Nvidia has long been known as the world’s leading AI computing company, powering everything from video games to cutting-edge research. Now, under CEO Jensen Huang’s vision, the company is aggressively expanding into physical applications of AI, namely self-driving cars and advanced robotics, in a bid to transform itself into a dominant deep tech superpower.
Huang believes that beyond artificial intelligence itself, robotics will be Nvidia’s biggest growth market, with autonomous vehicles as the first major commercial application.
International Finance will examine how Nvidia is leveraging its AI prowess to drive breakthroughs in autonomous driving and robotics, and what it means for the company’s future and the tech industry at large.
From GPU king to AI powerhouse
Nvidia has a very underdog story. It was established in 1993 and was almost set to be named GeForce. The name Nvidia comes from the Latin word ‘invidia’, which means envy (or green with envy). In early 2006, they put everything they had into CUDA (Compute Unified Device Architecture), which began paving the way for AI’s rise. It’s important to note that there would be no ChatGPT without Nvidia.
The company was a household name in the early 2000s, being the top supplier of graphics cards that were necessary to play high-quality video games. Their top competition back then was AMD, which in the 90s had considered acquiring Nvidia. Today, Nvidia dwarfs AMD, as they are single-handedly pioneering AI, robotics, self-driving cars, chip production, and even building supercomputers. The company is considered a major source of inspiration for Silicon Valley enthusiasts. Notably, Nvidia CEO Jensen Huang was once rejected by Apple, yet he went on to become a titan in the industry and has been referred to as the “Steve Jobs of AI Hardware.”
However, the visionary cofounder of Nvidia knows the semiconductor industry can be cyclical, with data centre investment coming in “booms and busts.” To secure Nvidia’s place in the tech stratosphere long term, Huang has been scouting the next big market beyond conventional “Big Tech” pursuits. At Computex 2024, he declared that two “high-volume” products will dominate robotics in the future: self-driving cars first, and eventually humanoid robots.
These technologies are converging thanks to advances in machine learning, and both require human-like perception, split-second decision-making, and immense computing power, precisely what Nvidia specialises in.
In Huang’s words, “Every single car company will have to be autonomous, or you’re not going to be a car company,” a bold prediction underscoring his belief that autonomy is the future of transportation.
Accelerating the self-driving car revolution
Nvidia has methodically embedded itself at every level of the self-driving technology stack. Huang often describes a three-part approach: one computer to train AI models, another to run simulations, and a third inside the vehicle for real-time driving. By providing the chips and software for all three stages, Nvidia aims to be the go-to enabler of autonomous vehicles rather than a consumer-facing automaker.
“Nvidia has strategically embedded itself in all three key steps that could make every car a self-driving car,” notes Business Insider.
While companies like Waymo and Tesla build robotaxis and personal electric vehicles, Nvidia positions itself as the behind-the-scenes supplier powering those efforts.
Nvidia Drive, the company’s end-to-end autonomous driving platform, exemplifies this strategy. It includes powerful automotive systems-on-chip (SoCs), like the current Drive Orin and next-generation Drive Thor, paired with the Drive OS software and toolkits for perception and mapping. At CES 2025, Toyota (the world’s largest automaker) announced it will integrate Nvidia’s Drive AGX Orin “supercomputer” and Drive OS into its upcoming vehicles to enable advanced driver-assistance and automated driving features.
This was a major win for Nvidia, as Toyota has been collaborating with the company since 2017 on AI for self-driving and even uses Nvidia’s cloud GPUs to train its models. The Toyota deal underscores Nvidia’s “cloud-to-car” strategy: first supplying chips for AI training in the data centre, and now supplying chips and software for intelligence inside the car.
Other automakers are also lining up. Mercedes-Benz, China’s BYD, Volvo, Hyundai, Lucid Motors, and many electric vehicle startups have adopted Nvidia’s Orin-based platforms for their next-gen cars. In early 2025, General Motors struck a broad partnership with Nvidia to use its GPUs and AI software across passenger vehicles, robotaxis, and even factory automation.
