Samsung looks to $20B war chest to beat open RAN rivalsSamsung looks to $20B war chest to beat open RAN rivals

Mobile network product development is not for the skinflint or penurious. Last year, Ericsson and Nokia spent a combined $9.2 billion (at today's exchange rate) on research and development (R&D), about 19% of their collective revenues. AT&T's Rob Soni, a big fan of Ericsson, previously reckoned as much as $5 billion of that went on the radio access network (RAN), a figure that also equals roughly 19% of their sales in that market. Competing against them on portfolio breadth and depth for a global audience would therefore require an R&D commitment by a lone player of around $2.5 billion. Outside China, perhaps no one besides Samsung looks capable.

Just to cover its annual R&D costs, a vendor spending $2.5 billion would need to capture a 7% share of a RAN market that's expected by Omdia, a Light Reading sister company, to generate $35 billion in total sales this year. Instead, the vendor could always specialize, pumping a smaller amount into one area. With open RAN's new interfaces, it could link to other vendors for the parts it does not provide. But specializing has gone out of fashion, partly because big telcos have shown they still want to buy most of their RAN products from one or two suppliers.

This briefly summarizes the commercial problem for most RAN challengers. Data from Omdia shows the world's seven largest RAN vendors last year served all but 2.7% of the market, leaving that 2.7% divided among all other players. Mavenir, one of the most prominent, would need more than half of that just to cover the $500 million it raised from Koch Industries to invest in R&D for radios several years ago.

'Massive R&D spend'

Of course, the $500 million might not have been spent in a single year. But a smaller annual commitment would make competing against Ericsson and Nokia even harder. And the Nordic kit makers spend the same amount year after year after year. Unsurprisingly, ratings agencies have recently warned that Mavenir is at risk of default or restructuring. "Despite its cost-saving initiatives, Mavenir must maintain high R&D spending to avoid harming its competitive position," said S&P in an October note.

The alternative to having a dependable stream of revenues to reinvest partially in R&D is having a giant backer unworried about short-term losses. Mavenir might gain this if a rumored Saudi investment of $1 billion ultimately materializes. But Samsung Networks can go one better. The South Korean vendor of 4G and 5G products – which in 2023 served 6.1% of the RAN market, according to Omdia – belongs to a group that last year invested about $20 billion in R&D. And spending by other parts of Samsung directly benefits networks, or so the company claims.

"Each business unit does have its own dedicated R&D, but we also have a fairly massive R&D spend that kind of crosses the Samsung Electronics families," said Alok Shah, the vice president of strategy, business development and marketing for Samsung Electronics America. These straddlers include Samsung Advanced Institute of Technology and Samsung Research, a dedicated R&D unit. "The benefits networks see tend to be pretty substantial when it comes to the umbrella spend," said Shah.

While Samsung Networks does not break out details of its profitability or R&D expenditure, it generated sales of about $2.64 billion in 2023 and is on course for a substantial decline this year, having made $1.4 billion over the first nine months. But with its parent and affiliates, it clearly has a rare ability to fund technology development and remain competitive while recording losses.

Silicon expertise

The current priorities include silicon. On the baseband side, it has teamed up with Marvell Technology, a US chipmaker, to develop a semiconductor for its purpose-built 5G range. Samsung is also collaborating with Marvell on the beamforming chips used in massive MIMO radios. But while some other 5G vendors look heavily reliant on the "merchant silicon" sold by third parties, Samsung also claims to have in-house options for modems, radiofrequency integrated circuits and digital front end chips used in 5G radios.

"The approach is generally to build our own silicon either where we feel like we can add value over the merchant silicon that's out there or where we feel there isn't a merchant silicon solution that works well," said Shah.

Those various activities mean Samsung is also not overly reliant on Intel, its main silicon partner for virtual RAN deployments based on general-purpose compute. Even so, while Samsung has engaged with Intel competitors including AMD, the 40,000 virtual RAN sites it currently supports are just about all based on Intel's x86 platform. And after a share price slump and decline in sales, the recent resignation of Pat Gelsinger from the CEO job has sparked renewed debate about Intel's survival in the RAN market.

Much like Ericsson, Samsung looks determined to ensure most of its virtual RAN software can be transferred to another platform with either no or minimal changes. Experts say this is difficult to accomplish when more demanding RAN functions are hosted on custom accelerators. "From our perspective, portability is important," said Shah. "Accelerators have APIs for the software and those APIs really do a lot of work to make sure you can establish some level of portability."

Of course, a possible danger of supporting numerous options in the portfolio is that Samsung ends up with multiple development tracks for purpose-built RAN, virtual RAN and the various permutations. But the strategy, Shah insists, is to maintain just one software "trunk," activating and deactivating features as and when those are needed by telco customers. "There are a lot of areas where we're providing capabilities to an operator, or to a set of operators, that maybe aren't needed by the rest of our operator customers," he said.

For telcos that have lost access to Chinese vendors amid geopolitical ructions, Samsung's continued involvement in the RAN probably looks more critical than ever. If Omdia turns out to be right about the market this year, annual sales will have fallen by nearly $10 billion in just two years. Mavenir is seemingly in poor financial health and other small and mid-size vendors have also struggled.

"We meet the Chinese competition, we meet Samsung, we meet Nokia," said Fredrik Jejdling, the head of Ericsson's networks business area, during a previous interview, as he reeled off the names of competitors he sees in worldwide tenders. It takes "a certain scale" to build critical infrastructure, he said. That could be interpreted as a dig at smaller rivals, and there seem to be few opportunities for potential disruptors. But no telco is calling for a duopoly.