Room for another one? SynaXG outlines Layer One and AI-RAN play

SynaXG believes operators are interested in Open RAN solutions built on better and more efficient systems software.

As the main vendors scoop up the vast majority of RAN business, including “Open RAN”, is  there a business opportunity developing software for specific layers of the disaggregated RAN, and in providing RAN software within the AI-RAN?

SynaXG is a Singapore-based company that develops L1 RAN software. That’s the part of the RAN baseband software stack that encompasses the compute-heavy physical layer and scheduling workloads. It says it is working with chip partners and RAN infrastructure vendors to create deployable solutions. It is also committed to the AI-RAN vision, and recently joined the AI-RAN Alliance. (It has also recently opened an R&D office in Malaysia to further explore AI-RAN.)

In a disaggregated RAN the Control Unit (CU) includes the RRC (radio resource control) and PDC (packet data convergence) protocols, while the Digital Unit (DU) includes the RLC (radio link control) and MAC (multiple access control) protocols along with the physical layer.  The lower you go in the protocol stack, the higher the demand on processing – Layer 1 & 2 combined comprise 90% of processing demands. 5G bandwidths together with massive MIMO technology exponentially increase the processing demand on L1 and beamforming.

For the main RAN vendors, the development and integration of Layer 1 RAN software with their own chosen chip platforms is where they derive much of their advantage.

Mantosh Malhotra, Chief Business Officer at SynaXG says that SynaXG is working to develop an alternative path for rival RAN vendors, with a “carrier grade “ L1 software offer. He describes the L1 software as the “critical component” of a RAN product – and says that although there are a few other companies working on L1 software, SynaXG’s is “battle-hardened” and deployed in the field. The company is deployed in a live network in Japan, Malhotra said.

SynaXG’s business model is to enable infrastructure partners outside of the big five RAN vendors with its software. Although he can’t name names, that puts SynaXG in play with the likes of Fujitsu, NEC, Rakuten Symphony, Mavenir, Kyocera and Tejas.

To support its software ambitions, the company has also developed a O-RAN vDU PHY accelerator card. At MWC 2024, the company demonstrated a vDU card solution with Ampere Computing, integrating with Wind River’s distribution of Starling X, an open source CaaS project. There was also a demo on the Vodafone booth where the infrastructure partner was Fujitsu. 

SynaXG is working on integrations with a variety of chip companies, including Marvell, NXP and AMD, with a solution for Arm-based platforms. To reduce time to market the company has developed an in-line accelerator card based on those chips, with L1 software integrated on cards which can be installed in a server to complete the baseband unit. Malhotra describes that as doing the “heavy lifting” for the infrastructure players. It also produces some appliance-based baseband units, which acts as a complete baseband.

“We’re predominantly a software company that has some hardware supporting those software ambitions. We are probably not the right candidate to support the hardware and deployment needs for large operators: their requirements from hardware partners and infrastructure deployment patterns are quite large and that’s not our core business.” 

Although major operators have been slow to give deals to non-traditional RAN vendors, Malhotra does believe there is an appetite for change.

“In the US we’ve seen announcements. Europe is keen to see multi-vendor  core-baseband-radio – and we’re hopeful it plays out like that. Asia and Japan seems to be more open. We are hopeful and our objective is enable those technology partners.”

One aspect that Malhotra does see as ripe for change is a desire for more optionality at the chip layer, where Intel has underpinned may Open RAN efforts.

“I feel that infrastructure players and operators are looking at alternative solutions. Nobody wants to be pegged to one supplier. Arm is pushing for Arm-based solutions with carrier partners and we work closely with Arm. Compared to this time three years ago a lot of operators are open to Arm-based platforms so that puts us in a  reasonable spot. From a technology advantage we have showcased that our solution is very low power compared to some other products available.”

AI-RAN

With the company joining the AI-RAN Alliance towards the end of 2024, Malhotra is keen to point out that leveraging a compute platform for both RAN software functions and AI workloads is not a new endeavour for SynaXG.

“In fact our AI & RAN platform is mature and deployed. We have a live demo in our Singapore office that was working before the advent of Grace Hopper, so that shows we have been working with Nvidia for a long time.”

For its AI RAN platform, SynaXG has built its solution on top of the Nvidia GPU platform. 

“We don’t do any hardware. And we use their SDK to take their Aerial L1 software and build upon that. Their software is the baseline, we take and improve and add features that are required by specific mobile operators.” 

“It’s the Nvidia chip, card and Aerial L1 software which we take and add more features and commercialise for our partners,” Malhotra said. 

That contrasts to SynaXG’s O-RAN solutions where it has written the L1 software from scratch. 

“We do believe that we will be able to contribute in the AI-RAN community. Obviously, there are some big and influential names there alongside us, but it remains to be seen how it plays out for us.”