Humanoid robotics: the end of the promise, the beginning of industrialization – Perspectives with Nicolas Halftermeyer

In 2010, at CES in Las Vegas, consumer drones were still just an industrial intuition. Fifteen years later, humanoid robotics is going through a comparable moment. Nicolas Halftermeyer, long-term expert, formerly Parrot then Aldebaran, shares with us in this new issue of Perspectives, his analysis of a field where innovation is reaching unprecedented speed, and which is integrated into all sectors of activity, all in a chessboard dominated by China.

A trajectory in contact with technological cycles

Nicolas Halftermeyer places his career in continuity: pioneer of the internet, then of the internet of things, he was head of marketing at Parrot, a milestone he claims as a founding moment: “I had the honor, with my boss Henri Seydoux, of launching drones at CES Las Vegas in 2010.” Ten years later, change of technological scale: “I took care of the robots Nao and Pepper at Aldebaran.”

Despite the eloquent demonstrations, he reminds us that robotics advances less through “fashion effects” than through long cycles, where the industrial time necessary to stabilize mechanics, electronics, the software chain, maintenance, and, above all, uses is required.

From Pepper to 2026: what has really changed

On the set, the hand of the robot Pepper materializes the gap: “Here is the hand of the robot Pepper, a Franco-Japanese robot, designed in Paris, but obviously under Japanese order from SoftBank.” Pepper, he recalls, costs “20,000 euros”, has 3D sensors, RGB, microphones, speakers, and a “pretty cool” hand, but which “couldn’t really catch much” at the time.

Ten years later, the pivot has less to do with “design” than with motor execution and on-board AI, the real key factors of success. It is on this point that the years to come will play out, particularly with Physical AI.

Count volumes, even imperfect ones

If industrial robotics is not new: “There are 500,000 (robots) installed per year.” Cobotics, robots outside cages, close to operators, is also progressing: “perhaps 50,000 per year.” The humanoid was until recently an exhibition object, “for stands (…) in Las Vegas or at VivaTech”.

In 2026, “It is estimated today that 13,000 robots (…) have been produced.” He talks about “real humanoids”, “our size”, “with two hands”. This volume remains low compared to Asian industrial chains, but it marks a change in category: the humanoid is no longer just an expensive prototype; it is entering a phase where we can pilot, deploy and learn in real conditions. It would be a mistake to say: ‘let’s wait until 2030’ to tackle the subject.

CES 2026: a battle for visibility and a geopolitical index

The CES reveals what the players want to show, and therefore where the marketing and industrial investment effort is located. On site, the observation is quantified: “Out of 40 developers present, 21 were Chinese.” A “dozen” Koreans, “a Singaporean”, then “a German, an English, two French”. The latter presented Miroki (Enchanted Tools) and Richie Mini (Pollen Robotics). As for the Americans, they were barely visible, Tesla and Figure, in particular, not being present.

Safety: robotics ceases to be a toy

The CES images raise a subject that the industry still approaches in a heterogeneous way, namely safety in human environments. When a humanoid falls, “it’s 80 kilos of metal that falls on your foot”. Suffice it to say that solid safeguards, a culture of emergency stopping, and “responsible” behavior must be imposed, particularly if we want robots in contact with the public.

Dual use: the military humanoid is soon here

In the defense sector, drones, he says, have already changed. Quadrupeds (inspection robot type) have concrete uses. The humanoid, on the other hand, is “not yet” the most credible military platform for various reasons, including limited batteries (“one hour, two hours maximum”), terrain constraints, and high sensitivity to conditions.

But he does not close the prospect, “In 10 years, will we really have combat robots? It’s quite probable.”

Europe: skills, but a high risk of dropping out.

We of course discussed the place of Europe, which after having been “very well positioned” with Aldebaran and laboratories (CEA, CNRS), is faced with conditions for scaling up that are too difficult, whether for industrial, financial but also regulatory reasons. He denounces the lack of “opening of the floodgates” on experimentation, in comparison with the United States, where he already sees robotaxis in circulation.

Among the nerve centers in Europe, he naturally cites Zurich as a robotics stronghold (ETH, EPFL) as well as Munich, and insists on the need to move away from a strictly national reading: “We will have to be with our Italian, Spanish, German friends, and our Swiss friends too.” The word that comes up is that of operational cooperation if we want to position ourselves against China where companies like Ubtech, Agibot and Unitree excel.