Applied Intuition CTO on 'Physical AI' on the Battlefield

Applied Intuition CTO Peter Ludwig told Bloomberg Businessweek Daily that the company is developing 'physical AI' applied to automotive systems and defense technologies, including drones and drone swarming. He also addressed ethical debates around military uses of the technology; no financial figures, contracts, or guidance were disclosed.

Analysis

The push toward “physical AI” materially reweights where value accrues in autonomy: not just to perception stacks but to large-scale simulation, orchestration, and verification infrastructure. Expect incremental datacenter GPU demand (H100-class workloads) and specialized inference accelerators to rise within 6–18 months as fleets move from lab pilots to scaled validation; that creates 20–40% higher attach rates for compute per vehicle/drone versus today’s pilots. Vendors that own the test/verification workflow capture higher-margin recurring revenue than pure sensor or integrator suppliers.

On the defense side, adoption expands TAM but stretches realization timelines — procurement, security certification, and systems integration mean meaningful revenue for primes and tier-1 suppliers will likely flow over 1–5 year windows, not quarters. Second-order winners include domestic avionics/comms/sensor vendors where on-shore sourcing and export controls raise effective entry barriers and increase attach rates; conversely, offshore or commodity hardware suppliers face compression as integrators vertically capture more software-defined value.

A key risk is regulatory and ethical pushback that can reroute budgets back into slower manual systems or strict certification regimes; a single high-profile autonomy incident or restrictive DoD policy could delay programs by 12–24 months. The market’s implicit timeline is optimistic — price in execution at 6–12 months and you overpay; price it over a multi-year adoption curve and you avoid negative surprises and identify acquisition targets earlier in the cycle.

Key Decisions for Investors

• Long NVDA (NVDA) via 6–12 month call spread or stock exposure: thesis is 20–40% incremental compute per autonomy deployment driving accelerated data-center GPU demand. Risk/reward: asymmetric — limited time premium vs multi-quarter cadence of H100 restocking; stop-loss at 12% on sudden datacenter demand slowdown.
• Overweight defense primes with on-shore electronics footprint — RTX (RTX) and L3Harris (LHX) — horizon 12–36 months: defensive revenue from integration and certification + higher attach rates. Risk/reward: lower beta with 8–15% upside if DoD pilot programs scale; downside is program delays, hedge with short-term puts on the position sized 20% of notional.
• Buy ANSYS (ANSS) or other simulation/verification software exposure for 9–18 months (stock or long-dated calls): simulation monetizes faster than hardware and is a choke-point for certification. Risk/reward: high margin growth if OEMs outsource validation; monitor OEM capital allocation and new RFPs as catalysts.
• Selective short of pure-play lidar/commodity sensor vendors (e.g., LAZR) over 6–12 months: thesis is margin compression and consolidation as software-defined stacks reduce attach rates for expensive sensors. Risk/reward: binary technology bets — cap losses with tight hedges or use options to limit downside if lidar wins a breakthrough.