Paris Fashion Week is no stranger to a gimmick. There was Coperni spraying a dress onto a model in 2022, followed by Schiaparelli’s faux animal heads a year later, and then Robert Wun’s blood-splattered “horror couture” last year.
This week’s event in the City of Light hewed to form as Chinese humanoid robot N2, created by Beijing-based Noetix Robotics, strutted awkwardly down a catwalk attired in waistcoat and pearls in the first outing of its kind outside of China.
“Humanoid robots look like human beings,” Noetix CEO Jiang Zheyuan tells TIME. “So maybe new sparks could be created through collaborating with the fashion world.”
Unfortunately, N2’s turn came after the close of the official Paris Fashion Week as a planned collaboration with a Chinese designer fell apart at the 11th hour over a funding dispute. Instead, N2 made do with modeling three outfits sourced from a local vintage shop at the popular UNESCO venue before the assembled press—as well as charming onlookers with street acrobatics.
Still, N2’s debut in Paris is another sign of the burgeoning global humanoid robotics industry, which Morgan Stanley predicts could be worth $5 trillion by 2050. Of course, N2 is far from the first robot involved in fashion. Back in 1999, Alexander McQueen featured industrial robot arms spray-painting a runway model, while 10 years later Tokyo Fashion Week was hosted by a black-haired female robot dubbed HRP-4C. In 2023, Coperni had robot dogs cavorting with human models, while Shanghai Fashion Week already debuted humanoid robots on the catwalk in March.
That Shanghai, the “Paris of the East,” scooped the original is, of course, unsurprising. Although the West has humanoid robotics players such as Boston Dynamics and Tesla, China occupies the industry’s vanguard thanks to its huge market, a mature industrial base of chips, sensors, and batteries, as well as robust state support.
In March, China unveiled a one trillion RMB ($137 billion) fund to support transformative technologies including AI and robotics. China’s cities and regions are also clamoring to seed a local robotics champion via a flurry of grants and subsidies. Last year, 31 Chinese companies unveiled 36 humanoid models—but only eight by their American counterparts.
And while N2’s Parisian outing seems at first blush like another gimmick—at 3 ft, 9 in, it certainly isn’t blessed with your quintessential model physique—it’s also an exercise that could bring concrete benefits. A significant problem for the development of humanoid robotics is identifying real uses. For while they dazzle on social media performing flips, running a half-marathon, dancing, boxing, and playing soccer, in terms of performing daily tasks that are, well, actually useful, they still pale in comparison to their breathing, sweating counterparts.
One key reason is a lack of data. While large language models such as ChatGPT or DeepSeek made giant leaps by parsing huge swathes of online information, this is practically all “2D” data: words, numbers, pictures, and sounds. By comparison, the “3D” data necessary to instruct humanoid robots how to move and react to stimuli in the real world—known as “embodied AI platforms”—is extremely sparse.
“If you want a so-called AI-empowered humanoid robot that actually moves like us, that needs scenario training and real life data,” says Grace Shao, a former Alibaba manager turned IT consultant who publishes the AI Proem newsletter. “So that’s a huge bottleneck for the industry.”
The race to dominate the humanoid robotics market is as much about harvesting 3D data as finessing the nuts and bolts. Last year, Shanghai authorities provided rent-free data collection premises for local firm AgiBot to have 100 robots repeating mundane tasks such as stacking shelves, folding clothes, and pouring cups of tea for 17 hours every day. However, it would be much faster and cost effective to harvest that data organically via industrial deployment—steadily improving processes while reaping the revenue necessary to foster economies of scale.
Could fashion be the answer? Chinese tech consultant Emma Meng thinks so. After graduating in French literature at Cambridge University, Meng worked as an assistant to the publisher of a New York fashion magazine—“just like the Devil Wears Prada!” she laughs—while also moonlighting as a fitting model, earning $1,000 a day to essentially have pieces of cloth pinned to her and move her limbs to identify pinch points and judge aesthetic flow.
Meng’s expertise at the crossroads of fashion and technology spurred her to bring Noetix to Paris. She believes that apparel is an ideal industry for robots to be deployed en masse, whether as more economical fitting models or dynamic mannequins that better show off clothes in retail shops. A chief reason is that fashion houses love eye-catching stunts and have cash to burn, she says, as illustrated by the $88.3 billion revenue posted last year by fashion powerhouse LVMH Moët Hennessy Louis Vuitton.
“They are ready to spend a lot of money just to get people’s attention,” says Meng. “So wouldn’t it be nice if they could commercialize these first-generation robots, even though they can only do flashy but pretty useless things.”
Case in point: the N2 may excel at flips but struggles with stairs, which limited N2 to walking single level venues in Paris. It’s a quirk that illustrates why, despite all the hype, global humanoid robot shipments will only reach 18,000 units in 2025, according to Merrill Lynch estimates. (Noetix hopes to ship just 1,000 units this year.)
Compared to 6.1 billion smartphone connections worldwide, that’s a minuscule pool of data, especially when spread across multiple competing companies. Speaking at the Beyond Expo tech conference in Macau in May, Michael Tam, chief brand officer of Shenzhen-based robotics firm UBtech, conceded it might take 20 years for mass consumer adoption. As such, the race is on to get these budding butlers, maids, and models into the workforce.
Certainly, China sees the opportunity. “In Beijing, the government asked us to build a robot that can walk the catwalk that will be used in the fashion industry,” says Jiang.
Still, Shao is not so convinced modelling would provide the requisite quality of data to fuel that next developmental leap. “For a mannequin, you can easily just preprogram five poses, or the same for doing a catwalk,” says Shao. “I just don’t think it’s actually that intelligent.”
Derek Zoolander would no doubt disagree. But it remains to be seen whether N2’s catwalk turn launches the next fashion craze or falls apart at the seams.
The post Paris Fashion Week’s Most Important Model Wasn’t Human appeared first on TIME.
