Vention Actually Builds

Founded in 2016 by CEO Etienne Lacroix and CTO Max Windisch, Vention offers a self-serve platform combining engineering software with plug-and-play industrial hardware. Manufacturing professionals design, order, and deploy automated equipment through the platform — no systems integrator required.

The pitch is “Zero-Shot Automation”: equipment that deploys without integration delays and works correctly on the first attempt. Vention claims its tools cut automation project timelines from months to days. The platform runs in more than 4,000 factories globally, including Boeing, L’Oreal, and Lockheed Martin.

Vention’s MachineMotion computing module runs on Nvidia’s Jetson system-on-chip, which explains the strategic investment. Nvidia is not just backing software AI. It is seeding the hardware layer that makes physical AI operational at scale.

The Physical AI Market Is Real Money

Goldman Sachs Research revised its 2035 humanoid robotics market projection sixfold — from $6 billion to $38 billion — citing breakthroughs in AI and plummeting hardware costs. Barclays goes further, projecting up to $200 billion by 2035 under optimistic scenarios. Manufacturing costs for humanoid robots have dropped 40%, and Goldman projects 250,000+ humanoid robot shipments by 2030.

Vention is not building humanoids. But it is building the platform layer that connects AI software to physical factory operations. The companies that control how AI meets the factory floor will capture value regardless of which robot form factor wins.

Why Nvidia Keeps Showing Up

Nvidia’s NVentures arm has been quietly assembling a physical AI portfolio. The pattern: back companies that put Nvidia silicon into real-world systems. Jetson chips in Vention’s hardware. Jetson chips in autonomous vehicles. Jetson chips in warehouse robots. Every physical AI deployment is another recurring customer for Nvidia’s edge computing stack.

The roughly 330-person company plans to use part of the capital for EMEA expansion, targeting European manufacturers facing labor shortages and regulatory pressure to automate.

Laterstack Editorial Take

Laterstack exists to sharpen critical thinking by connecting tech, policy, and power to everyday life — across class, industry, and influence. The people funding physical AI are not solving a labor shortage. They are repricing labor permanently. When Nvidia backs a platform that runs on its own chips, that is not venture capital — it is vertical integration with a press release. Lawmakers drafting workforce policy and industry leaders allocating automation budgets need to see this clearly: the timeline just accelerated, and the capital concentration in AI is now spilling into the physical world.

What This Means for Everyday People

Physical AI is where automation stops being abstract and starts replacing specific jobs. Goldman’s base case of 250,000 humanoid shipments by 2030 — nearly all industrial — is a labor market event, not just a technology milestone. The assembly line worker, the warehouse picker, the quality inspector face the most immediate displacement.

The upside is real: safer working conditions, lower manufacturing costs, reshored production. But the transition will not be evenly distributed. Vention’s $110 million is modest by AI standards. But the convergence of Nvidia, a sovereign fund, and institutional investors on factory-floor AI tells you where the next phase is heading. Not chatbots. Not image generators. Machines that move things.

---

GRAND JUNCTION, CO – Sunday Robotics, a robotics startup founded in April 2024, has unveiled Memo, an autonomous home robot capable of completing everyday household tasks like clearing the dinner table, loading the dishwasher, and even operating an espresso machine.

Unlike most robotics startups that rely on teleoperation or synthetic data simulations, Sunday Robotics developed a proprietary glove system to train Memo. Human operators wear gloves that mimic the robot’s Lego-like hands to perform tasks, generating real-world data for Memo to learn from.

Efficient and Cost-Effective Training

CEO Tony Zhao explained that the glove-based method provides “two orders of magnitude higher capital efficiency compared to teleoperation ($200 vs $20,000).” The approach allows for large-scale data collection without physically moving Memo, making the training process highly scalable. Currently, the startup employs more than 500 human data collectors across the United States to improve Memo’s dexterity and AI performance.

Breaking New Ground in Dexterity

Training robots to handle delicate objects has been a longstanding challenge in the robotics field. During live demonstrations, Memo successfully lifted fragile wine glasses, folded socks, and loaded the espresso machine without breaking any items. Zhao highlighted that replicating human hand capabilities, which involve thousands of touch receptors, is a major engineering feat.

