Writing in The Telegraph, a right-wing-leaning UK newspaper, Arm co-founder Hermann Hauser gave a blunt assessment (again) on what he saw as negative and far-reaching implications of selling the UK-based firm to US chipmaker Nvidia.
He fears that Arm's neutral licensing model for the design of central processing units, used in many of the world's most popular smartphones, will be put in jeopardy and that Britain's arch tech company – albeit owned by Japan's SoftBank – will fall into decline and find it difficult to expand into other areas.
The result, argued Hauser, would not only be tantamount to Britain throwing in the towel when it came to upholding tech sovereignty in "crucial" smartphone processor components – Nvidia, according to this line of thinking, would have little obvious incentive to see Arm's technology sold to its competitors, and Arm's customers might balk at paying a rival – but also restrict the country's supply chains in a number of other important areas, particularly PCs, next-generation data center servers and AI graph processors.
Avoid supply-chain squeeze
"To understand the implications of such a takeover, we need to consider the markets for the three basic types of processor: scalar processors that power servers, PCs, smartphones and the IoT, and vector processors for graphics – not to be confused with graph processors for AI," wrote Hauser.
"The UK has good access to scalar processors with Arm for smartphones and IoT, but it suffers from the duopoly of Intel and AMD in PCs."
Hauser thinks Arm has a chance to increase competition for PC processors, not least because Apple has just moved from Intel to Arm for its iMac PCs. He fears that ditching the neutral licensing model, however, will stymie Arm expansion and reduce manufacturing competition.
"The real battleground of the future, though, is data centers using servers, and graph processors for AI," added Hauser. "Intel has a dominant position, together with AMD, but has stumbled at the latest semiconductor nodes. This is the great opportunity for Arm to resume its growth by developing new server processors and licensing them around the world. This would increase market choice beyond Intel and AMD."
Nvidia's vision, according to Hauser, does not chime with UK interests. The US company made its bid to buy Arm, he said, "to address the data centre server market – the core business of a failing Intel." The Nvidia plan, he added, was to "take the microprocessor crown from Intel by winning the server market based on Arm architecture."
What's the alternative?
Hauser applauded the UK's ambition to create a £1 trillion ($1.3 trillion) firm in Britain, and thought that Arm was best-placed to achieve that through expansion into next-generation server products, and "especially if combined with Graphcore's AI capabilities."
Graphcore, another British company, has designed an architecture custom-made for AI graph problems. It is regarded by many, including Hauser, as the world's leading developer of graph processors.
This path to growth, argued Hauser, would make Arm "the most comprehensive supplier of processors to the world."
Rather than a sale to Nvidia, he thought that Arm should be listed on the London Stock Exchange and that the UK government could be the "cornerstone investor," pumping in $1 billion to $2 billion, and also orchestrating investments from Arm's main licensees, such as Apple, Samsung, Sony, Infineon, NXP and Qualcomm.
"They all have an interest in Arm's neutral status, rather than a Nvidia monopoly," said Hauser.
— Ken Wieland, contributing editor, special to Light Reading