Robotics and Autonomy

Zornitsa Todorova opens by widening the frame: physical AI isn’t just headline-grabbing humanoid robots, it’s a category Barclays breaks into four technologies (humanoids, autonomous vehicles, industrial automation, and drones) that together could scale from a small market today to roughly a trillion-dollar opportunity by 2035. Her framing for why capital should follow this space isn’t “dual use,” a term she argues undersells what’s happening, but “multiplicity”: AI compute spending, energy availability, critical minerals, and defense budgets all have to converge for the category to scale.

Scott Sanders pushes back on “dual use” from a different angle entirely: in his experience, it’s rarely real for startups, who need to commit to either the commercial or defense sales motion because the two move at completely different speeds and require different organizational muscle. Ira Renfrew adds the founder’s perspective, arguing recent advances in generative AI and reinforcement learning have meaningfully shortened how long it takes to train a robot for real-world tasks, from months of developer work to overnight retraining.

The panel also debates where value actually accrues in the robotics supply chain. Todorova frames it as “the three Bs,” brains (AI models), brawn (actuators, sensors, and the mechanical dexterity to interact with a world built for humans), and batteries (safe, sufficient power), arguing brawn and batteries get less attention than they deserve relative to AI models. On workforce, all three agree the technical and vocational education pipeline needs investment to keep pace. And when Ken asks what advice they’d give states and cities courting these companies, the answer is less about any single incentive and more about policy consistency over time.

Key Takeaways from the Conversation

• Physical AI is bigger than humanoids. Barclays frames it across four categories, humanoid robots, autonomous vehicles, industrial automation, and drones, together representing what could become a roughly trillion-dollar market by 2035.
• Panelists disagree on “dual use.” One guest reframes it as “multiplicity,” the convergence of AI compute, energy, critical minerals, and defense spending; another argues dual use isn’t practically real for startups, who need to commit to one customer base at a time.
• The “three Bs” framework is a useful lens for where value builds up: brains (AI models), brawn (the actuators and sensors that give a robot dexterity), and batteries (safe, sufficient power) — with brawn and batteries described as the more overlooked of the three.
• Drones are currently the most investable, mature corner of the market, having roughly doubled in market size over the past five years.
• The panel’s advice to states and cities: policy consistency matters more than any single incentive, since frequent shifts in direction make it harder for companies to plan long-term investment.

Why This Conversation Matters

Robotics and autonomy sit downstream of nearly every theme this season, reshoring, workforce, capital formation, making this conversation a useful bridge between the industrial policy conversations and the technology actually being built to support it.

About the Guests

Zornitsa Todorova is Head of Thematic FICC Research at Barclays.

Scott Sanders is Chief Growth Officer at Forterra, a defense technology company based in the DC region.

Ira Renfrew is Co-Founder of Familiar Machines & Magic. He has spent over a decade in robotics, with prior roles at iRobot, Amazon, and Outrider.