- The world is entering a new era of AI that’s focused on the physical realm – robots, factories and infrastructure
- T-Mobile’s CTO says physical AI requires a new type of token and telecom operators are in a prime position to distribute them
- Not coincidentally, T-Mobile stands to be at the forefront of physical AI due to its network advancements
Physical AI, which is expected to be a hot topic at next week’s Mobile World Congress 2026 in Barcelona, presents a unique opportunity for telecom operators to showcase their assets, especially as 6G approaches.
That’s according to T-Mobile President of Technology and Chief Technology Officer John Saw, who says this new era of AI is going to require a new kind of token: kinetic. He first publicly spoke about kinetic tokens briefly during T-Mobile’s Capital Markets Day in early February and elaborated on them in a blog post last week.
Fierce caught up with him for an exclusive deep dive into exactly what the heck he’s talking about with these new tokens and why they’re relevant to the wireless industry as a whole.
To be sure, tokens are nothing new in the world of AI. They’re basic compute units. Nvidia talks about them extensively as the language and currency of AI.
But in the future, the nature of tokens is going to look different than what we might call informational tokens, which are basically units of data that describe, summarize or predict, according to Saw.
Kinetic tokens don’t just represent information but take some kind of physical action, like movement, adaptation or coordination with the real world. Think drones, robots and autonomous cars – and they need to act in milliseconds.
Kinetic tokens are not just a T-Mobile thing, either, although it’s positioning itself to be in a good spot to take advantage of them. (More on that later.) It’s an industry-wide phenomenon.
“When tokens move – when things move with AI – I think it gives telecom operators a license to play in a big opportunity with physical AI,” Saw said.
Carriers vs. hyperscalers
This seems like as good a time as any to delve into the relationship between hyperscalers and telecom operators when it comes to the future of AI. Who will dominate?
“I think it takes a whole village. We’re not going to replace the data centers,” he said. “If you really want to support physical AI, data centers need to be supplemented by the edge. And this is what I’m talking about – edge AI for inferencing – because it is at the edge where the action happens.”
Nvidia President and CEO Jensen Huang said as much during a video clip he provided for T-Mobile’s Capital Markets Day event, where he talked about how physical AI will be distributed at the edge, and the wireless radio network will serve as the distribution hub.
And mobile operators happen to be very good at the time-space coherency that kinetic tokens require.
“This is what we do every day for mobile networks,” Saw said. “When I make a phone call or when I send a message, everything has to be perfectly synchronized – the entire system – to deliver that message to you. Mobile operators have been doing this for a living for the longest time. This is one thing that clouds don't do well.”
Roy Chua, founder of AvidThink, said Saw’s blog post and the subsequent attention it got on LinkedIn were the first references he’s seen of “kinetic tokens” as a term.
While the term may be new, the concept isn’t. “The wider industry recognizes that as AI moves from generating content to orchestrating physical actions – robots, autonomous vehicles, industry automation – the data shifts from being informational to operational, carrying intent that triggers real-world movement. And this is tied to physical AI, which Jensen and Nvidia have been increasingly promoting over the last two years,” Chua told Fierce.
Wireless carriers inherently know where subscribers and connected devices are at a granularity and scale that cloud providers can’t easily match without leveraging an app asking for access to a user’s mobile location. “When edge-driven workloads like a remotely managed drone, or robots need real-time coordination, a network that can recognize the workloads and provide the necessary QoS allows the carriers to add value,” he said.
6G as the connective tissue
6G is being designed as the first AI-native “G” in wireless. It’s still in relatively early stages of development with standards bodies and organizations like 3GPP, where specifications get studied and hammered out over the course of years before the next generation of wireless technology is deployed.
Of course, T-Mobile believes it’s particularly well-prepared for 6G. It was the first to launch a nationwide 5G Standalone (SA) core in 2020, as well as the first with nationwide 5G Advanced, which it announced last April. Both are seen as foundational for 6G.
And T-Mobile is a founding member of the AI-RAN Alliance.
“We’re not waiting for 6G,” Saw said. “We’re already using AI for RAN today in our 5G Advanced network. The way we design our network is using what we call customer defined coverage, where we use AI to tell us where’s the best [place] to put the next cell site. That’s based on our customers’ user experience.”
Chua said T-Mobile has a legitimate head start here over rivals AT&T and Verizon.
“Kinetic tokens is a good framing move by Saw and T-Mobile that positions carriers as a key player in the physical AI wave rather than on its periphery. And physical AI is the next hot thing which Nvidia is backing,” he noted.
“We’ll be watching how T-Mobile translates this into concrete use cases and pricing models enterprises will actually pay for,” he said.
Conflicting ways to handle AI
Speaking of AI, Fierce last week reported how Nokia and Ericsson fundamentally disagree on how and where to deploy AI in the mobile network. T-Mobile works with both Nokia and Ericsson, as well as Nvidia, which last year announced a $1 billion investment in Nokia.
So, how is T-Mobile handling all this?
“From a T-Mobile perspective, all we really wanted was a future network that processes bits and tokens at the same time with the best cost and best efficiency. We shall see,” Saw said. “We are not going to be prescriptive in terms of ‘here’s the exact hardware and here’s the exact chips you need to use.’ We have never done that. That’s why we pick world-class vendors. It’s their decision. I give them the requirements.”
He also reflected on why T-Mobile got involved in this in the first place.
“We wanted to bring in AI RAN because we wanted to challenge the industry by rethinking how we do telco computation. Before GPUs became popular, they were only using ASICs and CPU, so we wanted to introduce what Nvidia has, and those guys are the world champions in GPU processing.”
Ericsson is working with Nvidia as well. “How they use it is something that I think both sides are still trying to figure out,” he said.
Read next: What Fierce Network is watching for at MWC 2026: AI, 5G and network transformation