Up until the last few years, Arm chips dominated the mobile device market yet were scarce in the data center. But guided by Arm CEO Rene Haas, the company has recently enjoyed remarkable success in the data center: the number of customers using Arm chips since 2021 has grown 14-fold, and now tallies 70,000.
Helping drive this dramatic ascent is the technology that is fueling most of today’s investment: artificial intelligence. Arm claims that 12 times as many startups, a cohort focusing on generative AI, are using its chips today than in 2021. The advantage for startups is that renting Arm chips tends to offer better cost-for-performance metrics than renting NVIDIA GPUs, even as NVIDIA remains the top high-performance choice for companies that can afford them.
Historically, the greatest competition in the data center for Arm has been chips built with the x86 architecture, made by AMD and Intel, which were the default for running routine enterprise workloads. Enterprise buyers favored x86 for its advanced software ecosystem and greater hardware compatibility.
Yet Arm has always been well regarded for its combination of low energy consumption and relatively high power, hence its dominance of the smartphone sector, not to mention microwaves, cars, and countless embedded devices. In the last few years, Arm has maintained these energy-power advantages and also designed heftier chips for the robust needs of the data center.
The hyperscalers noticed. Microsoft, AWS and Google have each built their own Arm chips and incorporated them in their global networks of data centers—running AI workloads. This has opened a wide door for the expansion of Arm chips for data-intensive uses.
At this point, Arm is clearly on a roll. In June the company earned a spot on Time Magazine’s 100 Most Influential Companies, which noted that more than 22 million developers are coding software designed for Arm, which in turn creates more demand for Arm chips. Additionally, the number of applications supported by Arm chips has just about doubled since 2021, now standing at 9 million.
Signifying the greatest change in the company’s strategy: this year Arm is releasing its own chip. The company has always designed chips for other companies, but by the end of this year it expects to start selling its own semiconductors, which will be fabricated by TSMC. Meta is reportedly a customer. The initiative is not without its risks, since selling its own chip places it in competition with its customers.
Arm’s recent business strategy has been rewarding. The company generated more than $4 billion in revenue from April 24 and the 12 months following. But all is not positive on the revenue front. When the company released financial results in early May, it was one of several semiconductor companies issuing weak revenue forecasts. Part of this was due to tariffs, which would impede the supply chain for a wide array of chips.
More specific to Arm, smartphones were expected to see lower sales due to the tariffs. However, it was exactly this heavy reliance on the smartphone market that has prompted Arm to push more aggressively into the data center sector.
