Robots broke another human record

Over the weekend in Beijing, a humanoid robot named “Lightning” ran a half-marathon… and won.

It finished the 21 km race in just 50 minutes. That’s faster than the human world record. The best human runner that day came in more than 17 minutes behind.

A year ago, robots were finishing races like this in over two and a half hours. Now they’re outrunning elite athletes.

2026 will be the year of the robot. Time to buckle up...

There are now millions of robots operating across factories worldwide, with the global fleet more than doubling over the past decade.

As you can see, annual installations have surged past half a million units and held that level for four straight years:

Source: World Robotics 2025

But these aren’t the versatile helpers we dreamed of. They’re more like industrial savants—brilliant at one specific task, useless at everything else.

These robots are incredibly precise, but incredibly dumb. A robot arm can weld the same spot perfectly a million times. But move the target by an inch, and it’s lost. Want it to do something new? That means weeks of reprogramming.

Good news. Robots just got a brain transplant, courtesy of artificial intelligence (AI). The ChatGPT moment for robotics is coming, and it’s about to change everything.

Think about how you’d teach a child to fold laundry. You wouldn’t hand them a thousand-page manual covering every possible way a shirt could be crumpled. You’d show them a few times, and they’d figure it out.

Until now, robots couldn’t learn like this. They needed that thousand-page manual for every task.

Figure AI recently showed off its humanoid robot working inside a BMW factory. Previous robots needed tightly controlled environments and rigid programming for every step.

Figure’s robot learns tasks by observing humans and refining its actions through AI models. It can pick up parts, place components, and adapt when something shifts out of position without needing to be reprogrammed.

A robot apprentice learning on the job.

At Google DeepMind, researchers developed a general-purpose robot system that can perform hundreds of different tasks after being trained on internet-scale video and data.

It doesn’t just follow instructions. It adapts in real time. If an object is moved or a task changes slightly, the robot adjusts and continues without breaking the workflow.

The secret behind these innovations is the same breakthrough that supercharged language models: the transformer.

Transformer algorithms allowed computers to not just read words one by one but understand context. How each word relates to every other, predicting what comes next.

Just as ChatGPT understands the relationships between words to write coherent sentences, these new AI robots understand the relationships between objects and actions in the real world. They’re gaining “physical intelligence.”

Do you recall when Nvidia (NVDA) CEO Jensen Huang said, “Everything that moves will be robotic.” 

I agree, and the revolution is already unfolding all around us.

Delivery drones humming overhead. Compact warehouse robots that can squeeze through narrow aisles. Robotic arms that can work alongside humans in restaurants and factories. Even unassuming robot vacuum cleaners, like Matic, are finally becoming genuinely effective home helpers.

A decade from now, purpose-built machines embedding AI and physical intelligence will seem as normal as smartphones are today.

Figure’s humanoid robots used to be like students memorizing a massive rulebook. See the cup, grab the cup. See the bag, move the bag. That made them impressive, but brittle.

Figure threw out the rulebook and replaced that hand-coded approach with a vision-language-action model called Helix. Instead of following fixed instructions, its robots can now pick up thousands of objects they’ve never seen before just by responding to natural language prompts.

In Figure’s tests, the robots handled cluttered household items ranging from glassware to toys, tools, and clothing without prior demonstrations or custom programming for each object.

As this kind of AI scales, we’ll see another big jump forward. The skeptics who claimed general-purpose robots were decades away are about to be humiliated.

Humanoid robots will likely be the first widespread AI machines that change physical work. And once they can reliably see, reason, and act in the real world, the future is going to arrive a lot faster than most people think.

And humanoid robots are just the visible frontier of the robo-revolution.

Beneath the ocean’s surface, Irish startup Ulysses is deploying teams of autonomous underwater drones—nicknamed “Robo Sharks”—to become marine gardeners. These underwater robots are planting seagrass, a vital ecosystem that captures carbon 35X more effectively than rainforests.

On land, precision-farming robots reduce herbicide use by 90%, their sensors detecting and treating individual weeds instead of spraying entire fields. Meanwhile, driverless tractors guided by satellites plow perfect rows through the night.

As Jensen said, “Everything that moves will be robotic.” Get ready for all kinds of weird and wonderful AI robots to enter our world.

Even robotic police dogs—like “Roscoe” in Massachusetts, which recently faced gunfire and likely saved human lives in a standoff.

Today, robots in factories are still mostly one-trick ponies.

Programmed for repetitive motions in choreographed settings, repeatedly making the same weld in the same spot on an assembly line or dropping the same item into the same box.

AI allows robots to learn new things.

Soon, the same robot that unloads trucks in the morning could help assemble products in the afternoon and organize inventory at night. We’re moving from single-purpose machines to versatile helpers that can understand and follow instructions.

For decades, American manufacturing has lost out to countries with cheaper labor. But AI changes the math. Now, robots can work 24/7, never get tired, and can learn new tasks overnight.

Suddenly, America’s advantages—cheap energy, strong infrastructure, and AI expertise—matter more than low wages.

Picture a handful of people overseeing thousands of robots running a factory. Now, imagine if America were full of these factories. We could produce 1,000X more stuff for a fraction of today’s costs. When THAT happens (it’s already started), it could create trillions of dollars in wealth.

Remember the old dream of “Made in USA”? It’s felt like a nostalgic slogan.

Take phones. A company called Purism created a fully “Made in USA” phone. Great idea, but it costs $1,600… and you can get an equivalent Android phone for $300. That’s the old math. With AI-powered robots, we can change that equation.

It’s no exaggeration to say self-learning robots can reindustrialize America. They can turn “Made in America” from a feel-good slogan into a badge of cutting-edge innovation.

This will be one of the biggest investment opportunities in the next decade. And it’s only getting started.