Beyond ChatGPT: Secret robotics plans and the $38 billion humanoid revolution

Global competition is heating up among large language models (LLMs), with the major players vying for dominance in AI reasoning capabilities and cost efficiency. OpenAI is leading the pack with ChatGPT and DeepSeek, both of which pushed the boundaries of artificial intelligence.

Amid this AI arms race, OpenAI’s latest trademark application with the United States Patent and Trademark Office (USPTO) shows that the organization has other goals beyond LLMs. The application lists various hardware such as AI-powered smart devices, augmented and virtual reality headsets, and even humanoid robots.

Sam Altman, CEO of OpenAI, confirmed to the media that the company is researching AI-powered consumer hardware and is working with several companies to do so. In addition, OpenAI is reportedly developing its own semiconductor technology to support these AI initiatives.

OpenAI’s foray into robotics

For industry tech advocates, the mention of humanoid robots in OpenAI’s trademark application signals a possible return to robotics. The company closed its own robotics division in 2021 and then invested in companies like Figure and 1X. Given advances in AI and sensor technology, OpenAI now seems to be revisiting the prospect of AI-driven humanoid robots for real-world applications.

This turnaround is not surprising, with Goldman Sachs Research, for example, predicting that the humanoid robot market could reach $38 billion by 2035 — a six-fold increase over earlier estimates.

NVIDIA’s AI and robotics offensive

OpenAI is not the only company, however, that is betting on and driving the increasing convergence of AI and robotics. At CES 2025, NVIDIA presented Cosmos, a development platform for World Foundation Models (WFM) that facilitates AI-driven decisions for robotics and autonomous vehicles. Cosmos enables AI models to simulate environments and generate real-world scenarios, accelerating training for humanoid robots.

NVIDIA also introduced the Isaac GR00T Blueprint, a tool for generating synthetic motion that supports the training of humanoid robots using imitation learning. By utilizing extensive synthetic data and reinforcement learning, NVIDIA aims to advance AI-driven physical automation. The Cosmos platform also includes AI models trained on massive datasets, including two million hours of footage on autonomous driving, robotics, and drones.

China’s rapid advances in humanoid robotics

China has been aggressively developing its humanoid robotics industry, with government-led initiatives advancing the goals of mass production by 2025. At a recent event in Beijing, nearly 30 companies showcased state-of-the-art humanoid models. Last year, 102 humanoid robots from 10 companies gathered at a 4,000-square-meter facility in Shanghai to demonstrate tasks such as walking, making beds, washing dishes and even welding.

China’s innovative strength is also highlighted by the integration of robotics into cultural events. During the Spring Festival Gala, humanoid robots performed the Yangge folk dance, combining traditional heritage with advanced AI-driven movement. This is an example of China’s vision of seamlessly integrating robots into everyday life.

The vision for humanoid robots

Elon Musk also remains optimistic about Tesla’s ambitions in the field of humanoid robots. When announcing the results for the fourth quarter of 2024, he announced that Tesla plans to produce thousands of Optimus-Humanoid robots in 2025. Subsequently, production is expected to grow exponentially.

Meanwhile, Figure AI, a California-based robotics company, has ended its partnership with OpenAI, which was only formed in February 2024. On X, Figure CEO and founder Brett Adcock explained that large language models (LLMs) are becoming more and more powerful, but also more and more commoditized. For Figure, LLMs quickly became the smallest piece of the puzzle, while the much more difficult part was developing new AI models for high-frequency robot control.

Few additional thoughts:

We can’t outsource AI for the same reason we don’t outsource actuators, batteries, or electronics – it’s too critical to performance

LLMs are getting smarter yet more commoditized. For us, LLM’s have quickly become the smallest piece of the puzzle, the… https://t.co/tAn7YfZhKl

— Brett Adcock (@adcock_brett) February 5, 2025

At the same time, Adcock announced that Figure had made a major breakthrough in a fully end-to-end robotic AI that was developed entirely in-house.

In the next 30 days, the company will show something that no one has ever seen before in a humanoid robot. Figure has secured several large companies as customers, including BMW. The company plans to deliver 100,000 robots over the next four years.

AI and robotics — a symbiotic development

The exponential advances in AI, particularly in large language models and machine learning, are laying the foundation for the next generation of humanoid robots. AI-driven humanoids show great promise for industrial applications, from manufacturing to logistics and beyond. However, challenges remain, including regulatory frameworks, cost efficiency, and hurdles to market adoption.

As AI and robotics continue to develop in parallel, the coming years will determine how seamlessly these intelligent machines can be integrated into human-centered environments. Whether in industrial automation, personal assistance or even entertainment, the AI-first approach is shaping a future in which humanoid robots will play an increasingly central role in society.

Jan Burian is a global speaker, analyst and author covering disruptive digital technologies and the industrial sector. He is currently the head of industry insights at Trask, responsible for overseeing and executing the go-to-market strategy from a marketing perspective, with a primary focus on the DACH region, particularly Germany. The role includes content marketing, developing media relationships and tracking IT and technology trends, especially in the automotive and industrial sectors.