NVIDIA agrees to buy Arm for $40 billion

NVIDIA agreed to acquire the chip-design firm Arm from SoftBank and the SoftBank Vision Fund for $40 billion — $21.5 billion in NVIDIA stock, $12 billion in cash, and up to a further $5 billion contingent on Arm meeting financial targets, with an additional $1.5 billion in equity set aside for Arm employees. It would have been the largest acquisition in semiconductor history.

Arm does not manufacture chips; it licenses the instruction-set architecture that underpins nearly every smartphone processor and an increasing share of servers, to customers including Apple, Qualcomm, Samsung and — pointedly — NVIDIA’s own GPU rivals. NVIDIA framed the deal as building “the premier computing company for the age of AI,” combining its accelerator business with the architecture that already sat inside most of the world’s edge and mobile devices, and pledged to keep Arm’s headquarters in Cambridge, preserve its open-licensing model, and build a new AI research facility in the UK.

That neutrality pledge was also the deal’s central problem. Arm’s customers were, in large part, NVIDIA’s competitors, and several of them — reportedly including Qualcomm and Google — lobbied regulators against a transaction that would hand a chip designer with obvious incentives control over the architecture their own products depended on. The agreement required sign-off from antitrust authorities in the UK, US, EU and China, and NVIDIA estimated the process would take around 18 months.

It took longer than that, and it did not succeed: the deal collapsed in February 2022 after the US Federal Trade Commission sued to block it and UK and EU regulators raised similar objections, leaving Arm to pursue a public listing instead. The attempt nonetheless signalled how central chip-architecture control had become to the AI buildout, well before the compute crunch of the following years made that dependence widely discussed.