Sony's Ace Robot Defeats Elite Table Tennis Players in Nature Study

A Landmark Moment for Robotics

Sony has published groundbreaking research in the journal Nature demonstrating that its robotic arm, named Ace, can defeat elite human table tennis players in regulation matches. The company describes it as the first time a robot has achieved human, expert-level play in a commonly played competitive sport in the physical world. The result lands during a week packed with humanoid robotics milestones across multiple continents.

How Ace Plays

Ace is a paddle-wielding robotic arm built around reinforcement learning and sophisticated sensing. It uses eight joints to control the racket with remarkable precision, while nine cameras positioned around the court track the ball's trajectory and spin in real time. This sensor array allows the system to anticipate shots, plan returns, and execute strokes at a speed and accuracy that rivals trained humans.

The Match Results

Under International Table Tennis Federation rules, Ace competed against five elite amateurs and two professional players. Against the elite group, it won three of five contests. After the paper was submitted for publication, follow-up matches in December 2025 and March 2026 saw Ace defeat professional-level opponents too, suggesting the system is continuing to improve beyond the original study conditions.

A Bigger Picture for Humanoids

The Sony result is one of three major robotics stories unfolding simultaneously. In Beijing, a bipedal robot named Lightning, built by smartphone maker Honor, completed a 21-kilometre half-marathon in 50 minutes and 26 seconds, beating the current human world record by nearly seven minutes. At Germany's Hannover Messe, humanoids such as Agile Robots' Agile ONE have been demonstrating factory-floor tasks alongside human workers, even greeting the German Chancellor.

Why It Matters

Table tennis demands millisecond-level reflexes, spatial awareness, and anticipation, qualities long considered difficult for machines. Ace's success suggests that reinforcement learning combined with rich multi-camera sensing can close the gap between simulation and the messy physical world. Combined with the marathon and factory-floor demonstrations, this week signals that embodied AI is moving from laboratory curiosity to real-world capability at a startling pace.