Listen to this article

Electric Sheep Robotics, an outdoor maintenance company, has acquired two more landscaping businesses. Westar Landscaping and Caliscapes are the newest additions to Electric Sheep’s landscaping portfolio.

With these new acquisitions, San Francisco-based Electric Sheep Robotics has acquired four landscaping businesses in recent months. In October, it acquired Phenix Landscape and Complete Landscaping. These acquisitions are part of the company’s long-term strategy to consolidate the landscaping industry. Electric Sheep plans to acquire traditional landscaping companies and progressively transform their operations to include its autonomous lawn mowers and other robots.

Since implementing this model, Electric Sheep said it has grown revenue eight times. And it doesn’t plan on slowing down anytime soon. The company said it has a growing pipeline of interested businesses that could enable it to grow by 10 times in 2024.

“Electric Sheep is at a critical point in its growth; acquiring Westar and Caliscapes builds on our successful business model of injecting advanced AI and robotics into traditional landscaping companies and significantly increasing their value,” said Nag Murty, CEO and co-founder of Electric Sheep. “We’re bringing a new business model to an industry that is ripe for innovation; by acquiring these businesses first and incorporating this full data and AI deployment engine, we are creating a fully scalable and sustainable business that is really a first in the outdoor services market.”

---

Learn from Agility Robotics, Amazon, Disney, Teradyne and many more.

---

Electric Sheep Robotics built to scale

Electric Sheep’s acquisition strategy is a key aspect of its long-term business plan. Acquiring already profitable businesses, and maintaining that profitability as Electric Sheep starts to introduce robotics to their operations, means that the company can be profitable from day one.

“The ESR [Electric Sheep Robotics] business model of acquiring landscaping businesses and improving their margins by augmenting workers with automation is radically innovative, and a sustainable, rapidly scalable way to build moonshot robotics,” said Pieter Abbeel, professor in AI and Robotics at UC Berkeley, co-founder and chief scientist at Covariant, and long-time scientific advisor to Electric Sheep. “As this model scales, ESR is poised to build an RL factory to train AI agents for sustainable outdoor work. I’m excited to support their mission.”

The company considers a few factors when it’s picking landscaping companies to acquire. The first is that it’s looking for a diverse set of data, which can include a diversity of workflows and sites. Jarrett Herold, Electric Sheep’s co-founder and COO, also said company culture is a large factor when picking companies to acquire. 

“Our intent is to roll this up to hundreds of millions, to almost a billion dollars, in revenue and deploy at the scale of tens of thousands of robots because we believe that the next shift in robotics will have to happen with this progressive automation and reinforcement learning done up to that scale,” Murty told The Robot Report in October.

Full stack approach

Electric Sheep uses machine learning models it designed to automate various physical tasks. These tasks can include anything from mowing and sweeping to knowledge work like inventory management, customer success, and marketing.

The company’s robots explore, map, navigate, and manipulate the physical world around them. They’re deployed in places like HOAs, parks, university campuses, and more. Electric Sheep is developing a full stack data channel that its robots are continually trained on.

While the company’s acquisition model does help it to maintain profitability, it also plays an important role in building its full-stack data channel. Murty said that acquisitions are a good way for the company to acquire crucial data it can use to build its AI models. Electric Sheep’s goal is to build a foundational model similar to large language models like ChatGPT, but for the physical world instead of for language.

“Today, AI models really understand language, but they still have no sense of physical reality,” Herold said. “If you take a cat, it navigates the physical space effortlessly, you don’t have to train a cat to go around your house and do X, Y, Z. In the same way, I think the model that we’re building at its core is fundamentally semantically aware of its surroundings.”