Cohere, the capacity booster, confident of RAN deal this year

Ray Dolan, the CEO of Cohere Technologies, once cited the Whispering Gallery effect at St Paul's Cathedral to explain beamforming. Visitors who enter the London landmark can literally talk to the wall and be heard by someone on the opposite side of its vast dome as if they were invisibly close. Beamforming, in a more complicated guise, is Cohere's specialty.

For a while, the company seemed destined to remain a far less popular attraction than Sir Christopher Wren's famous building. More than two years since it raised $46 million in funds from multiple investors, Cohere has yet to land a commercial deal. But the market this year looks as ripe for a Cohere breakthrough as it ever has.

The absence of a commercial deal so far has not, it seems, been for lack of telco interest in what Cohere can do. During recent field trials in the Spanish city of Ciudad Real, Vodafone said Cohere's Universal Spectrum Multiplier (USM) software boosted capacity on a 5G network by 50%. Cohere has worked in test environments with Telstra in Australia and with Bell Canada. Deutsche Telekom, Europe's biggest service provider, has also fiddled with Cohere's tech in its labs and said positive things about the experience.

So, why have deals not materialized some four years since those Deutsche Telekom trials? The blame lies partly with industry inertia, telecom's habit of gushing about radical change and then carrying on as usual. On open radio access network (RAN) technology – an attempt to break a perceived oligopoly in the mobile infrastructure market – the best the industry can advertise so far is a $14 billion contract between AT&T and Ericsson. That baffles many observers because Ericsson is supposedly one of the oligopolists. Until this time last year, the 150-year-old Swedish company was also seen as an opponent of open RAN.

Open or closed, Ericsson has been no friend to Cohere, and that is the bigger problem for Cohere's business. Its USM software cannot be installed on a user handset like WhatsApp can. It needs either to be directly integrated with a RAN vendor's technology or deployed as an app in the network. That network apps market remains immature, lacks standards and is already in flux. And the big RAN vendors have shown no obvious desire to work with Cohere.

Cohere on the outside

As things stand, none of the giants is doing especially well. Huawei can be immediately dismissed because the same geopolitical forces at work in open RAN are squeezing the Chinese vendor out of the Western markets that Cohere would naturally target. Ericsson and Nokia, meanwhile, have been hobbled by spending cuts among telcos. Revenue at Ericsson's mobile networks business fell 23% for the final quarter of 2023. Nokia's equivalent unit reported a 17% drop. Both vendors are laying off thousands of employees to protect margins.

What they really want is for operators to start spending money on their latest massive MIMO radios. Massive MIMO is the Formula 1 of the RAN, replete with dozens of antennas for a 5G performance boost. It also comes with beamforming technologies developed by Ericsson and Nokia. The routine message from Börje Ekholm, Ericsson's boss, is that operators in numerous Western markets have underinvested in 5G compared with their Asian peers. Unless they increase investment, those countries will suffer as mobile networks fill up with gigabytes and become a digital version of gridlocked LA.

The last thing those vendors need is Cohere. Its pitch has been largely about the boost USM could bring amid less advanced radios already in the field. If those can effectively be supercharged by Cohere's software, telcos would see even less reason to invest in new radio capacity. The drought for Ericsson and Nokia might last even longer.

But spend or suffer is not a very upbeat message for telcos that haven't seen any meaningful sales growth since octogenarians and toddlers acquired smartphones. And telcos have shown a willingness in recent years to lean harder on their suppliers. Open RAN is evidence of that, and both Ericsson and Nokia now profess support for the concept. Having already beaten vendors into complying with open RAN specifications, telcos like Vodafone could exert similar pressure on them to admit Cohere.

Vodafone's big tender

Dolan struck an optimistic note at Mobile World Congress, where he met with Light Reading. "Until the pain of same is greater than the pain of change, you don't change," he said. "We've arrived at a point where this industry has enough pain in it to look at opportunities. I am convinced we will be fully integrated with at least one of the RAN players this year."

He is, of course, not saying anything about conversations with specific vendors. But the biggest opportunity on the horizon is probably the 170,000-site tender issued this year by Vodafone, a Cohere advocate. The operator's publicly announced goal is to have open RAN installed across 30% of its network footprint by 2030.