GM will equip future cars with Nvidia’s “AI brain,” including Drive SoCs running the safety-certified Drive OS (based on Nvidia’s latest Blackwell GPU architecture), to enable hands-free driving and autonomy. Notably, GM is also leveraging Nvidia’s Omniverse 3D simulation platform to create virtual assembly lines and train its industrial robots, aiming to boost manufacturing efficiency with AI. This includes the company’s Drive AGX system-on-a-chip (SoC), similar to Tesla’s Full Self-Driving chip or Intel’s Mobileye EyeQ.
The SoC runs the “safety-certified” Drive OS operating system, built on the Blackwell GPU architecture, and capable of delivering 1,000 trillion operations per second (TOPS) of high-performance compute.
This partnership came on the heels of GM’s decision to wind down its Cruise robotaxi unit after safety incidents, signalling that GM now prefers to pivot toward consumer vehicles with advanced autonomy, and it’s tapping Nvidia to help make that happen.
Winning over these automotive giants could translate into huge business for Nvidia. The company projects its automotive division will reach a $5 billion annual run rate by FY2025, a fivefold increase from 2023. That’s still modest next to Nvidia’s booming data centre revenue, but the growth trajectory is clear. Nvidia’s automotive VP Ali Kani remarked that the car business is “still in its infancy,” contributing Nvidia chips to under 1% of cars on the road today, but he calls it a “trillion-dollar opportunity” long term.
Industry analysts have taken note: McKinsey estimates assisted and autonomous driving could be a $400 billion market by 2035. And after a few gloomy years when automakers dialled back self-driving investments (Ford and VW shuttered Argo AI in 2022, GM pulled back on Cruise in 2024), Jensen Huang’s confident CES showcase was seen as a “shot in the arm” for the sector.
“Nvidia has reversed that and just gave autonomous driving an absolute shot in the arm,” said one automotive consultant, noting that hearing a tech leader evangelise self-driving renewed investors’ interest in the space.
Crucially, Nvidia’s advantage is its full-stack approach. Few companies can provide the training-side infrastructure (massive AI supercomputers to train driving models) and the in-car chips to execute those models on the road.
Tesla, for instance, trains its Autopilot AI on Nvidia GPUs in the data centre, even though it builds custom chips for its cars.
This “cloud + edge” synergy makes Nvidia a natural partner for any firm aiming to deploy autonomous vehicles at scale. It’s no surprise Huang says Nvidia is “absolutely positioning [itself] as the leader for autonomous technologies, period.”
Indeed, from passenger cars and long-haul trucks to robotaxis, Nvidia’s silicon and software are increasingly becoming the standard toolkit for autonomy. As Huang put it at CES 2025, self-driving cars are no longer perpetually “coming” — “they’re already here,” citing the commercial progress of Waymo and Tesla as proof.
Building an ecosystem of robots
If self-driving cars are essentially “robots on wheels,” Nvidia’s ambitions don’t stop at transportation. The company is simultaneously assembling an expansive ecosystem for robotics and automation in other domains. Huang believes the world is on the cusp of an era of “physical AI,” intelligent machines performing tasks in the real world, and he wants Nvidia to provide the brains of those robots. At Computex, flanked by virtual humanoid figures, Huang proclaimed, “Robotics is here. Physical AI is here. This is not science fiction.”
So, what is Nvidia’s game plan for robotics? In short, to offer a full technology stack akin to its automotive approach. Simulation is one pillar: the company’s Omniverse platform creates rich virtual worlds where robots can be trained and tested safely. Built atop Omniverse is Isaac, described as a “gym” for robots, which lets developers put virtual robots through their paces to practice tasks or generate synthetic training data.
Then comes the edge hardware: Nvidia’s Jetson line of AI chips (with a forthcoming flagship called Jetson Thor) provides the onboard compute for robots to perceive and act in real time. Finally, tying it together are AI models and software frameworks that give robots their smarts.
