Clad in his ubiquitous leather jacket, CEO Jensen Huang strode onto the stage in front of a large screen displaying a row of humanoid robots straight out of a Philip K. Dick novel.
After a beat, something cuter emerged from back stage: two knee-high robots reminiscent of Star Wars’ R2-D2 waddled out, emitting beeps and boops.
This is the business Huang has been talking up for much of 2024, what he has framed as the next wave of artificial intelligence (AI). Robots will bring AI that “understands the laws of physics,” and how to interpret the world, he told Jim Cramer earlier this year.
All factories will be robotic, and they’ll be building products that are robotic. “Billions” of humanoid robots will be shipped in the coming years, Huang has said.
Having captured one of the most lucrative markets in recent times for chips that can train and run Generative AI systems, Huang has been talking up three other areas he’s now eyeing for growth: autonomous vehicles, quantum computing and, the most plausible contender when it comes to available technology, robots.
You could argue Huang doesn’t need to be thinking about new markets, given the staggering profitability of his AI chip business. Nvidia earned more than $16 billion in net income for the second quarter, up nearly 170% from the previous year.
Much of the $56 billion that Microsoft, Alphabet, Amazon.com and Meta Platforms are projected to have spent in the third quarter on data centres and other resources for building AI were earmarked for Nvidia. Such a heavily concentrated business is risky.
If Nvidia’s handful of customers stop buying AI chips, or start developing their own, or if AI computing demands change in some way, Nvidia would suddenly look vulnerable. And as any veteran of the cyclical semiconductor business knows, the industry is typified by booms and busts.
AI chip demand might seem insatiable now, but it won’t be forever.
One rumour among chip firms has been that Nvidia will build its own robots from the ground up. The idea is that it could take advantage of its powerful chips, known as graphics processing units (GPUs), and the tools it already sells to developers to build its own robots. More than 100 robotics firms are using Nvidia’s Isaac suite of software tools and AI models to test robot applications.
Building the chips, the software and hardware for robots would mean the company controls the entire technology ‘stack’ behind such machines. How might that benefit Nvidia?
In theory, it could develop all those components in tandem so that they work seamlessly with one another, allowing the company to build the best robots on the market in much the same way Apple does with phones.
In practice, building robots would be a terrible idea. Antitrust regulators would probably be all over Nvidia’s efforts, given its dominance in AI chips.
The company also lacks the supply-chain and manufacturing expertise to build robotic hardware. And jumping into the business would gnaw deeply into its fat profit margins, which swelled to 55.3% in the third quarter.
“Nvidia is more focused on creating new markets for their chips than building robots themselves,” says David Reger, CEO of Germany’s NEURA Robotics, who added that partnering with large manufacturers made more sense.
Huang has described himself as a “market maker, not a market taker,” so the idea of fighting other robotics firms over a nascent market might not even appeal to him on a personal level.
But framing yourself as a platform hub for new-fangled technology comes with challenges too. Arm Holdings, the British chip designer whose instruction sets are found in most smartphones today, positioned itself in the mid-2010s as the low-powered, beating heart of the internet of things revolution.
Instead, it struggled over several years to sell the necessary software to manufacturers like Robert Bosch to make good on its vision. It has since pivoted to framing itself as central to the next phase in the AI boom.
To its credit, Nvidia has entered new markets before with astounding success. Huang transformed gaming graphics cards into the bedrock of the AI revolution by becoming everyone’s essential partner, not their competitor.
For robotics, the playbook should probably be the same: Don’t build the bodies, but build the brains and tools that everyone else’s robots can’t live without.
The profits from that endeavour probably won’t match today’s AI gold rush, but when that boom cools, Nvidia may be glad it plugged itself into robots too. ©bloomberg
#Nvidias #R2D2like #robot #focus #developing #brain #body
