Yesterday, numerous outlets reported publicly: “Tesla’s third-generation robot has arrived at the AWE show in Shanghai!” Soon after, Tesla officially responded: the robot on display was still the second-generation model. The third generation has not been released at all.

The episode has been called the biggest tech-sector mix-up of the year. Behind it was little more than another round of Elon Musk’s high-stakes hype-building...

Super-Sized Mix-Up

On the morning of March 12, as soon as the AWE show in Shanghai opened, media and visitors flooded toward Tesla’s booth.

A number of tech bloggers rushed in with livestreaming gear, shouting that they were about to “witness history.” “The scene was crazy. You simply couldn’t squeeze in,” one said.

That account came from a tech blogger at the venue. By noon, Cailian Press had published a story saying, “Tesla’s third-generation robot makes its debut.” A wave of other outlets quickly followed.

The puzzling part was that the photos used by the media “looked exactly the same as the second-generation robot.”

By 5 p.m., major outlets began issuing clarifications: Tesla had officially responded that the robot shown on-site was the second-generation model.

“The third-generation unit is currently undergoing final mass-production testing at the Fremont factory and has not been shipped to Shanghai.” The industry was immediately hit with a heavy dose of cold water. How did such a major misunderstanding happen?

In fact, the backdrop was Musk’s relentless hype-building.

At Tesla’s earnings call in January this year, Musk said clearly that the company would “launch the third-generation robot in the first quarter of 2026,” begin production before year-end, and target annual capacity of 1 million units.

That made the AWE show in March the timing many people assumed was inevitable.

At the “Abundance Summit” on March 11, Musk turned up the volume again.

He said the third generation was close to completion, would enter production this summer, and called it “the most advanced robot in the world so far.”

Soon after, on the same day, Tesla’s official Weibo account posted a photo of a Tesla robot making a heart gesture. The industry immediately lost its composure, reading it as a “warm-up for the third-generation robot launch,” which pushed expectations even higher.

By then, the industry was almost certain that the AWE show would be the first appearance of Tesla’s third-generation robot.

After the mix-up, online sentiment quickly shifted from “feverish anticipation” to ridicule and skepticism.

“Musk is good at this kind of over-marketing and hype,” one comment said. “If it didn’t even appear in March, how could annual production of 1 million units possibly happen?”

A Heart Gesture Was Enough to Rattle the Industry

The intense enthusiasm around the third-generation robot came not only from Musk’s hype, but also from the shock of that “heart gesture” photo.

Industry observers said: “One photo of a heart gesture rattled the entire sector.” If you didn’t look at the full body, you might think it was a human hand wearing a glove.

Tesla’s third-generation robot, Gen 3, is said to use a hand with 22 joints and 50 precision actuators, bringing its dexterity close to that of a real human hand.

Is that hand impressive? China can also build 22 joints.

Domestic players such as AgiBot and Unitree have already built dexterous hands with 19 or 20 degrees of freedom, while Xpeng’s dexterous hand, like Tesla’s, has 22 degrees of freedom.

But Tesla’s hand is different in two ways. First, it is lighter. Earlier dexterous hands placed motors in the palm, making them bulky and thick.

Tesla’s current hand has no heavy motor inside it. Instead, “wires” in the forearm pull the fingers, much like controlling a puppet.

Second, it can sense. Many domestic hands with 19 to 22 degrees of freedom can only move.

Tesla’s hand, however, is covered with tactile sensors. In other words, it can not only move, but also feel.

It is said to be able to peel an egg and catch a tennis ball. This hand is where the real fingertip magic lies. But after the amazement come very practical questions.

The first is durability. Fifty actuators mean 50 potential points of failure. If the hand is put into a factory and works eight hours a day, how fast will it wear out?

In October 2025, Fast Technology reported that the main reason Gen 3 had been delayed again was its hand: under high-intensity training for parcel sorting, it lasted only six weeks.

When Figure 02, whose hand structure is simpler than Gen 3’s, was “retired” from the factory, both of its hands were already badly worn.

The second issue is cost. Although more than 60% of Tesla’s core components are now backed by China’s supply chain, bringing the estimated cost of a single dexterous hand down from $6,000 to under $4,500, Musk has set a target price of around $20,000 for the entire Optimus robot. That leaves only about $11,000 to cover the torso, legs, computing unit and all other parts, making the budget extremely tight.

And even if the lifespan of the dexterous hand has improved significantly, some transmission components still last less than two months. No wonder some investment banks have said bluntly that this is unnecessary “overdesign”: attractive, but uneconomical.

Still, judging by the specifications, this pair of hands does appear to sit at the technical ceiling.

Mass Production Only by Year-End

Beyond doubts about the dexterous hand, attention is also focused on Gen 3’s payload and battery life.

Tesla says Gen 3’s weight has fallen from 63 kg to 57 kg. A lighter robot becomes faster: Tesla says its “top speed can reach 1.2 meters per second.”

It may be lighter, but its payload is weaker. Gen 3 is said to have an overall load capacity of about 20 kg. By contrast, China’s Galaxea AI, which focuses on heavy-duty handling, says its robot can steadily lift 50 kg with both arms.

Current tests show that humanoid robots such as Gen 3 typically last only about two hours under heavy workloads.

But Tesla says “Gen 3 is expected to have a battery life of about eight hours.” Many Chinese vendors, by contrast, use “fast battery swapping.” UBTech’s Walker S2, for example, can unplug and replace its own battery when it runs low, reportedly completing the process in three minutes.

In January this year, although some media outlets loudly claimed that the factory had already “started mass production,” Musk was candid: “The robots are not yet doing truly useful work, and large-scale mass production will have to wait until the end of the year.”

The activity for now is most likely just a pilot production line. Musk now claims Gen 3 will “fully get rid of remote operation and achieve true full autonomy,” and even work “for 10 consecutive hours.”

Everyone wants to know whether Musk is exaggerating: how long can Gen 3 actually work independently without any human intervention?

As of today, Tesla has never disclosed that figure.

Musk’s hype machine has never lacked marketing skill. But the technical exploration behind that “heart gesture” photo has already redrawn the competitive map for the industry.

There is still a long way to go from “concept-driven frenzy” to practical deployment...