Key Takeaways
While Arm and Nvidia operate in distinct semiconductor markets, the two businesses are similar enough that regulators’ antitrust concerns stymied an attempted merger in 2022.
Two years later, Arm is expected to unveil a new data center strategy at the end of the month that could bring it into more direct competition with Nvidia, suggesting that monopoly fears may have been justified.
In the 21st century, no companies have done more to disrupt the global semiconductor market than Arm and Nvidia.
The former is the go-to CPU provider for mobile computing and almost every smartphone in the world uses components designed by Arm. Meanwhile, the latter is unrivaled in the GPU space, where it has come to dominate the provision of high-end AI chips for data centers.
Having agreed to buy Arm from its majority owner SoftBank for $70 billion, Nvidia’s plans were scuppered by “significant regulatory challenges,” but the two firms still collaborate closely.
However, at times the line between collaboration and competition has become blurred as Arm’s moves in the AI chip sector have grown increasingly bold in recent years.
While most people associate AI data center workloads with GPU computing, there is a growing demand for more energy-efficient CPU-based solutions. This presents an opportunity for CPU players like Arm to cut into a market that is currently dominated by Nvidia’s GPUs. (Conversely, Nvidia’s plans to develop its own ARM-based CPUs will provide additional revenue to Arm, which holds the intellectual property rights to ARM designs.)
Of course, GPUs are expected to remain critical in AI training for the foreseeable future. But solutions like Arm Neoverse have proven that less computationally intense AI inference (i.e. the process of running live data through a trained AI model) can be done much more efficiently with a CPU architecture.
As the AI market evolves, rising adoption could play directly into Arm’s hands as inference supplants training as the primary growth driver.
Just as the AI training boom has fueled a surge in demand for Nvidia’s GPUs in recent years, as data centers look to ramp up their inference capacity, Arm’s AI-optimized CPUs will become more sought-after.
Post-IPO, Arm is currently valued at around $180 billion, more than double what Nvidia was going to pay for it in 2022.
The Japanese bank has done so well from its investment in the company that it is now doubling down on the semiconductor market.
In February, CEO Masayoshi Son announced plans for a $100 billion AI chip venture that could challenge Nvidia’s dominance . And on Thursday, July 12, Softbank closed a $500 million deal to acquire the UK-based AI accelerator startup Graphcore.
While it would be reductive to say that Arm’s gain is Nvidia’s loss, two years after the latter’s failed takeover bid, SoftBank’s shareholders may be thankful that the deal fell through.