In September 2020, Nvidia announced an agreement to buy Arm from SoftBank for approximately $40 billion. The deal would combine the dominant GPU company with the company whose architecture underpinned almost every smartphone and an increasing share of laptops and servers. The reaction across the chip industry was immediate and almost entirely sceptical.
Arm itself was an unusual company. It did not manufacture chips. It licensed its architectures to companies that did, including Apple, Qualcomm, Samsung, MediaTek, and many others. The neutrality of Arm was central to its business model. The chip designers paying for Arm licences trusted Arm not to favour one customer over another, because Arm had no chip business of its own to favour.
Nvidia, by contrast, had a very large chip business of its own. The acquisition would put a major chip designer in control of the architecture used by its potential competitors. The structural conflict of interest was obvious enough that most of the major Arm customers raised concerns publicly within days of the announcement.
The regulatory review was going to be extensive. Deals at that scale and with that kind of competition implications normally take a long time to clear, and this one had to navigate competition authorities in the US, the UK, the EU, and China. The optimistic view was that with appropriate behavioural commitments the deal could close in eighteen months. The realistic view was that the deal might not close at all.
What the announcement revealed was that the semiconductor industry had become strategically important enough that consolidations of this kind would be examined under a much harsher light than they had been a decade earlier. The combination of national security considerations, geopolitical tensions over chip supply chains, and the increasingly central role of AI accelerators all meant that chip M&A was no longer just about competition policy.
The deal would, ultimately, fall apart. By early 2022, Nvidia and SoftBank would mutually agree to terminate the agreement after it became clear that regulatory clearance was not going to happen on any reasonable timeline. Arm would eventually go public on its own, and the architecture would remain neutral. But the deal being attempted at all said something about where the chip industry sat in 2020.