SoftBank snaps up UK chipmaker Graphcore

Bristol-based Graphcore makes high-performance semiconductors that are essential to AI. Financial details of the deal were not disclosed, although it was reported to be worth around $600m.

That would be a marked discount to an earlier valuation of $2.5bn, however, which Graphcore was valued at in 2022 follow a $222m fundraising.

The eight-year old firm made a pre-tax loss of $205m on sales of $2.7m in 2022, the most recent year for which accounts are available. In October 2023, it acknowledged it would need to raise new capital.

Its backers include Microsoft, Molten Ventures and OpenAI co-founder Ilya Sutskever.

Nigel Toon, co-founder and chief executive, said SoftBank deal was a "tremendous endorsement" of Graphcore’s ability to build "truly transformative AI technologies at scale".

He continued: "Demand for AI compute is vast and continues to grow. There remains much to do to improve efficiency, resilience and a computational power to unlock the full potential of AI.

"In SoftBank, we have a partner that can enable the Graphcore team to redefine the landscape for AI technology."

Vikas Parekh, managing partner at SoftBank Investment Advisors, said: "Next generation semiconductors and compute systems are essential in the AGI [artificial general intelligence] journey. We’re pleased to collaborate with Graphcore in this mission."

Toon and co-founder Simon Knowles, the firm’s chief technology officer, will stay on after the deal. Graphcore will retain its name and remain headquartered in Bristol.

Toon told the Financial Times that SoftBank would provide a "huge amount" of resources, allowing it to take on US market leaders Nvidia and AMD, including increasing its UK headcount.

Masayoshi Son, SoftBank’s founder, is looking to position the firm at the centre of AI’s development.

SoftBank already majority owns fellow UK chip maker Arm.

Derren Nathan, head of equity research at Hargreaves Lansdown, said: "Graphcore’s commercial traction has been underwhelming, but its intelligent processing units are claimed to have some significant advantages over Nvidia when it comes to training AI models.

"That remains to be seen, but this is a hot space and it’s a somewhat disappointing to see British innovation move to overseas ownership. In days of yore, this would have been a prime candidate for floatation Aim, but the risk appetite in the London market has all but dried up."