Others, however, saw the potential for a new front in a battle with Elon Musk over robots.

In Silicon Valley these days, it can be hard to keep track of how the various tech lords are trying to get a leg up on each other, and what they’re competing over at any given moment. The Musk-Zuckerberg rivalry—once headed to the Colosseum for a cage fight—could be one for the ages, up there with the feud between Bill Gates and the late Steve Jobs.

So far, however, the stakes between them have been comparably small. Mostly, just bragging rights about the sizes of their manliness or fortunes. The 2023 introduction by Meta Platforms of a text-based social-media platform, Threads, to compete with Musk’s X added some meat to things.

Now, things could get really interesting if Zuckerberg’s Meta and Musk’s Tesla are headed toward a collision involving humanoid robots.

Thanks in part to Musk’s showmanship, the prospect of humanoid robots is one of the hottest fields in tech.

Long the stuff of sci-fi fantasy, the technology holds the promise of becoming more real with recent advances in artificial intelligence. Whereas robots previously could be programmed to perform specific tasks, such as stacking boxes, it was a rigid and narrow set of abilities. Now, roboticists say they can use AI that learns from watching humans perform tasks, greatly opening up the use cases for robots.

“The promise of that is basically now we can very quickly add a whole host of new tasks where anything becomes fair game,” Jeff Cardenas, co-founder of a humanoid robot company called Apptronik, told me for Friday’s “Bold Names” podcast. “And that kind of changes everything about what robots are capable of and where they can be applied in the future.”

Part of the bull case for why vehicle maker Tesla is so well positioned in robotics owes to its work to develop computer vision to help robot cars see and navigate the world. It already has eight million vehicles on the roadways, collecting video data that helps improve its systems.

A natural evolution, Musk has argued, is humanoid robots.

In May, Milan Kovac, then a Tesla vice president, provided some insight into how the company was working to develop its robot, dubbed Optimus. “One of our goals is to have Optimus learn straight from internet videos of humans doing tasks. Those are often 3rd person views captured by random cameras etc.,” he posted on X. “We recently had a significant breakthrough along that journey, and can now transfer a big chunk of the learning directly from human videos to the bots (1st person views for now).”

With its own first-person views, Meta’s glasses, too, could be well positioned to collect reams and reams of video data crucial for robot development.

The newest version of those glasses goes on sale Tuesday and includes a video display within the lens and a built-in camera able to capture what the user sees. Meta tells users that certain video and audio data can be used for improving its products.

One of the demonstrations during the Meta event on Sept. 17 was supposed to include a chef cooking food for a party with the help of instructions from the AI watching him through the glasses. A technical hiccup, however, ended that effort.

Still, Adam Jonas, the well-known Morgan Stanley analyst, argues that within two years Meta could have 20 million pairs of its glasses in use around the world—almost twice the number of Tesla vehicles expected to be on the roadways by then.

“Every Meta glasses user may be training a humanoid avatar iterated in simulation across billions of scenarios in a digital omniverse,” Jonas wrote to investors in a note this past week.

In other words, Meta could be sitting on a gold mine of the world’s most boring home videos that could be useful to robots trying to learn how to navigate domestic life.

That’s when things could get really meta. Imagine: The Meta AI instructing you step-by-step how to make dinner while in turn gathering video to later train a robot how those instructions manifest in the physical world.

While Meta hasn’t said if Live AI, what it calls its feature in the consumer glasses, will be used to gather data for robot training, it has other wearables aimed at just that. The new generation of its Project Aria research glasses, announced earlier this year, were designed, according to the company, to gather data for AI and robotics.

By 2040, Musk predicts, there will be at least 10 billion humanoid robots in the world, remaking the idea of work and life—even if some roboticists say such technology is still a work in progress.

So far, Zuckerberg’s ambitions for robots are less clear.

Meta’s efforts are still in the early stages, having emerged from its broader work in AI and extended reality devices. “If you’re thinking about having an AI that is always on, using cameras and microphones to assess the situation that it’s faced with, to be an assistant to you through a wearable device—that’s actually pretty similar to a robot who’s got the same kind of packages in terms of how it should understand the world,” Andrew Bosworth, Meta chief technology officer, said in a social-media post in June.

Earlier this year, Meta hired Marc Whitten, the former CEO of General Motors’ autonomous-car subsidiary Cruise, to oversee a new robotics effort. Among other recent recruits is Sangbae Kim, a Massachusetts Institute of Technology robotics professor, to lead research and development as Meta’s robotics architect. He had attracted attention for his MIT lab’s development of a small four-legged “cheetah” robot that could run quickly.

Those hires have garnered far less attention than the more-recent spate of big AI names—and a recruitment drive that has included hundred-million-dollar pay packages. That new effort is aimed at helping Meta catch up with the likes of OpenAI and others in developing advanced AI—“superintelligence” as Zuckerberg calls it. He has said such technology is at the heart of the glasses and his effort to enable a new computing paradigm.

Amid worries of an AI bubble, Zuckerberg says he’s trying to make the right bet.

“If we end up misspending a couple of hundred billion dollars, I think that is going to be very unfortunate obviously, but what I’d say is I actually think the risk is higher on the other side,” Zuckerberg said on the “ACCESS” podcast earlier this month. “If you build too slowly…then you’re just out of position on what I think is going to be the most important technology that enables the most new products and innovation and value creation in history.”

Whatever the cost, a robot cage match could be fun.

Write to Tim Higgins at tim.higgins@wsj.com