We've migrated — and merged — into something bigger

Physical AI is not a trend. It is the next computing platform. That's why I'm proud to announce the migration of the F50 Newsletter from Mailchimp and Physical AI Builders Newsletter into one unified home: Physical AI Builders at 50builders.ai. Two networks, one mission — to accelerate the people building the physical AI era. Opt out anytime.

And we are bringing that ecosystem together in person. April 29 in Silicon Valley. June 11 in Dallas. The F50 Physical AI Summit is where the real conversations happen — free for select founders, investors, and operators who are actually building this future. RSVP links bellow. I hope to see you there.

The future is physical. Let's build it together.

— David Cao Curator, Physical AI Builders · F50

Announcement 01 · Robotics Foundational Models

Isaac GR00T N1.7 Available + GR00T N2 Preview

GR00T N1.7 is now in early access with a commercial license, focused on dexterity and generalized skills. The preview of GR00T N2 — built on a "World Action Model" architecture — helps robots succeed at new tasks in unfamiliar environments more than twice as often as leading alternatives. It currently ranks #1 on MolmoSpaces and RoboArena benchmarks.

Announcement 02 · World Foundation Models

Cosmos 3 — A Unified World Foundation Model

Cosmos 3 is the first world foundation model that unifies synthetic world generation, vision reasoning, and action simulation in a single architecture, built for robotics and autonomous systems. LG Electronics and Milestone Systems are already adopting it for physical AI workloads.

Announcement 03 · Physics Simulation

Newton 1.0 — Open-Source Physics Engine

Co-built with Google DeepMind and Disney Research, Newton is an open-source physics engine purpose-built for developing robots. NVIDIA now describes three computers for physical AI: Isaac GR00T as the robot's brain, Newton simulating its body, and NVIDIA Omniverse as its training ground.

Announcement 04 · Data Infrastructure

Physical AI Data Factory Blueprint

An open reference architecture that automates training data pipelines in three stages: Cosmos Curator for curation, Cosmos Transfer for augmentation, and Cosmos Evaluator for quality assessment. NVIDIA's goal: turn robotics' data problem into a compute problem.

Announcement 05 · Autonomous Vehicles

Alpamayo 1.5 + New Automaker Partners

Alpamayo 1.5 is a specialized model for AVs with an interactive steerable reasoning engine. New partners including BYD, Hyundai, Nissan, and Geely have joined the NVIDIA DRIVE Hyperion platform, with Uber deploying autonomous vehicles directly into its ride-hailing network.

Announcement 06 · Industrial Robotics

ABB, FANUC, YASKAWA & KUKA Fully In

The four largest industrial robot makers — collectively operating well over 2 million installed units globally — are now integrating NVIDIA Omniverse libraries and Isaac simulation frameworks into their commissioning solutions, with Jetson modules being built directly into robot controllers for real-time edge AI inference.

Announcement 07 · Healthcare

Surgical Robotics Platform: Open-H, Cosmos-H, GR00T-H

NVIDIA launched a full domain-specific suite for healthcare robotics — including Open-H (the world's largest healthcare robotics dataset), Cosmos-H for physics-based surgical data generation, and GR00T-H, a vision-language-action model that converts text commands into precise robot motion for clinical tasks.

Announcement 08 · Demo Moment

Olaf Walks the GTC Stage

Disney Research's Olaf robot — trained in Isaac Lab, powered by Newton, running on a Jetson chip — walked onto the SAP Center stage as a live demo of the full physical AI stack. "You learned how to walk inside Omniverse," Huang told the snowman. A preview of AI characters in theme parks — and a perfect metaphor for NVIDIA's platform ambition.

① Signal

Turning the Data Problem into a Compute Problem

The most strategically significant move at GTC 2026 was NVIDIA reframing physical AI's core bottleneck. By combining Cosmos 3, the Data Factory Blueprint, and Isaac Lab 3.0, NVIDIA argues that synthetic data + massive GPU compute is a viable shortcut to capable physical AI — replacing years of real-world fleet data collection. This is a fundamentally powerful business case for selling more Blackwell and Vera Rubin GPUs.

