Physical AI, AD and Robotics

Newsletter | Technical Briefing

Curated insights for AD professionals, Roboticists, Physical-AI engineers, Founders & Tech leaders

🚗 Waymo and Uber end Phoenix robotaxi pilot

• ●Waymo and Uber have discontinued their joint robotaxi pilot program in Phoenix, ending a partnership that allowed Uber app users to hail Waymo’s autonomous vehicles.
• ●The shutdown marks a strategic shift for both companies as they reassess their autonomous vehicle deployment models in competitive markets where multiple robotaxi operators are vying for riders.
• ●Watch whether Waymo pursues direct-to-consumer operations in Phoenix or pivots to other markets where it maintains exclusive partnerships with ride-hailing platforms.
• 🔍 Read More →
• What matters: Partnership dissolutions signal that robotaxi operators are moving from experimental collaborations to hardened competitive strategies.

🧠 Galbot’s Cang Yu: Embodied AI data dilemma harder than autonomous driving

• ●Galbot CEO Cang Yu argues that embodied AI faces a “brain ready, hands lagging” problem, where foundation models outpace the hardware and data pipelines needed to train manipulation skills.
• ●The data challenge for humanoid robots exceeds autonomous driving’s complexity because manipulation requires diverse, high-quality interaction data across countless object types and environments—far harder to collect at scale than driving scenarios.
• ●Expect the industry to prioritize data collection infrastructure and simulation-to-real transfer methods as the next bottleneck after model architecture improvements plateau.
• 🔍 Read More →
• What matters: The embodied AI race will be won by whoever solves data collection and curation, not just model scaling.

🚗 Drones taught to “feel pain” could prevent self-driving cars from self-harm

• ●Researchers have developed a pain-sensing system for drones that uses sensors to detect physical damage in real-time, triggering protective responses similar to biological pain reflexes.
• ●The technology could extend to autonomous vehicles, enabling them to detect and respond to mechanical stress, collisions, or component failures before catastrophic damage occurs—essentially giving robots self-preservation instincts.
• ●If commercialized, pain-sensing systems could reduce maintenance costs and improve safety by allowing autonomous systems to self-diagnose and adapt behavior when operating near their physical limits.
• 🔍 Read More →
• What matters: Biomimetic sensing could shift autonomous systems from reactive to self-protective, reducing downtime and extending operational lifespans.

🧠 Avride deploys cloud VLMs as safety net for delivery robots

• ●Avride uses cloud-based vision-language models (VLMs) to provide contextual awareness for its delivery robots, acting as a fallback when onboard perception systems encounter ambiguous or novel scenarios.
• ●The hybrid architecture offloads complex reasoning to the cloud while maintaining low-latency edge compute for real-time navigation, balancing safety and responsiveness in unpredictable urban environments.
• ●This approach suggests a broader industry trend: edge-cloud hybrid architectures that leverage foundation models for edge cases without sacrificing the speed required for autonomous operation.
• 🔍 Read More →
• What matters: Cloud VLMs enable robots to handle long-tail scenarios without requiring massive onboard compute or exhaustive training datasets.

🤖 Zoox robotaxi redesign brings rider-focused upgrades

• ●Zoox has unveiled a redesigned robotaxi with significant interior improvements aimed at enhancing passenger comfort and user experience in its bidirectional, purpose-built autonomous vehicle.
• ●The updates reflect lessons learned from early deployments, prioritizing rider-facing features as Zoox prepares to scale commercial operations beyond its current test markets.
• ●As robotaxi operators move from technical validation to consumer adoption, expect more emphasis on interior design, ride quality, and user interface refinements to differentiate services.
• 🔍 Read More →
• What matters: Robotaxi competition is shifting from autonomy capability to passenger experience as the technology matures.

🧠 NVIDIA expands Physical AI with Verkada partnership, names Microsoft veteran Parker

• ●NVIDIA has partnered with Verkada to integrate its Physical AI platform into enterprise security and building management systems, while appointing Microsoft veteran Parker to a leadership role in its Physical AI division.
• ●The Verkada deal extends NVIDIA’s reach beyond robotics and autonomous vehicles into the massive enterprise IoT market, where AI-powered cameras and sensors generate actionable insights from physical spaces.
• ●NVIDIA’s aggressive Physical AI expansion signals its bet that the next wave of AI growth will come from embodied systems—robots, vehicles, and smart infrastructure—not just software.
• 🔍 Read More →
• What matters: NVIDIA is positioning Physical AI as the next trillion-dollar compute market, with enterprise IoT as the beachhead.

🚗 WeRide and Uber launch commercial robotaxi service in Zurich

• ●WeRide and Uber have launched a commercial robotaxi service in Zurich, marking WeRide’s first European deployment and Uber’s continued push to integrate autonomous vehicles into its ride-hailing platform across multiple markets.
• ●The Zurich launch is part of a broader European expansion strategy, testing regulatory frameworks and consumer acceptance in a region with stricter autonomous vehicle rules than the U.S. or China.
• ●Success in Zurich could accelerate WeRide’s entry into other European cities, while demonstrating that Chinese AV companies can compete globally despite geopolitical headwinds.
• 🔍 Read More →
• What matters: Europe is becoming the next battleground for robotaxi operators seeking to prove their technology works under diverse regulatory regimes.

⚡Partnership fragmentation::Robotaxi operators are ending experimental collaborations and hardening competitive strategies as markets mature.

⚡Data bottleneck::Embodied AI’s next frontier is data collection infrastructure, not model architecture—manipulation data is harder to scale than driving data.

⚡Hybrid architectures::Edge-cloud systems using foundation models for edge cases are becoming the standard for autonomous robots operating in unpredictable environments.

⚡Physical AI expansion::NVIDIA’s enterprise IoT push signals that the next AI compute wave will come from embodied systems, not software alone.

⚡Global regulatory test::Europe’s stricter AV rules make it the proving ground for robotaxi operators seeking to demonstrate global viability—watch Zurich closely.

Worth forwarding to a colleague? Pass it along.

© 2026 Physical AI, AD and Robotics · DriveTech AI. All rights reserved. Privacy Policy