Maven Robotics just came out of stealth with $100 million in funding and a plan to put hundreds of warehouse robots to work. According to TechCrunch AI, the startup raised the round from RoboStrategy, LocalGlobe, Vine Ventures, and XTX Markets Ventures, and it’s already running as many as eight robots at a customer site 16 hours a day with 99% or higher uptime.
That’s a striking debut for a company that, in 2024, was little more than “a cartoon of a robot and a team of people,” as CEO and co-founder Hamza Derbas told TechCrunch AI. The team won its first major contract by doing something most robot startups don’t bother with: they asked to walk the customer’s factories and warehouses before pitching anything.
What Maven Actually Does
Maven’s robots aren’t humanoids. They sit on wheeled bases that move up to 10 miles an hour, with two arms that lift up to 30 kilograms using vacuum suckers to grab and stack boxes.
The core job is “mixed palletizing.” Here’s the workflow in plain terms:
- Pallets of boxed goods arrive at a distribution center from different factories.
- Retail stores order a custom mix of products based on real-time demand.
- The robot builds a fresh pallet with that exact mix, then sends it out.
- Today, humans do this by running around the warehouse grabbing one of this, one of that.
Maven’s system hooks directly into the warehouse management software on one end and loads trucks on the other. Derbas frames it as taking on the whole task, not solving “a single robot problem.”
Why This Matters
The robotics field is crowded, and most companies lead with the hardware or a research agenda. Maven is betting on the opposite: industrial operations first, ROI above all.
That positioning came straight from its founders’ backgrounds. Derbas spent nine years at Apple’s special projects group, widely believed to have been its self-driving car effort before it shut down in 2024. He started Maven that same year with his brother Khalid, a former private equity investor who now serves as CFO. Like many physical AI companies, Maven leans on self-driving veterans, because those teams built the most mature pipelines for training autonomous hardware on real-world data. The loop matters: data comes back from working robots within minutes or hours, the team retrains, runs ablation studies, redeploys, and repeats.
Jack Pearson, the RoboStrategy investor who backed the company, told TechCrunch AI that Maven’s edge is its industrial systems background rather than a research culture “optimized for learning.”
The Competitive Angle
Derbas didn’t shy away from comparisons. He named Agility Robotics, the company going public this fall in a $2.5 billion SPAC deal, as the most similarly positioned firm. But Agility’s robots stand on two legs, and Derbas argues that makes “zero sense” for warehouse work, calling bipedal designs complex, unreliable, and costly. His pitch is blunt: “ROI is the name of the game here.”
What stands out to me is how contrarian that is. The industry keeps chasing general-purpose humanoids, and Maven is going the other way, task by task, market by market.
What Comes Next
With the new capital, Maven plans to build 250 of its third-generation robots and start designing a fourth. Palletizing alone is an $80 billion market, but the next tasks the company wants, material handling and eventually fabrication, will need manipulation skills that don’t exist yet.
To get there, Maven is collecting more data, tapping third-party providers, and has even built pincer-like gloves so humans can demonstrate the gripper motions it wants to teach. The strategy is deliberate: pick sizeable problems one at a time, each a multibillion-dollar market, and let the data pile up.
The risk Derbas acknowledges is real. A powerful new physical AI model from a frontier lab could leapfrog the whole task-by-task approach. His counter: “We’re not in the race for models, we’re in the race to solve industrial labor.”
Whether that focus holds as Maven scales is the question worth watching. For now, it’s shipping robots that work, which is more than most of the field can say. Full details are available at the original TechCrunch AI report.