“Today, we present a step-change in robotic AI,” Zhao said. “Memo demonstrates that autonomous home robots can perform complex tasks efficiently and safely.”

Alternative to Traditional Training Methods

Most robotics companies train AI through teleoperation, where humans control robots via joysticks, or by using simulated data. Sunday Robotics’ glove system bypasses these methods, providing more accurate and context-rich training data.

Michael Cheng, co-founder of Sunday Robotics, noted that “relying solely on teleoperation could take decades to gather sufficient training data for a robot like Memo.” By contrast, the glove method allows for widespread, distributed training.

Looking Ahead

The startup plans to continue refining Memo’s abilities, aiming to expand its applications in home automation. With breakthroughs in dexterity and autonomous AI learning, Sunday Robotics positions itself as a leader in the next generation of consumer robotics.

Related stories on Laterstack:

solar startup Africa is quietly building the future the rest of the world keeps talking about

Startup CEO accidentally names new project after famous OnlyFans creator

In a small factory outside Shanghai, a new kind of worker is learning how to move. Its name is AgiBot, a humanoid robot that can be trained by people in real time to assemble electronics, test parts, and pass them down a production line.

AgiBot’s idea is simple but radical. Instead of teaching robots through endless simulation, the company pairs each machine with a human trainer who guides it through a task for about ten minutes. Then the robot learns to repeat it alone. This process, known as real-world reinforcement learning, blends human intuition with machine precision.

It is already being tested by Longcheer Technology, a major Chinese manufacturer that builds smartphones, VR headsets, and other electronics. The AgiBot system allows robots to take on repetitive but high-volume tasks, such as moving components from testing machines to assembly lines, while still adapting to shifting workflows.

Unlike traditional industrial robots that perform rigid motions, AgiBot’s machines learn through touch, vision, and trial. They are not coded to complete one routine forever. They evolve through repetition, much like human workers do.

Behind the system is Jianlan Luo, a UC Berkeley researcher turned entrepreneur. Luo helped pioneer human-in-the-loop robotics research in California before bringing the idea home to Shanghai. At AgiBot, his team of engineers and teleoperators trains robots for different factories across China, from electronics to consumer goods.

Training robots this way takes a surprising amount of human effort. In AgiBot’s training center, hundreds of operators guide robot arms through tasks, generating data that improves the company’s learning models. It is part of a growing trend in robotics where human labor fuels machine intelligence.

“Robots are not replacing workers,” said Yuheng Feng, an AgiBot representative. “They are learning from them.”

Each robot session creates more adaptable code, faster learning cycles, and smarter machines that can move to new production lines without weeks of reprogramming. For manufacturers, that flexibility is gold.

China’s government has made robotics a core focus in its latest five-year plan, alongside artificial intelligence and automation. The country already operates more industrial robots than the rest of the world combined, giving startups like AgiBot a vast playground for scaling quickly.

Experts say this fusion of human skill and robotic learning could define the next phase of manufacturing. “AgiBot is using some of the most advanced reinforcement learning seen outside a lab,” said Jeff Schneider, a Carnegie Mellon roboticist. “If it works as described, it could reshape how factories operate.”

Across the Pacific, startups in the United States are racing to catch up. Companies like Physical Intelligence and Skild are developing similar models that teach robots to adapt to new shapes, arms, and environments. But China’s scale and production speed may give AgiBot a lasting advantage.

AgiBot’s long-term goal is to create humanoid robots that can walk, handle tools, and work alongside people safely. For now, its focus remains clear: give robots a human touch and let them learn from the people who know the work best.

The quiet revolution is already underway, and it is not happening in a lab. It is happening on a factory floor where humans and machines are learning to build the future together.

Read the original coverage at WIRED.

Explore more Laterstack startup insights:

• Cluely AI cofounder says startups fail because engineers cannot make viral content

• Saquon Barkley joins Nvidia in $1.4 billion investment in data center startup Crusoe

[thb_gap height=”150″]