TL;DR: Standardization is no longer just about nuts and bolts; it is about data and protocols. The value of a robotics company is no longer found in its proprietary motors or closed software silos, but in its ability to integrate with the global ecosystem. We are entering a robotics renaissance where the barrier to entry has never been lower, and the potential for general-purpose automation has never been higher.
So who wins? Our belief is that in the end, the customer will win. With customers having more access to better robotics due to more companies being able to enter the space customers will have choice, access, and cost-savings. From the company perspective, the companies that will win are the ones best at leveraging, fine-tuning, and contributing meaningful data back to these shared pools - most likely Nvidia…
The Robotic Standard
For decades, robotics followed a "hardware-first" model, where physical robots and their software were inseparable, locked away in proprietary silos. Today, we are witnessing a fundamental shift: hardware is becoming a commodity, while a standardized "intelligence stack" is becoming the industry's true value driver.
Proprietary History
In the 90s, companies like ABB (using RAPID) and FANUC (using KAREL) built "walled gardens." Every sensor upgrade required a custom driver (software that lets the operating system interact with specific hardware or devices), and every robot spoke a unique language.
While proprietary tech was once a moat, it is now a liability. In an era of hyper-innovation, no single company can outpace the collective intelligence of the global open-source community; if you cannot beat the community's speed, joining it is the only way to stay relevant.
Unlike autonomous vehicles, which operate in the highly structured, "solved" environment of paved roads and traffic laws, general robotics must navigate the "infinite entropy" of homes and warehouses. Companies like Tesla and Waymo succeeded in silos because they solved a constrained operational design domain problem (specific routes, road types, conditions, etc). In contrast, general robotics faces an open-ended diversity problem: the sheer variety of objects, textures, and human interactions is too vast for any one brand-specific silo to map.
By restricting data to closed ecosystems, companies trap their robots in a cycle of diminishing returns. Realizing general-purpose intelligence requires an open foundation where shared data drives a universal increase in machine capability, turning individual breakthroughs into industry-wide standards.
The Intelligence Stack: ROS 2 and DDS
As hardware commoditizes, the software framework becomes the "gold standard." ROS 2 (Robot Operating System), managed by the non-profit Open Source Robotics Foundation, acts as a universal translator, allowing developers to mix and match hardware from different vendors.
The Framework: ROS 2 provides the interoperable software layer.
The Protocol: It utilizes DDS (Data Distribution Service), a peer-to-peer discovery system.
Instead of relying on a central server, DDS allows a sensor and a motor to "find" each other and share data automatically. This modularity ensures that the intelligence is decoupled from the physical frame.
Hardware as an Embodiment Platform
Historically, the sheer physical diversity of robots (wheels, legs, or arms) made software sharing impossible. Standardized APIs and physical interfaces have finally bridged that gap.
We are seeing a rapid adoption of universal standards:
Physical Connectivity: Standards like ISO 9409 define how parts from different manufacturers physically bolt together.
Safety: Modern ISO standards ensure these modular machines can safely operate alongside humans.
This shift transforms robots from single-task machines into embodied platforms. Companies can now invest in a chassis today, knowing they can upgrade the "limbs" or "eyes" tomorrow without rebuilding the brain.
The Brain: The Rise of VLA Models
The most significant breakthrough in robotic logic is the transition from rigid "if/then" programming to Vision-Language-Action (VLA) models. VLA models are unified AI systems that combine visual and linguistic data to directly output physical robotic commands. Unlike traditional AI that uses separate modules for sensing, planning, and moving, VLAs use a single integrated "brain" to process inputs and actions simultaneously for better real-world adaptability.
In the past, picking up a cup required an engineer to hard-code specific coordinates and force sensor thresholds. With VLAs, such as NVIDIA’s Isaac GR00T, a robot can understand a natural-language command: "Carefully pick up the red cup."
The Power of "Positive Transfer"
Perhaps the most counter-intuitive discovery in modern robotics is Positive Transfer. Research via the Open X-Embodiment dataset has shown that a robot’s performance improves when it is trained on data from entirely different robots.
Data from a diverse pool of hardware is more valuable than more data from a single robot. As we add more hardware to open-source datasets, every robot in the ecosystem gets smarter simultaneously.
Takeaway: Standardization is no longer just about nuts and bolts; it is about data and protocols. The value of a robotics company is no longer found in its proprietary motors or closed software silos, but in its ability to integrate with the global ecosystem. We are entering a robotics renaissance where the barrier to entry has never been lower, and the potential for general-purpose automation has never been higher.
So who wins? Our belief is that in the end, the customer will win. With customers having more access to better robotics due to more companies being able to enter the space, customers will have choice, access, and cost-savings. The companies that will win are the ones best at leveraging, fine-tuning, and contributing meaningful data back to these shared pools - most likely Nvidia…
Have a great weekend,
Josh