To kickoff the year, we have released our Physical AI Podcast episode: “Physical AI 2026 Outlook”—a fast-moving, technical conversation with advisory committee members about what’s actually going to matter as AI leaves the cloud and enters factories, fleets, farms, jobsites, and robots.

Featured Physical AI Advisory committee member guests in this episode (and why they matter):

More about the Physical AI Advisory Committee—a new leadership group bringing together builders and decision-makers across embodied AI, robotics, industrial automation, manufacturing, energy, infrastructure, logistics, and construction to shape what comes next.

Below are the 2026 outlook highlights (edited into newsletter form from a longer conversation):

🔹 Apurv (NVIDIA): Simulation-first becomes the default path to real-world autonomy Apurv’s outlook is that the next acceleration wave will come from teams who treat simulation as the training ground for everything—using world models to generate scenarios, bootstrap data, and shorten time-to-skill across new environments. In one clip, he describes how “world foundation models” can generate physics-aware scenarios from a prompt, and flags compute cost as a practical limiter on how far teams can push “everything in simulation.” Why this matters: the broader ecosystem is investing heavily in physics-aware synthetic data generation and simulation pipelines, which is consistent with NVIDIA’s world foundation model framing and simulation/synthetic data positioning.

🔹 Anupam: 2026 is when enterprises escape AI pilot purgatory and scale into physical operations Anupam’s forecast focuses on the “enterprise reality check”: scaling Physical AI requires reliability, predictability, and latency performance at the edge. Why this matters: industrial systems are explicitly constrained by safety/performance requirements, and reliability + latency are recurring “hard requirements” in OT and industrial wireless contexts—especially once AI is making real-time decisions.

🔹 Harnish: Physical AI expands beyond controlled indoor settings into messy outdoor domains Harnish’s 2026 view is a resurgence of robots and AI systems leaving structured indoor environments and moving into unpredictable outdoor settings—construction, agriculture, and other real-world “long tail” conditions—driven by better edge deployment and more capable multimodal action models. Why this matters: vision-language-action progress is making it more credible to unify perception + language + action planning in a single model family—crucial for unconstrained environments.

🔹 David: The next bottleneck won’t just be models—it’ll be the robotics supply chain and services layer David flags a practical constraint: physical deployment scales through supply chains, integrators, testing, maintenance, logistics, and parts—an ecosystem that must mature alongside the models. (If you’re building in the “picks and shovels” layer—this is your opening.)

🔹 Keith: Data advantage shifts toward 3D workflows—and governance becomes a moat Keith’s 2026 emphasis is that experimentation and proprietary data loops will determine who wins—and that 3D data around products, facilities, and fleets (plus licensing/protection) will become a major competitive differentiator. Why this matters: digital twin frameworks and standards are evolving because the world is pushing toward reusable, interoperable “digital representations of physical systems,” not just one-off demos.

Full Episode on Youtube

Physical AI is entering a phase where the hard problems aren’t only model architecture—they’re data pipelines, simulation-to-real transfer, edge reliability, safety constraints, and the industrial deployment surface area. That’s exactly why the advisory committee exists: to keep the ecosystem grounded in what works, what breaks, and what’s worth funding or building next.

If you’re leading in any of these areas—robotics, embodied AI, industrial automation, fleet autonomy, smart infrastructure, energy systems, data centers, edge AI, manufacturing ops, or supply-chain modernization—we want your voice in the room.

Coming next: Product Alpha in the Age of Physical AI: Lessons from Groq’s $20B NVIDIA Deal with SC Moatti.

Which 2026 prediction do you agree with most—and what are we missing? Drop your take in the comments. 👇