TOKYO, Aug 21 (Reuters) – When SoftBank Group-owned (9984.T) chip designer Arm files for a Nasdaq IPO on Monday, investors are set to hone in on a key question – will the company have “exponential growth” due to the boom in artificial intelligence as CEO Masayoshi Son claims?
Since acquiring the chip designer in 2016, Son has positioned Arm as the tech investment conglomerate’s crown jewel asset, and in recent months has enthusiastically talked up the role it could play in AI.
Arm “is at the centre of a group of AI-related companies to generate synergies” and “85% of SoftBank Group assets are AI-related companies overseas,” he told investors in June.
The billionaire also said he has spent months creating hundreds of inventions with AI-powered ChatGPT that he believes can be realised through Arm.
But Son has been short on details. Investors are hoping that the filing, which sources have said is due later in the day, will yield further clues about SoftBank’s AI strategy and whether Arm is worth as much as or perhaps more than a pre-IPO valuation of $64 billion.
That’s the figure SoftBank valued Arm at with the purchase of a 25% holding – the only part it did not directly own – from its Vision Fund unit, sources said on Friday.
Son has been known to over-tout Arm’s prospects before.
When SoftBank acquired the company in 2016, he said the Internet of Things (IoT) would be the “greatest paradigm shift in the history of humankind,” and Arm-designed chips would power every object newly connected to the internet. The shift has yet to materialise and IoT makes up a fraction of Arm’s revenues.
Analysts also caution that Arm does not sit at the centre of the AI boom but is more AI-adjacent.
“The excitement that kicked this all off was really on the software side and the platform side, with OpenAI coming out with tools that could take advantage of large language models to create content,” says Kirk Boodry of Astris Advisory Japan.
“That’s not what Arm is, it’s not in any way related to that.”
Boodry, who values Arm at around $47 billion, said SoftBank’s $64 billion valuation figure was likely motivated in part to reward its Vision Fund limited partner investors, which include Saudi Arabia’s and UAE’s sovereign wealth funds.
“An intra-company transaction isn’t as strong a comparable as if it were with a third party,” he said.
In contrast, graphics chips specialist Nvidia (NVDA.O) has emerged as the biggest beneficiary of the AI boom, with its advanced semiconductors powering the data centres behind large language models such as ChatGPT.
To some extent, Arm can ride on Nvidia’s coattails as Nvidia’s chips must be coupled with energy-efficient central processing units (CPUs) – Arm’s speciality.
Nvidia has developed a “superchip” for use in data centres – the GH200, which contains CPUs based on Arm architecture. That said, the chip must compete with a host of alternatives.
“Not every Nvidia GPU has to be sold with an Arm CPU, they just happen to offer a superchip that combines the two,” said Rolf Bulk, an analyst at New Street Research.
Arm customers are also making inroads in AI. Qualcomm (QCOM.O) and Apple (AAPL.O) have designed portions of chips geared toward AI processing and cloud computing companies like Amazon (AMZN.O) and Alphabet’s (GOOGL.O) Google have built AI-focused chips that don’t use Arm technology.
But Bulk said Arm’s opportunity lies in AI and machine learning moving away from centralised cloud servers towards the devices used by end users, such as phones, home appliances and machinery components.
These devices will require specific intellectual property of the kind Arm has developed extremely successfully for other architectures in the past, he added.
The potential for AI synergies between SoftBank portfolio companies is, however, another matter and analysts question whether, as Son says, 85% of those firms can be described as AI-related.
“That is a stretch. Many SoftBank portfolio companies will embrace and apply generative AI but that does not make them AI companies,” said Victor Galliano of Galliano’s Latin Notes, who publishes on Smartkarma.
Reporting by Anton Bridge; Editing by Edwina Gibbs
Our Standards: The Thomson Reuters Trust Principles.