Nvidia bets on robotics to drive future growth

State of Robotics Market & Business Models

• Robotics seen as long-promising but historically low-margin, reliability-focused, and slow-growing, especially in industrial settings.
• “Robots-as-a-service” is emerging: vendors deploy, maintain, and remotely monitor robots, charging per operating hour or per unit of work, aligning incentives and lowering adoption barriers.
• Industrial robots are mostly arms with vision systems or mobile bases, not humanoids; successful automation is often invisible or rebranded as something else.
• Some argue battery, AI, and cheap semiconductors now remove key historical bottlenecks; others note we still lack a mass-market “household robot.”

Nvidia’s Robotics Bet & Market Size Question

• Nvidia is pushing a full stack (GPUs, Jetson/DRIVE, software, simulation) to be more than a compute vendor and “own” the robotics AI layer.
• Skeptics doubt factory robots alone can materially move a trillion‑dollar company: annual industrial installations are modest, many tasks don’t need large GPUs, and China (a big robot buyer) faces export limits.
• Supporters counter that AI-driven, flexible “android-like” robots across industry and services could greatly expand demand for on-device compute.

AI, GPUs, and Technical Shifts in Robotics

• Discussion centers on large vision–language(-action) models, imitation learning, RL, and massive simulation as the main “GPU-driven” breakthroughs.
• GPUs accelerate mapping, dense cost grids, vision, end‑to‑end planning, and policy learning; embedded platforms like Jetson make this practical on robots.
• Some see general-purpose robotics as achievable with huge amounts of real-world data and bigger models; others argue imitation learning doesn’t generalize and current demos are brittle and overhyped.

Self‑Driving as Robotics Example

• Strong disagreement over whether self-driving is “close to solved.”
• Waymo is cited as highly capable but geographically and operationally constrained, occasionally getting stuck or having incidents.
• Tesla FSD users report impressive performance, but others stress that short personal experience is meaningless for safety and that many 9s of reliability are still missing.
• Broad view: autonomous driving is robotics, but real-world deployment remains limited by safety, reliability, and operations.

Hardware, Cost, and Developer Experience

• Jetson Orin praised for power and ease of use; price cuts make it more competitive, though still not “$10 GPU” cheap.
• Some prefer x86 mini‑PCs for software compatibility; CUDA requirements complicate alternatives.
• Microcontrollers (e.g., ESP32-class) can run tiny models but are too slow for serious convnet workloads; dedicated neural accelerators on MCUs are emerging.
• Complaints about robotics stacks like ROS lagging modern dev practices; others respond that target users are hardware-heavy industries, not web/app engineers.

Safety, Ethics, and Militarization

• Concerns raised about lack of robust methods to verify safety and reliability of ML-driven robots.
• Fears that autonomous or semi-autonomous armed robots (e.g., gun drones, armed quadrupeds for border control) will be attractive to states seeking to distance humans from violence and reduce political risk.
• Some speculate this could shift war toward leader‑targeting and reduce mass casualties; others note history of targeted killings hasn’t prevented broader conflicts.

Hype, Economics, and Future Outlook

• Several comments see Nvidia as surfing successive hype waves (crypto → LLMs → robotics) to sustain GPU demand, with uncertain long-term economics.
• Others argue AI already delivers real value; failures often stem from trying to retrofit old workflows instead of designing new ones.
• Unclear whether “physical AI” via robotics will match the scale of the LLM boom, but many expect significant growth as perception and control keep improving.