Vodafone tender for 100k+ sites puts Cohere in open RAN spotlight

The medieval Spanish city of Ciudad Real, just south of Madrid, seemed a mischievous choice of location when Vodafone took Cohere Technologies out of the lab and into the field. Run since 2018 by Ray Dolan, Cohere claimed its technology could turbocharge the performance of existing mobile networks – and, by implication, the value of the spectrum they use – through a mere software update. It sounded unrealistic, even quixotic, a word bequeathed to the English language by the great Spanish writer Miguel de Cervantes. A museum dedicated to his novel Don Quixote is among Ciudad Real's attractions.

Vodafone, though, did not think Cohere was tilting at windmills. Operators pay billions of dollars for their spectrum licenses and mobile network equipment. New-generation network products are expensive. And revenues are not growing much in saturated smartphone markets. "It's so important for us how much you can squeeze out of the spectrum," said Francisco (Paco) MartÍn Pignatelli, Vodafone's head of open RAN (or O-RAN). Even a 30% improvement in spectral efficiency would be good, he told Light Reading at last week's FYUZ event in Dublin.

But the field trials on test equipment in Ciudad Real showed much better results than that. "Most of the time, we saw a 50% improvement, and we also checked the mobility, the number of devices, and it was all easy to integrate," said Pignatelli. "For us, that was an important achievement on the journey to make it commercial right then."

Unrestrained in his praise for the San Jose-based Cohere, Pignatelli has become an eager advocate, name-checking the company several times at FYUZ while on stage. "We are seeing promising innovations and one of them is Cohere," he said during a panel that also featured Parallel Wireless and Wind River, other players active in the market.

Yet Cohere Technologies was founded about 13 years ago and has been led for the last six by Dolan, who made his name in 2006 with the $805 million sale of Flarion, a radio innovator he cofounded, to Qualcomm. If Cohere's technology is as valuable as Flarion's seemed to be, why is there so little evidence of it in commercial networks?

Unfortunately, what's good for Cohere and Vodafone is not necessarily good for the companies that today earn billions from the sale of RAN products. If it boosts the performance and prolongs the life of those products, Cohere's Universal Spectrum Multiplier (USM), the software that so impresses Pignatelli, could become a low-cost alternative to investments of the traditional kind.

Unlocking the gates

Today's RAN market is dominated by three big kit vendors. Together, Huawei, Ericsson and Nokia accounted for about 75% of all RAN product sales last year, according to data from Omdia, a Light Reading sister company. Cohere's USM software cannot percolate through deployed networks unless those vendors unlock the gates. Pignatelli understands why they would be concerned. "Probably you cannot expect some of those suppliers to be the first ones striving for this, because there is nothing to gain initially," he told Light Reading.

Cohere has two routes into the network, both of which demand willing partners. The first is through an xApp, a near-real time application hosted on a RAN Intelligent Controller (RIC), a network platform. Yet most of the industry attention and progress has recently been on a non-real time alternative called an rApp, more distanced from the basestation nodes and functionality.

"We see the different views in the industry about xApps and how feasible they are, because you have multiple applications touching the same component of the radio in real time," said Pignatelli. "Who controls, who polices, all that? That is the question. And some suppliers are taking a clear view on that – they see difficult implementations, even though the idea was good."

Comparatively few RICs supporting xApps as opposed to rApps have taken shape and Ericsson remains philosophically opposed to the entire xApps concept. "The functionality of xApps is provided by the RAN itself," the company recently told Light Reading by email. "At this point, we don't see benefits of xApps on top of what our RAN is doing."

Its stance fuels criticism Ericsson is paying only lip service to the concept of open RAN, whose proponents push for greater use of non-proprietary interfaces between different RAN components. Just last week, Ericsson also confirmed it would not support open RAN for massive MIMO, an advanced 5G technology, in some of its most cutting-edge products.

The non-xApp route for Cohere would involve direct integration with a RAN vendor's compute systems. "This implementation of Cohere's is so deep down into the radio that in some cases it could be potentially better to just integrate it natively in the software," said Pignatelli. The drawback here is the potential for Cohere's technology to be limited to one vendor.

"Nobody gains if it's only for one supplier," said Pignatelli. "You want this to be enabled across multiple suppliers, at least from our side. There's always an advantage if someone does it, but the point is to make it portable if you can. So doing it as an rApp or through an open interface should be the right way to do it." Whether USM could be engineered to work as an rApp is currently unclear.

Big RAN refresh

Still, these are potentially exciting times for Cohere as Vodafone prepares to award new RAN contracts across its African and European footprint. This refresh, likely to happen in phases over several years, was originally supposed to cover 170,000 basestations. But the number has inevitably fallen with Vodafone's divestment of assets in key European markets including Italy and Spain, Pignatelli confirmed. When it comes to open RAN, Vodafone's ambition is unchanged – to reach 30% of all European sites with the technology by 2030.

Deployment today is limited to a relatively small number of sites in the UK and Romania. Nevertheless, in both countries, Vodafone is replacing Huawei, a controversial Chinese vendor it must evict under government orders, with products from Samsung, NEC and various IT component and software suppliers. After success in the UK and Romania, Samsung has high hopes. It also appears to have a proven template for integration, using Intel for chips, Dell for servers, Wind River as a virtualization platform and NEC as an occasional partner for radio units (RUs). Other radios come from Samsung, which is also providing the all-important RAN software.

Rather like Ericsson, however, Samsung has previously sounded hostile toward the concept of xApps. "I think it is much tougher to meet performance expectations," said Alok Shah, the head of strategy, business development and marketing for the networks business of Samsung Electronics America, when he met Light Reading at Mobile World Congress early this year. Its own RIC did not support xApps at the time and it was keeping its own near-real-time functionality integrated with its RAN software.

This could all work out favorably, then, for Mavenir. The US open RAN developer has responded to Vodafone's tender, according to sources, and was known to be meeting with Pignatelli at FYUZ. It is also part-funded by Koch Investments Group, one of the same backers of Cohere, and boasts an xApp-supporting RIC that accommodates Cohere's USM software, the companies announced last year.

But Mavenir's RIC could not simply be linked to Ericsson's baseband units. For a start, Ericsson's antipathy toward xApps implies it is not supporting E2, the critical open RAN interface that would expose its nodes to a third party. Even if it did, Mavenir and Ericsson would probably need to collaborate on integration, especially for equipment already in the field. The advantage of a Cohere xApp that works only on Mavenir's products is limited by Mavenir's negligible presence in the RAN market, including Vodafone's European footprint.

Pignatelli's enthusiasm for Cohere clearly lines it up for a potentially big role in the site tender. But it also seems aimed at putting Vodafone's traditional vendors under some pressure to accommodate its demands. "It depends on the solution they propose to us and that's a work in progress," said Pignatelli when asked if the open RAN part of the tender is open to Ericsson and Nokia. "We didn't do O-RAN to exclude them but to transform the way the radio is done, so it's great they are changing. They are welcome to the party."

Update: The original version of this story gave $600 million as the amount Qualcomm paid for Flarion. The correct amount was $805 million. This has now been changed.