In 2023, Nvidia unveiled Project “GR00T,” a “moonshot” effort to develop a foundation AI model for humanoid robots. In 2025, this bore fruit in the form of Isaac GR00T N1, billed as the world’s first open-source generalist model for robot intelligence.
GR00T N1 is essentially a robot “brain” that has been pretrained on vast data, not just text or images, but demonstrations of physical actions. It uses a dual-system architecture inspired by human cognition: a “slow-thinking” module that reasons and plans, and a “fast-thinking” module that executes reflexive actions.
This mimics psychologist Daniel Kahneman’s concept of thinking fast and slow. In practice, GR00T N1 can observe its environment (through sensors and cameras), interpret instructions, plan a sequence of actions, and then control a robot’s limbs to carry out complex tasks. Nvidia pretrained it on both real-world human motion data and millions of synthetic scenarios generated in simulation.
Importantly, GR00T N1 is customisable. Developers can fine-tune it with additional data so a robot learns specialised skills. Huang declared that “the age of generalist robotics is here” as he opened up GR00T N1 to the world’s robot makers.
Complementing GR00T is another key piece called Nvidia Cosmos. Unveiled at CES 2025, Cosmos is a family of foundational models focused on modelling the physical world.
Whereas language models ingest books and websites, Cosmos was trained on 20 million hours of video of humans and objects in motion. It generates highly realistic images, simulations, and 3D scenarios, for example, showing boxes falling off a shelf in a warehouse, which can be used to teach robots what to expect and how to respond in the real world.
“It’s not about generating creative content, but teaching the AI to understand the physical world,” Huang explained.
Companies are already using Cosmos: humanoid robot startups like Agility Robotics and Figure, and self-driving car developers like Waabi and Wayve, are leveraging it to accelerate their training and simulation. In essence, Cosmos gives robots common sense about physics and environments, while GR00T gives them the decision-making and motor skills, together aiming to dramatically lower the barrier to robotics development.
Nvidia’s Jensen Huang envisions a future where advanced AI chips and software power fleets of autonomous vehicles and humanoid robots, a vision already taking shape through partnerships with automakers, electronics manufacturers, and even entertainment companies.
Nvidia is backing up these platforms with real-world pilot projects to showcase what’s possible. One headline-grabbing example is its collaboration with Foxconn, the world’s largest electronics manufacturer.
In 2025, Nvidia and Foxconn announced plans for a “robotic factory” in Houston that will use humanoid robots to assemble Nvidia’s own next-gen AI servers.
This would be the first time Nvidia products are built with the help of humanoid machines, and one of the first such deployments in any electronics factory. Foxconn has been co-developing humanoid robots with Nvidia’s hardware and software, including one bipedal model and one wheeled model, and training them for tasks like picking up components, inserting cables, and performing assembly.
The goal is to have a small number of these robots operational by Q1 2026 when the Houston plant begins production of Nvidia’s “GB300” AI servers. If successful, this could herald a new era of AI-driven manufacturing.
Observers note it as a prestige project for both firms: Nvidia would solidify its position not only as a chip and server leader but as a platform provider for robotics, while Foxconn would demonstrate high-tech manufacturing innovation on American soil.
Nvidia’s robotics push goes beyond factories. The company is moving into service and entertainment robots, even magic-infused Disney creations.
In March 2025, it emerged that Nvidia, Disney Research, and Google DeepMind are teaming up on a project codenamed “Newton” to create a new generation of interactive robotic characters.
Newton is essentially an open-source physics engine that will help Disney’s robots learn complex movements with precision. Disney Imagineering’s vision is to bring more lifelike robots to its theme parks, for example, free-roaming droids like the Star Wars-inspired “BDX” robots that were previewed during Huang’s keynote.
“This collaboration will allow us to create a new generation of robotic characters that are more expressive and engaging than ever before,” said Disney Imagineering’s senior R&D VP.
In fact, Disney’s first batch of AI-powered BDX droids has already been play-tested on a cruise ship and will soon appear in Walt Disney World and other parks.
