Automation
for Energy Infrastructure.

Energy is the basis of our society. Everything we do, from the food we eat to the devices we use, depends on it. Electricity in particular will play a bigger role in the future. The electrification of transport, heating and industry must and will happen, and along with the growth in compute needs, it will push demand for electricity up[1]. Among the ways to produce that electricity, renewables should and will take a bigger share. Solar alone is projected to account for almost 80 percent of that growth[2].

Yet the critical infrastructure on which our economy depends still has no scalable, cost-effective way to be maintained. Automation, for good reasons, has been confined to factories and warehouses. Energy infrastructure like solar farms, battery storage and substations, spanning hundreds of acres in places that are hard to reach, has been left behind. This needs to change if we want to build a strong basis for our society.

Robotics has been around for decades. Autonomous robots, though, have mostly stayed indoors, in warehouses and factories built to make their job easy. That is changing. On navigation, self-driving cars now handle city streets, an environment far messier than a solar field or a substation. On hardware, the sensors that used to price outdoor robots out of the market have collapsed in cost: automotive lidar has gone from around $75,000 a decade ago to a few hundred dollars today[3]. The case for working outdoors is no longer a hardware problem.

On manipulation, the same shift is underway. Foundation models trained on demonstration data, instead of hand-coded motion, are starting to let robots learn a physical task by watching it rather than being programmed joint by joint. These models are not yet reliable enough for the hardest tasks, and the data to get them there barely exists: most robot foundation models today are trained indoors, on tabletops, not outdoors on real infrastructure.
The data flywheel has to start somewhere. Better field data will produce better models, and better models will make robots capable of more work.

Put navigation, manipulation and a hardware cost curve that finally works together, and we think this is the decade robots leave the warehouse.

Following our previous thoughts, we came to the logical conclusion that:

In 10 years from now, energy infrastructure will be built, monitored and maintained by autonomous robots always available on site, effectively lowering the cost of energy and increasing the availability and safety of these critical assets.

We are building the hardware and software required to deploy autonomous robots in the field and make them useful from day one. Our machines are designed around the work they need to perform: form follows function.
We stay close to operators, solve problems that matter today, and iterate quickly as the capabilities of robotics improve. We are starting with the tasks where robots can already create meaningful value, while building the technology, field experience and data required for them to take on increasingly complex work.
Our ambition is not to build robots for demonstration. It is to build the machines that energy infrastructure comes to depend on.