Physical AI Forms an App-Infrastructure Loop
oyhsu · x · 2026-07-11
The author reviews the evolution of the physical AI / robotics sector over the past 2–3 years, highlighting three main trends:
1. The Robotics Stack is Becoming Layered
In the past, many teams built their entire infrastructure in-house. However, since 2023–2024, an increasing number of teams focus solely on a single layer built on top of general-purpose platforms. The author believes this means robotics might see an apps > infrastructure loop, similar to the web, mobile, and crypto industries.
2. 2024–2025 Focus Shifts to Data and Platforms
During this phase, the industry primarily tackled how to scale the collection of robotic data. We also saw the emergence of more robotic hardware platforms, components, and tools tailored for "robot developers."
3. 2025–2026 Application Layer Takes Shape
More entrepreneurs and engineers are shifting towards the deployment layer, giving rise to "neo-systems integrator" companies dedicated to actually deploying next-generation learning-based robots in the field. Large robotics labs are also starting to handle both proprietary and third-party deployments, clarifying the relationship between applications and infrastructure.
Research Frontiers
The author also summarizes key directions in general robotics research, including:
- VLA (Vision-Language-Action)
- world state prediction
- sim / simulation
- human motion transfer
- online RL
- domain randomization
- high-low level action bridging
They also mention that there is growing evidence supporting scaling laws for robotic actions.
Finally, the author notes a shift in focus towards the broader frontiers of physical AI: integrating physical reasoning and physical modalities into overall AI capabilities.
More from Embodied
- Musk: Cybercab certified at 165 Wh/mi, the most efficient production EV ever — elonmusk · 2026-09-11
- MKBHD goes hands-on with the first folding iPhone; $2,000 48MP selfie cam mocked — alexmacgregor__ · 2026-09-11
- NTU spin-off Ropedia launches HOMIE Gen 2 wearable system to train robots from human experience — liuziwei7 · 2026-09-11
- M.2-Oculink eGPU Link Silently Downgrades to PCIe Gen1 — Here's How to Check — El_90 · 2026-09-11
- Replit Agent in a robot builds and publishes websites autonomously via MCP — amasad · 2026-09-11
- TARS Robotics unveils embodied foundation model AWE: 15+ tasks, one model, zero retraining — heyshrutimishra · 2026-09-11