② Signal

The Full Stack Lock-in — Isaac + Newton + Omniverse

NVIDIA isn't selling chips. It's selling a vertically integrated operating system: GR00T as the brain, Newton as the body simulator, Omniverse as the training ground. By open-sourcing GR00T on Hugging Face, NVIDIA captures developers early, then monetizes through compute and deployment layers. The strategy mirrors how CUDA captured scientific computing — a flywheel that compounds over years.

③ Signal

Winning the Industrial Giants Means Decade-Long Revenue

ABB, FANUC, KUKA, YASKAWA, Universal Robots — these are companies with 20–50 year product cycles. Embedding NVIDIA Jetson and Omniverse into their robot controllers and commissioning pipelines means NVIDIA becomes the invisible compute layer inside factories that don't upgrade for a generation. It's the stickiest revenue stream in hardware.

④ Signal

Healthcare Robotics — The Sleeper Frontier

While AVs and manufacturing dominate physical AI coverage, NVIDIA's surgical robotics suite signals a deliberate third frontier. With regulatory moats, high-stakes precision requirements, and enormous capital budgets, healthcare robotics could become as strategically important as autonomous vehicles — and the barriers to competition are even higher.

⑤ Signal

The $1 Trillion Compute Bet is Partly a Physical AI Story

Jensen Huang projected at least $1 trillion in revenue through 2027. Physical AI is a key demand driver — robots, AVs, and edge inference devices require continuous compute at scale. Unlike one-time model training runs, physical AI deployments generate persistent, ongoing inference demand. That's recurring GPU revenue at a scale the industry hasn't seen before.

⬛ Bottom Line

GTC 2026 confirmed that NVIDIA is not merely an AI chip company. It is systematically building the operating system of the physical world — from simulation to foundation models to edge hardware — and signing up every major industrial player to run on its stack. Physical AI is no longer a future bet. It is NVIDIA's present business.

📅 F50 Physical AI Summit is Calling

Two landmark gatherings. One movement. Join the builders shaping the physical AI era.

APR 29 F50 Physical AI Summit

📍 Silicon Valley, CA — 2026

JUN 11 F50 Physical AI Summit

📍 Dallas, TX — 2026

📌 Recent Meetups

Feb 25 Physical AI Builders Meetup 350 RSVPs

Mar 5 Dallas Meetup 190 RSVPs

Mar 12 Austin (SXSW) Meetup 800+ RSVPs 🔥

⚡ Upcoming MAR 25 Post-GTC PAB Meetup

​Present or demo your work at PAB meetup:

​👉 Apply here: https://app.sve.io/call

​Join the Physical AI Builders Community

​Continue the conversation beyond the event:

​🎙️ Physical AI Podcast: youtube.com/@f50f50

​Sponsorship Opportunities

​Support the community shaping Physical AI's future and gain prime visibility.

If you do not like this email,Opt out anytime. You can still register our event at lu.ma/physical-ai

Austin-based Apptronik closed a $520M extension in February, bringing its total Series A to over $935 million. The round was co-led by Google and B Capital, with participation from Mercedes-Benz, AT&T Ventures, John Deere, and the Qatar Investment Authority — valuing the company at $5.5B+.

SERVE ROBOTICS SHOWCASING AT SXSW — ACQUIRED DILIGENT ROBOTICS IN 2026

Autonomous delivery robot company Serve Robotics is presenting at both SXSW and NVIDIA GTC this week. After acquiring Diligent Robotics, Serve now spans both sidewalk delivery and hospital service robots — reaching approximately 3 million people across the U.S., serving 3,600+ restaurants.

🏙️ PHYSICAL AI at SXSW

Multiple signals confirm Austin's ascent: