- Building radios for cellular networks is not a new thing for 1Finity, a Fujitsu subsidiary
- But it needed a way into the U.S. market, and open RAN provided the fuel
- With Boost Mobile in the rearview, it’s now focused on the next big thing: AI RAN
Building and selling cellular radios is a tough business. Take it from 1Finity, a Fujitsu company.
For years, Fujitsu worked to expand its customer base beyond Japanese operators in its home country. Several years ago, it saw open RAN as the way into North America. It quickly got behind the open RAN movement and eventually won a contract with Dish Network to supply radios for the Boost Mobile 5G network.
In total, Fujitsu, now 1Finity – the mobile networks and photonics business was spun out a year ago – supplied about 100,000 radios for Dish.
Boost’s 5G open RAN network is one of the biggest innovations that 1Finity’s Femi Adeyemi has seen in over 30 years in the RAN business.
“It’s a cloud-native network,” he told Fierce at 1Finity’s North American campus in Richardson, Texas. “Nobody has that today, so we were happy to be a part of that.”
Sadly, those radios are now sitting dormant on towers across the country as Boost Mobile traffic is now running on AT&T’s RAN, and Dish parent EchoStar is decommissioning its 5G network. That came after the Federal Communications Commission (FCC) last year opened an inquiry into Dish’s use of its spectrum, an investigation that ended after EchoStar agreed to sell spectrum to AT&T and SpaceX.
The big question now is who’s going to remove all the wireless gear sitting on those towers and other infrastructure?
It’s unknown whether the radios will ever be re-used or recycled, but the intellectual property (IP) behind them will live on for years to come, said Adeyemi, who serves as head of Strategic Product Marketing at 1Finity.
“As you look at 6G, the next generation of networks, they have to be cloud-native based,” he said. “We’ll be seeing a lot more of that approach.”
1Finiti’s open RAN biz
Beyond Boost, 1Finiti has successfully integrated its open RAN-based radios for several mobile network operators (MNOs), including with Ericsson for AT&T’s network and with Nokia for Deutsche Telekom in Germany. More recently, Rakuten announced a major collaboration to deploy 1Finity massive MIMO open RAN radios in Japan.
“We actually believe in the value of open, so we don't want to force our entire solution on MNOs. If you want the radios, excellent. If you are only interested in the DUs [distributed units] on the software side of the house, that’s good,” he said.
Obviously, radios play a key role in connecting consumers to mobile networks, but they’re one of the toughest things to build. Others have tried and failed.
“The reason why you are not seeing many [radio] suppliers is because this is not a walk in the park. It's really not. Others have tried it,” Adeyemi said. “Reliability is actually very, very important. It's not an easy thing to build.”
Open RAN is also not a walk in the park. From the start, open RAN has seen its share of detractors. In the early days, people were skeptical that Ericsson and Nokia would ever support open RAN specifications. Eventually, they did, then people questioned if they were really open.
Adeyemi stressed that open RAN is real. “Open RAN works. It’s commercial. It’s ongoing,” he said.
Since 1Finity works with both Ericsson and Nokia, Fierce asked how their styles differ and if it’s easier to work with one more than the other.
Adeyemi chose the diplomatic route. “I will only speak for 1Finity,” he said. “We are flexible to work with anyone and everybody, as long as you have an open interface. The open interface is what dictates how we work with you.”
Next big thing: AI RAN
Now that 1Finiti has established its place in the open RAN market, its next mission is focused on virtual RAN (vRAN) and AI. “That’s now our focus – the AI RAN solution into the North American market, from the stack to how we integrate that with GPUs or CPUs to support the AI RAN solutions,” he said.
Does that mean GPUs in base stations? Maybe. It all depends on what the MNO wants.
“What you want to run is your AI workloads. You can do that on GPUs. You can also do that on CPUs. What is prevailing in the industry is a dichotomy. When we talk about GPUs and AI RAN, the next connection is Nvidia. But there are others that are providing compute, so we don't want to lock ourselves into either it's Nvidia or AMD or someone else," he said.
“We just want to run AI workloads. If I'm able to do that successfully in GPUs, fine. Our solutions work on GPUs. If we want to do that on CPUs, we can do that on CPU as well,” he concluded.