Can Figure Catch Optimus?

We are witnessing a meteoric rise of technological capability…

So naturally, in yesterday’s Bleeding Edge – Will AI Take Our Jobs?, we explored a question many of us have been asking.

Artificial intelligence (AI), as software, will enable the largest productivity growth in history, augmenting human capabilities and performing those time-consuming, menial tasks we all loathe doing, freeing up time for more productive and enjoyable endeavors.

Today, we’re going to consider this from a different angle.

What about the labor shortages around the world that hold back economic growth and productivity? What about those millions of unfilled jobs in essential roles in warehouses, logistics facilities, manufacturing, and transportation… where labor shortages are acute and attrition is high?

Take the trucking industry for example. There are millions of unfilled trucking jobs in developed countries around the world. In the U.S. alone, there is an annual labor shortage in trucking of around 100,000 drivers, and the number increases by the year. By the end of this decade, it will be more than 150,000 drivers.

This is why AI – manifested in the form of a self-driving truck – is such a critically important technology. It’s an answer to how to fill the gap of the labor shortage.

The reality is that it is very difficult to attract new, young drivers into the industry. This problem won’t go away. Self-driving technology is the clear solution.

Tesla Semi | Source: Tesla

But what about those tasks that require dexterity and movement – the kinds of things our human bodies are so adept at performing?

I’ve been thinking a lot about this, as 2025 is the year of commercializing another autonomous technology…

Humanoid robots.

“Something No One Has Ever Seen”

We’re probably one to two months away from hearing about Tesla’s third-generation Optimus design, which is already employed in its offices and factories.

And just last week, Figure AI – Tesla’s closest competitor – announced that it has signed its second enterprise customer, which it states is one of the largest U.S. companies.

Figure’s first enterprise customer is BMW, which I’ve written about in the past in The Bleeding Edge and Outer Limits. BMW started using Figure’s humanoid robots at its Spartanburg manufacturing plant in South Carolina and plans to expand the use of Figure’s technology to other locations.

Figure Humanoid Robots Working at BMW | Source:  Figure AI

Between BMW and this new, yet-to-be-named customer, Figure estimates that there is the potential to ship 100,000 humanoid robots to just these two customers over the next four years.

That’s just for two enterprise customers. The market potential for humanoid products is easily more than 1 billion units over the next 10–15 years.

Even more significant than Figure AI’s recent customer win, though, was a hint given by Figure’s CEO, Brett Adcock, recently, when he made this statement…

Last week, we successfully began running an end-to-end neural network on the new client’s use case.

And as a further indication of Figure’s confidence in this new approach, yesterday Adcock announced that Figure would be ending its collaboration with OpenAI.

Source: X @adcock_brett

This may not sound like much…

But it is.

OpenAI was a significant investor in Figure’s Series B round in February 2024.

Collaboration was part of the deal.

The Ultimate Consumer Device

The reality is that Figure has been way behind Tesla on the artificial intelligence software required for autonomous, learning humanoid robots. It had been relying heavily on OpenAI’s multi-modal AI models to fill that deficit…

But I always saw that as a stop-gap measure. This approach couldn’t compare to Optimus’s own on-board neural network – a literal synthetic brain.

In order to manifest a general-purpose humanoid robot – capable of performing tasks autonomously in an enterprise setting or a home setting – the solution must be a neural network with advanced reinforcement learning.

In other words, predicting and programming millions of individual tasks into a robot’s software is simply not scalable. Instead, end-to-end neural networks – capable of learning and improving through visual data sets and trial-and-error – can scale quickly.

Once a single Optimus or Figure humanoid robot masters a single task, every robot in the fleet has mastered that task with a simple download of the latest software update.  It’s the same concept with our smart phones, laptops, or Tesla electric vehicle.

That’s why having multiple enterprise customers and scaling production is so important. Every humanoid robot is a data collection and learning “computer,” much in the same way that Tesla’s “fleet” of more than 7 million vehicles collects and learns from video data every single day.

Figure’s software architecture change is a major upgrade and step forward for the company.  It’s also a tacit admission that Tesla’s approach to autonomy – regardless of whether or not it is self-driving cars or humanoid robots – is the right path forward.

Others will follow quickly.

Necessitating Humanoids

It was the Greek philosopher Plato who wrote, “Our need will be the real creator,” which morphed into the proverb, “Necessity is the mother of invention.”

The “need” has long been solving the problem of the labor shortages – almost 8 million unfilled jobs in the U.S. market alone.

Source:  U.S. Chamber of Commerce

And despite there being 7.1 million unemployed, most of these jobs are still unfilled. The reasons vary:

• Almost half of those who lost their jobs during the pandemic are unwilling to take a job if it is not remote work
• 17% have simply decided not to work any longer
• 14% have decided to only work part-time
• Almost 25% say government benefits incentivize them to not actively look for work
• 36% of the younger workforce aged 25-34 decided to focus on “prioritizing personal growth” before re-entering the workforce

We can clearly see the impact of these trends in the labor force participation rate. It had been on a healthy uptrend until February 2020… and then the pandemic turned everything upside down. And while labor participation rebounded through 2023, it has been falling again for the last year.

Source:  U.S. Chamber of Commerce

Even if labor force participation returned to levels seen in February 2020, it would only return 2.75 million workers to the labor force. That would still leave almost 5 million jobs unfilled.

Source:  U.S. Chamber of Commerce

Irrespective of the reasons for those staying out of the workforce, our reality is that labor shortages won’t go away. Many of the trends that we are seeing now will continue. Many jobs are simply undesirable.

Said another way, the need for a solution will only get stronger.

Why hasn’t this need been filled? Well, even two years ago, the technology didn’t exist to solve the problem. We knew what we needed to invent, we just hadn’t invented the technology yet.

Now we have…

And now we can scale.

Jeff