Agility Robotics is launching a 60,000-square-foot facility dedicated to the training of its humanoid robots in Fremont, California, located not far from the Tesla factory where manufacturing of the Optimus robots is anticipated to commence this year.

Tesla has increasingly invested in Optimus. Elon Musk recently indicated that he anticipates it to become “the largest product ever” once it’s “useful outside of Tesla sometime next year.”

Although Agility may not possess Tesla’s financial resources, it has the working robot, Digit, which is already proving beneficial in various real-world applications. This robot is currently generating income by handling totes and bins in manufacturing and warehouse operations for clients such as Amazon, GXO, Schaeffler, and Toyota Motor Manufacturing Canada. The organization claims it has obtained $300 million in contract orders for its robots.

“Having [Tesla] in proximity to us is fantastic because Agility was initially out there alone, and it’s beneficial to have others in the humanoid sector,” CEO Peggy Johnson shared with TechCrunch. “We have commercialized. We now understand what it takes to integrate into these facilities and satisfy their safety, regulatory requirements, compliance, and IT infrastructure along with their warehouse management system.”

Agility has not shared the number of Digits it has constructed or deployed, but outside analysts estimate that several dozen have been active in pilot or profit-generating projects. For instance, the company has noted that Digits have facilitated the movement of 100,000 totes within a GXO logistics site.

Johnson is currently navigating Agility through a reverse merger that is expected to position it as the first pure-play humanoid robot firm on public markets by the end of this year. Established in 2015 by a team of researchers who developed innovative methods for enabling robots to walk on two legs safely, Agility aims to leverage its advantage over a newer wave of AI-driven robotics startups like Figure, 1X, the Bot Company, and Sunday Robotics.

While the introduction of transformer-based neural networks contributing to the emergence of LLMs is poised to bring significant advancements in robotic behavior, Agility is adopting a pragmatic stance toward autonomy.

“Considering self-driving cars as a non-humanoid example, you definitely wouldn’t want the anti-lock brake controller managed by AI,” Agility co-founder and chairman Damion Shelton remarked to TechCrunch. “For humanoids, all safety components must follow a path that isn’t reliant on generative AI, correct? Creativity should not be involved in your safety stack.”

Nevertheless, what AI does provide is the potential for scalability.

“One of the earliest occasions when [Bruce Leak, the Quicktime inventor who serves on Agility’s board] inquired how we would approach coding applications for the robot, we didn’t have a solid answer,” Shelton stated. “The array of tasks a robot can be envisioned to perform far exceeds the number of engineers capable of programming robots. Generative AI resolves that issue definitively.”

The new facility is intended to expedite the company’s robot deployments. Johnson mentions that over 30 customers are currently negotiating with the company about deploying Digit, and the new facility will serve as a training ground for the six-foot-tall robot to acquire new skills in environments similar to those it will encounter in the field.

In contrast to many of the newer players in the humanoid arena, Agility does not plan to introduce in-home humanoid robots in the near future. This perspective aligns with that of most independent robotics specialists, who believe the most powerful robots available today are not sufficiently safe for consumer use. Digit currently functions in areas devoid of humans, but version 5, expected to be announced this fall, will feature human-sensing capabilities and won’t have to operate in robot-only zones.

Co-founder and chief robot officer Jonathan Hurst expressed there is ample work to keep Agility engaged in manufacturing and logistics alone.

“Let’s initially focus on the bins and totes, and then we can tackle picking and kitting,” Hurst relayed to TechCrunch. “Subsequently, we can start addressing cardboard, which is quite challenging, along with loading and unloading tractor trailers and similar tasks. At that transition, we’ll be talking about 100 million robots, you know? A trillion-dollar enterprise.”