On September 13, 2020, Nvidia announced its intention to purchase Arm Holdings from Japan’s SoftBank for up to $40 billion. It will be the largest purchase in the global semiconductor industry if the deal goes through. The purchase will potentially bring Nvidia a competitive advantage over a technology that powers everything from mobile devices to data centers.
Nvidia and Arm have very different expertise and business models although they share an industry. Nvidia makes graphics processing units (GPUs), expensive and specialized accelerator chips that are originally developed to handle video streams. Nowadays, GPUs have become the main computing engines used to train artificial intelligence systems, making them a key technology in data centers. In contrast, Arm licenses blueprints for general-purpose chips that support a variety of devices ranging from smartphones to cars. Arm’s customers ship more than 20 billion Arm designed chips every year.
One motive for Nvidia’s purchase is to use Arm’s expertise and build its own version of the general-purpose processors that power the data-center computers. Jensen Huang, Chief Executive of Nvidia, commented that Arm’s technology was “quite extraordinary” because it is very energy efficient. Data-intensive tasks like artificial intelligence can cause massive increase in energy demand. Therefore, Arm’s low-power design fits well for data centers and can be a key weapon in the competition with Intel in the lucrative data center market.
Before announcing the deal with Arm, Nvidia had already reported strong demand from data center customers during the pandemic and had seen a 50% increase in revenue in the latest quarter. Its shares had risen by about 56% in the three months since its earnings report in April, propelling it past Intel to become the world’s most valuable chipmaker.
However, the deal has caused concerns. Major customers of Arm including Apple, Samsung, and Qualcomm have not shown any sign of supporting the deal. Arm’s business model relies on licensing designs for a variety of chips, not on competing with its customers by selling chips directly. Customers worry that once the deal is done, Arm will lose this neutrality and give Nvidia, a major rival for Arm’s customers, preferential access to its designs.
As a response, Nvidia has expressed the intention to keep Arm’s business model intact. As Nvidia strives to supply a complete data center computing platform, it would sell parts of the technology separately, Mr. Huang said. Other companies would still be able to license its intellectual property for use in their own chips.
Most of Arm’s customers choose to wait and see whether Mr. Huang will live up to his promise. Nevertheless, if the two sides push through the deal, this transaction will have the potential to change the landscape of the semiconductor industry in the near future.