T-Mobile, AT&T, and Verizon are all building AI-RAN. They cannot even agree on whether the radio needs a GPU.

By Brian Newman, Contributing Editor

On September 18, reporting surfaced on how differently the three U.S. carriers are approaching AI in the radio access network. T-Mobile is the only one of the three publicly committed to Nokia as its primary RAN vendor, working with Nvidia GPUs, and targeting 50 percent spectral efficiency gains by the end of 2027 and 100 percent by 2028, with early trials already showing 30 percent plus improvements.

Verizon's network chief Yago Tenorio took the opposite position in public: "Do we need to put them in the radio? Absolutely not." Verizon's view is that CPUs already handle current RAN workloads fine, and GPUs earn their keep in applications, not inside the base station.

AT&T, running primarily on Ericsson, is staying in the middle. Rob Soni put it plainly: "We won't just do it for the sake of doing it." AT&T will adopt GPU-based RAN when the business case proves out, not before.

I spent many years watching three carriers make three different bets on the same underlying technology shift, from 3G to VoLTE to Open RAN. The pattern rarely changes. One operator moves first and either sets the standard everyone else adopts, or absorbs the cost of a path nobody follows.

The most likely outcome is that T-Mobile's early spectral efficiency numbers keep improving through 2027, and either Verizon and AT&T follow with their own GPU trials, or T-Mobile's gains plateau against a workload that never needed the extra silicon in the first place. However, if Verizon is right that CPUs are enough, T-Mobile will have spent two years and a vendor partnership chasing a specification advantage that shows up nowhere a customer can feel it.

Where does GPU-based RAN processing actually change what a subscriber experiences, and where is it a spectral efficiency number that looks good in a press release and nowhere else?

 

Brian C. Newman is a telecom and AI strategy consultant, course creator, and former Verizon technology leader with more than 30 years of experience across wireless networks, 5G, network operations, infrastructure modernization, and emerging technologies. He helps organizations understand how AI, connectivity, edge computing, and digital infrastructure are reshaping business operations, real estate, public safety, and customer experience.

Previous
Previous

How Should CRE and Telecom Leaders View Regulatory Uncertainty for AI?

Next
Next

Ericsson Puts Forth its Vision for the 6G Core