On March 5, 2026, Wu Changzheng, former president of “Magic Atom,” confirmed his departure. Some media outlets said he is preparing to start his own company.

Corporate registration records show that Wu began a phased exit in mid-January this year: he first stepped down as general manager of the Wuxi headquarters, then successively relinquished his legal representative roles at three subsidiaries in Beijing, Shanghai and Jiangsu.

Co-founder and CTO Chen Chunyu has now taken over the core roles, while the legal representative of Magic Atom Robot Technology (Wuxi) Co., Ltd. has also been changed to Chen.

As for the reason for Wu’s departure, some media cited outside sources saying it stemmed from “differences over development philosophy,” but Magic Atom has explicitly denied that claim.

More certain than unverified rumors is this: at a critical industrialization window for humanoid robots, the founder has chosen a new arena and is re-entering the game.

Two Months, a Gradual Exit

Wu Changzheng’s departure looks less like a sudden personnel shock than a carefully prepared handover.

From stepping down as general manager of the Wuxi company in mid-January 2026 to completing the legal representative change on March 5, Wu spent nearly two months transferring management roles at several affiliated companies to Chen Chunyu one by one.

The process followed a clear cadence. Notably, the handover coincided with Magic Atom’s brightest and most crucial moment.

On January 23, Magic Atom announced it had become a strategic partner of CCTV’s Spring Festival Gala. On Lunar New Year’s Eve, two of its humanoid robots appeared on stage and even completed what the company called the industry’s first “Thomas 360°” stunt.

It was the company’s biggest brand exposure since its founding two years earlier. Also on January 23, co-founder Gu Shitao said publicly that the company was moving as fast as possible on its schedule and could seek an IPO as early as 2026.

Seen another way, the Spring Festival Gala needed a founder to stand behind the brand, while an IPO requires a stable management transition.

As a core executive, Wu completed a smooth transfer of all his roles after the Spring Festival Gala project landed successfully and before the IPO process fully opened up. That may have been the safest way to keep the company’s formation intact.

One Company, Two Narratives

Public interviews with Magic Atom executives make visible a split inside the company over the pace of commercialization.

Wu Changzheng, Magic Atom’s former president and a technologist by training, is a thorough long-termist. He has insisted on spending heavily to push through fundamental R&D for general-purpose humanoid robots, tackling hard problems such as high-torque joints and whole-body motion control algorithms.

“He [Wu Changzheng] has always refused to compromise for short-term commercialization,” one person familiar with the company said.

The shareholders, however, were anxious. They wanted Magic Atom to accelerate commercialization and the IPO process, and to bet on 2026 as the first year of industry-scale deliveries.

Capital’s ultimate goal, of course, is to exit and monetize in time.

Co-founder Gu Shitao has said clearly: “We are advancing according to the fastest listing timetable and will complete this task in 2026.”

“There were also internal disagreements over whether to spend 100 million yuan on the Spring Festival Gala,” the person familiar with the matter said. Investors supported the Gala appearance, seeing it as a necessary route to lift valuation and speed up the IPO.

One side wanted to build steadily; the other wanted to cash out in time. These are two fundamentally opposed paths. The person familiar with the matter said: “Magic Atom was tied down by capital, forcing the founder out.”

But Magic Atom has explicitly denied the claim, directly rejecting the idea of “differences in philosophy” and saying “that is not the case.”

Embodied AI in 2026 Waits for No One

The broader backdrop to Wu Changzheng’s departure is a brutal shakeout across the embodied AI industry.

The market from 2025 to the start of 2026 has been a study in extremes.

On one side is clearing: more than 20 robotics companies globally have faced bankruptcy or layoffs, with Aldebaran, CloudMinds and others falling one after another;

On the other side is fundraising: leading players are accelerating. Galaxea AI completed a new 2.5 billion yuan financing round in March 2026, holding the top valuation among unlisted Chinese humanoid robot companies.

Capital is concentrating faster around a very small number of giants.

For second-tier players, the choice is either to go public quickly and secure “lifeline money,” or be cleared out. Magic Atom’s decision to push hard for an IPO in the current window makes full commercial sense.

But an IPO sprint often means delivering better-looking short-term financial numbers.

That creates a certain mismatch with Wu Changzheng’s commitment to “long-termism,” because foundational technical breakthroughs require sustained cash burn and have extremely long payback cycles.

This is the single-choice question facing every technology-driven company: keep grinding on long-horizon technology, or narrow the front and prioritize commercialization? Wu chose the former, just from a different arena.

Given his hard-core technical background at Xiaomi’s “CyberDog” and Pudu Robotics, his new direction will most likely remain within embodied AI. Magic Atom, meanwhile, is now under Chen Chunyu’s operational control.

The company’s external signal is also clear: the team is stable, and the IPO timetable remains unchanged.

With that, technical conviction and commercial reality have completed a “peaceful” split.

From 2025 to 2026, similar stories have played out more than once: technologists leave, operators take over, and capital pushes for a listing.

As embodied AI reaches the eve of industrialization, how much time is still left for “slow refinement”?

Now, Wu Changzheng may take his technical obsession to the next table, while Magic Atom keeps running forward with its IPO timetable.

Which route is right, “the best technology” or “the fastest listing,” remains to be seen.