That naturally brings opportunities for new entrants more receptive to Cohere. Even the RAN incumbents may put up less resistance to it when they are not defending basestations but on the attack. And Huawei, increasingly unpopular in Europe, still has a major presence in the Vodafone network. In a report published in 2020, Strand Consult, a Danish analyst firm, reckoned it accounted for 62% of Vodafone's entire European footprint.

Cohere has already announced a tie-up with Mavenir, a US developer of RAN software previously involved in Vodafone trials. "That would be like a rip and replace," said Dolan, citing Mavenir as an example of Cohere's "greenfield" route into the RAN. Besides Mavenir, though, Cohere has not publicly named any baseband software partners. And Samsung may be at the front of the queue in the 170,000-site tender.

Currently replacing Huawei at Vodafone sites in both the UK and Romania, the South Korean electronics giant is held in high regard by Santiago Tenorio, the Vodafone executive in charge of network strategy. Unlike smaller companies, it has a complete portfolio of RAN products. With a RAN market share outside China that has fluctuated between 9% and 11% in the last three years, according to Omdia (a Light Reading sister company), Samsung is very much the challenger to Ericsson and Nokia, with less existing business to protect and more reason to be disruptive.

Death of the RIC?

Besides developing software that any baseband vendor can directly integrate into its products, Cohere also provides its USM in xApp form. The x, in this case, distinguishes it as "near real-time" from the rApp category of "non-real-time" software. In theory, an operator should be able to use something called the RAN Intelligent Controller (RIC) to select a variety of network-specific apps such as Cohere's. Juniper and VMware, an investor in Cohere, both offer RICs that host Cohere, effectively making them app stores for the network.

In a properly open network, then, a telco could access an xApp, wherever it is hosted, and run it in conjunction with any supplier's equipment and software. Right? This certainly fits with Dolan's vision of how the market will ultimately evolve. "You take the local USM that sits under the tower, and you move it to the cloud in the form of an xApp," he said, while ruminating on the future.

Yet there are doubts the RIC concept will survive. "The RIC is not as universally supported as it maybe was initially as part of open RAN," said Dolan. "There are operators still strongly in favor of deploying a RIC, and there are operators that are no longer in favor or never were."

The dispute is not about whether apps for the network are desirable. That much is clear, insists Dolan. Rather, it is about the need for the RIC as the place where these apps are hosted. "The notion of an app store still resonates, whether it is hosted in the RIC or in the cloud or starts locally at the basestation," said Dolan.

On that note, Microsoft is now positioning a telco cloud platform called Janus as a future commercial store for xApps, said Dolan. "At some point in time the RAN is fully virtualized and runs applications from the cloud," he said. Cohere has long insisted that its USM xApp has the "latency budget" to reside in the cloud, meaning there would be no performance drawbacks of hosting it there.

Even so, this Janus development may exacerbate the short-term anxiety about a proliferation of app stores, which threatens to be a commercial turn-off for developers. "There are a lot of players," said Alok Shah, the head of strategy, business development and marketing for the networks business of Samsung Electronics America. "The operators will help to drive that number down, ultimately."

Standardization also appears to have lagged in the xApps area. "I think it is much tougher to meet performance expectations," said Shah. Samsung has instead thrown its efforts into rApps. Its own RIC does not support xApps, and it has decided to keep any of its own "near real-time" functionality integrated with its RAN software. But all this may reflect concern about third-party xApps as a competitive threat to basestation products sold at high prices.

So far, RIC relations with Juniper and VMware have not produced commercial business for Cohere. Given the uncertainties associated with the RIC, and the relative immaturity of open RAN, Cohere's best bet would still seem to be a potential direct integration with one or more RAN vendors.

Can it afford to wait? Cohere has raised a total of $81.5 million, according to Crunchbase, including the $46 million it pocketed in 2024. "We have two years of capital on the balance sheet," said Dolan. Yet while he insists that is enough to carry on as before, he anticipates another possible funding round this year to help Cohere "scale faster." He will hope the Vodafones of this world are leaning on their suppliers with zeal.