VC experts on why Physical AI funding is heating up | E2333

What It Covers

Physical AI venture funding reached nearly $50B in H1 2026—up 80% year-over-year and exceeding 2022–2024 combined totals. VCs Paige Dougherty (Behind Genius Ventures) and Caitlin Holloway (776) discuss orbital data centers, defense tech adoption, EU AI regulation, frontier model data sovereignty risks, and early-stage strategies for capital-intensive hardware companies.

Key Questions Answered

• •Physical AI funding inflection: $45B across 450 deals in H1 2026 versus under $5B in all of 2023 represents a structural shift, not a cyclical spike. Aerospace and defense lead category allocation. Early-stage investors entering hardware now face a different calculus: dilution from multi-billion-dollar raises can compress seed ownership to near-zero, making verticalized applications—where companies reach revenue faster with less capital—a more defensible entry point than general-purpose robotics platforms.
• •Verticalized physical AI as dilution hedge: Seed funds investing in general-purpose robotics risk severe ownership dilution as companies raise billions. Behind Genius Ventures targets verticalized physical AI applications—like Maniva's factory-floor quality assurance AI or Knox Metals' next-day metal servicing platform—where commercialization timelines are shorter, capital requirements are lower, and asymmetric information advantages at pre-seed and seed stages remain intact before valuations become astronomical.
• •Defense tech bottleneck has shifted: The historical barrier to defense investment was winning government contracts, which took years. That barrier is largely solved. The new bottleneck is a 30-person startup successfully executing against a large federal contract, managing sudden non-dilutive capital inflows, and building the operational capacity to sustain and renew that revenue. Founders are surprising themselves by winning contracts they weren't fully prepared to deliver.
• •Frontier model data sovereignty risk: Startups paying tens of millions annually in API tokens to frontier labs risk training their eventual competitors. Anthropic reversed a data retention policy under enterprise pressure, now offering 30-day retention windows on customer cloud environments. The strategic response—already adopted by legal AI firm Ligura—is building proprietary vertical language models or deploying on-premise LLMs, removing dependency on labs that face pressure to launch competing vertical products.
• •Non-consensus bet mechanics inside a VC firm: 776's investment in Star Cloud—orbital GPU data centers—required Caitlin Holloway to independently learn thermal dissipation physics in space and make an internal case against vocal external critics who built websites calling the concept the worst idea ever. The firm uses the concept of "skin rippy" conviction as a threshold signal. Within six months of the investment, Elon Musk, Jeff Bezos, and Sam Altman publicly validated orbital data centers as a core infrastructure solution.
• •Launch cost trajectory unlocks space infrastructure: Cost per kilogram to orbit dropped from roughly $10,000 in 2014 to $4,000 today, with projections reaching $1,500 by 2030 and potentially $273 by 2040—driven almost entirely by SpaceX reusability. This cost curve is the foundational thesis for orbital data centers, reusable second-stage rockets like Stoke Space's Nova, and the broader commercialization of space infrastructure. Investors should track this metric as the primary demand unlock for any space-adjacent hardware company.