For Nvidia, this alliance is a testament to the scope of its robotics prowess, reaching into animatronics and arts, away from factory floors. It also highlights a familiar refrain: Nvidia is not making the robots or cars themselves; rather, it provides the enabler technology that allows industry leaders (be they Toyota, Foxconn, or Disney) to fulfil their AI-fueled dreams.
Towards a deep tech superpower
All these moves point to Nvidia’s evolution from a pure “Big Tech” company into something broader, a deep technology platform spanning hardware, software, and services for the AI-driven future.
“We stopped thinking of ourselves as a chip company long ago,” Jensen Huang recently remarked.
He prefers to describe Nvidia as an “AI infrastructure” and “computing platform” provider, one that now delivers cloud services, simulation software, and development toolkits in addition to silicon.
By weaving itself into the fabric of emerging industries like autonomous vehicles, robotics, and even defence (Nvidia is a partner in a European project led by Nokia to use drones and robots for critical infrastructure protection), Nvidia is staking a claim as a leader in “physical AI,” the application of artificial intelligence in the physical world.
Huang predicts that in the not-so-distant future, there will be “billions of robots, hundreds of millions of autonomous vehicles, and hundreds of thousands of robotic factories” all powered by Nvidia technology. It’s an audacious vision, but one the company is investing heavily to realise.
Financially, these sectors are still ramping up. Nvidia only recently began reporting its automotive and robotics revenues together, and they accounted for roughly $567 million last quarter (about 1% of total sales), albeit growing 72% year-on-year. The company’s present profitability still relies on data centre AI chips, but investors are keenly watching these nascent divisions.
For tech-savvy investors, Nvidia’s foray into self-driving cars and robotics represents the opening of new multi-billion (even trillion) dollar TAMs (Total Addressable Markets) over the coming decade. It pits Nvidia not against the usual FAANG companies, but against (or alongside) players in automotive, manufacturing, healthcare, and aerospace, industries hungry for AI solutions.
The company’s strategy of partnering with incumbents (rather than competing directly) could yield a sprawling customer base without massive capex on its part. As one set of analysts wrote, Nvidia’s strengths in robotics and digital twins could “scale into massive businesses themselves,” potentially driving decades of growth.
Naturally, challenges abound. Robotics, especially humanoid robots, remain harder than web or mobile tech; they involve complex mechanics, safety concerns, and huge data requirements to function reliably in unstructured environments. Sceptics point out that despite Nvidia’s impressive tools, fully autonomous robots are still in the early stages and demand significant R&D.
Even Huang acknowledges that growing the robotics market “isn’t just a matter of time” or inevitability. It will require continued advances in artificial intelligence algorithms, sensors, and perhaps most importantly, cost reduction, to see robots proliferate beyond pilot projects.
Nvidia’s strategy of open-sourcing models like GR00T N1 and collaborating widely is meant to accelerate this progress by seeding an ecosystem (much as OpenAI frameworks spurred the machine learning boom).
Huang believes that by reducing barriers and providing substantial computing power, Nvidia can do for robotics what it accomplished for AI software: make it accessible enough for an explosion of innovation.
Nvidia has gone from selling graphics chips to becoming the linchpin of modern AI, and now it’s charging into autonomous machines on our roads and in our factories. This pivot could fundamentally reshape Nvidia’s identity, no longer simply a supplier to “big tech” companies, but a superpower in deep tech, commanding influence in the next generation of industries built on AI, from smart cars to smart robots.
Huang’s remarks at a recent shareholder meeting put it best: “AI and robotics are the two largest opportunities, representing a multitrillion-dollar growth opportunity.”
If Nvidia succeeds, it won’t just be leading in AI computing; it will be everywhere that advanced computing meets the real world.
In an era when “everything that moves” is poised to become autonomous, Nvidia appears determined to supply the engines of that revolution, thereby laying the basis of its transformation as a force dominating deep tech innovation.