Intel Bets on Outmuscling Arm in the RAN

The US chipmaker reckons the RAN industry is too small to keep pace with design changes that will benefit its general-purpose processors, but not everyone is convinced.

Iain Morris, International Editor

April 5, 2023

2 Min Read
Intel logo
Getty Images

Perform surgery on Nokia's latest radio equipment and you would quickly stumble across transistors based on relatively new 5-nanometer process nodes. They were probably made at a foundry in Taiwan, the island source of most advanced components. Go back three years and you would not have encountered them in the telecom sector outside the most expensive smartphones. Their introduction into radio access network (RAN) equipment shows how fast kit makers have moved to exploit state-of-the-art silicon.

When it comes to process nodes, the smaller the measurement, the better the chip. Shrinkage, very crudely, allows a chipmaker to cram more transistors onto the same bit of silicon. For a company like Nokia, this means higher-performance products that guzzle less power. Tommi Uitto, the head of the Finnish vendor's mobile networks business group, ticks off all the different areas to benefit – capacity, connectivity, energy consumption and product cost.

But there is doubt this market for customized RAN silicon – so-called application-specific integrated circuits (ASICs) – will be able to adapt so quickly in future. Each year, the broader RAN market generates between $40 billion and $45 billion in sales, depending on which analyst firm you ask. The overall market for semiconductors, meanwhile, hauled in about $574 billion last year, according to the Semiconductor Industry Association (SIA). At its dwarfish size, the RAN market simply lacks the research and development (R&D) muscle to keep jumping quickly from one process node to the next.

The Limits of Performance

Such is the claim of Intel, which spies an opportunity to divert billions of dollars in future RAN spending away from custom silicon products and into its x86, "general purpose" processors. "We keep delivering and refreshing every 18 to 24 months," said Sachin Katti, the head of Intel's network and edge group, during an interview with Light Reading at this year's Mobile World Congress. "Custom silicon cannot afford to refresh its design every 18 to 24 months because the market is much smaller."

The logic is compelling. Intel reported total revenues of $63.1 billion last year, making it substantially bigger than the entire RAN industry. At its data center and AI group, which mainly supplies x86 chips for server central processing units (CPUs), sales were about $19.2 billion. The same underlying technology is used across these chips and those Intel is pitching for the RAN – hence the general-purpose label. This gives Intel a much larger addressable market and a bigger pot of money to reinvest.


Continue reading this article on Light Reading

About the Author(s)

Iain Morris

International Editor, Light Reading

A skilled writer and editor with experience of writing about technology, telecommunications, business and sport for publications including Light Reading, The Observer, The Economist, reports from the Economist Intelligence Unit and Telecommunications magazine.

Subscribe to the Data Center Knowledge Newsletter
Get analysis and expert insight on the latest in data center business and technology delivered to your inbox daily.

You May Also Like