Nvidia is betting on robotics to drive future growth

Unlock Editor’s Digest for free

Roula Khalaf, editor of the FT, picks her favorite stories in this weekly newsletter.

Nvidia is betting on robotics as its next big growth driver as the world’s most valuable semiconductor company faces increasing competition in its core artificial intelligence chipmaking business.

The US tech giant, best known for the infrastructure that has supported the AI ​​boom, will launch its latest generation of compact computers for humanoid robots – called Jetson Thor – in the first half of 2025.

Nvidia is positioning itself as the leading platform for what the technology group believes is an impending robotics revolution. The company sells a “full-stack” solution, from the software layers used to train AI-powered robots to the chips built into them.

“The ChatGPT moment for physical AI and robotics is upon us,” Deepu Talla, Nvidia’s vice president of robotics, told the Financial Times, adding that he believes the market has reached a “tipping point.”

The push into robotics comes as Nvidia faces increasing competition for its powerful AI chips from rival chipmakers such as AMD, as well as cloud computing giants such as Amazon, Microsoft and Google, which are looking to reduce their dependence on the US semiconductor giant.

Nvidia, whose value has risen to over $3 trillion due to huge demand for its AI chips, has positioned itself as an investor in the “physical AI” space to help grow the next generation of robotics companies.

In February, it was one of several companies, including Microsoft and OpenAI, to invest $2.6 billion in humanoid robotics company Figure AI.

Robotics has so far remained an emerging niche that is not yet generating major returns. Many startups in this space are struggling to scale, reduce costs, and increase the accuracy of robotic products.

Nvidia doesn’t provide a breakdown of robotics product sales, but they currently represent a relatively small share of total revenue. Data center sales, which include the coveted AI GPU chips, accounted for about 88 percent of the group’s total revenue of $35.1 billion in the third quarter.

However, Talla said a shift in the robotics market is being driven by two technological breakthroughs: the explosion of generative AI models and the ability to train robots using simulated environments based on these basic models.

The latter is a particularly significant development because it helps close what roboticists call the “sim-to-real gap” and ensure that robots trained in virtual environments can work effectively in the real world, he said.

“In the last 12 months. . . (This gap) has matured to the point where we can now do simulation experiments combined with generative AI, which was not possible two years ago,” Talla said. “We provide the platform that allows all of these companies to do all of these things.”

Talla joined Nvidia in 2013 to work on its “Tegra” chip, originally intended for the smartphone market. However, the company quickly changed, and Talla oversaw the redeployment of about 3,000 engineers into “AI and autonomous training (e.g. for vehicles).” This was the origin of Jetson, Nvidia’s line of robotic “brain” modules that launched in 2014.

Nvidia offers tools in three phases of robotics development: basic model training software, derived from Nvidia’s “DGX” system; simulations of real-world environments in its “Omniverse” platform; and the hardware that enters the robots as “brains”.

Apptronik, which relies on Nvidia’s technology to develop humanoid robots, also announced a strategic partnership with Google DeepMind in December to improve its products.

According to US market researchers BCC, the global robotics market is currently valued at around $78 billion and is expected to reach $165 billion by the end of 2029.

Amazon has already used Nvidia’s robotics simulation technology for three of its warehouses in the US, and Toyota and Boston Dynamics are other customers using Nvidia’s training software.

David Rosen, who directs the Robust Autonomy Lab at Northeastern University, said the robotics market still faces major challenges, including training the models and verifying their safety in use.

“Currently we do not have very effective tools for checking the safety and reliability properties of machine learning systems, especially in robotics. “This is an important open scientific question in this field,” Rosen said.