Amid the excitement surrounding AI, which is prompting developers to create related products that demand more Arm-based ASIC chips, Arm anticipates reaping the benefits of this surge. Additionally, the rise of DeepSeek is expected to further benefit Arm.
During the earnings call on February 5, Arm CEO Rene Haas highlighted the growing trend of hyperscalers customizing silicon based on Armv9 and CSS technology to handle increased chip complexity.
Haas stated Arm is gaining significant market share in the data center, with major players like AWS, Microsoft, Google, and Nvidia adopting ARM-based chips such as Graviton, Cobalt, Axion, and Grace. AWS recently revealed that over 50% of new CPU capacity installed in the past two years was on Graviton, with more than 90% of its top 1,000 EC2 customers using the technology. Arm also boasts the largest global developer community, with over 20 million developers.
Haas also pointed to Nvidia's launch of Project Digits, which integrates the ARM-based Grace CPU and Blackwell GPU into the GB10, powering the world's smallest AI supercomputer. The Grace CPU, based on ARM CSS, features 20 power-efficient Arm cores, and the new platform will allow developers, data scientists, and AI researchers to build and deploy inference models more efficiently.
He confirmed that AWS's Graviton chip is being used by over 90% of its top 1,000 EC2 customers. When discussing Microsoft's Cobalt 100 chip, Haas noted that he would defer to Microsoft for specific shipment details but emphasized the growing momentum for the Cobalt platform and broader Arm-based products in the cloud.
Continued investments in R&D to retain competitiveness
Arm CFO Jason Child provided an update on the company's financial performance, highlighting key metrics such as annualized contract value (ACV), remaining performance obligations (RPO), and research and development (R&D) spending. ACV for the third quarter rose 9% year-over-year, slightly below the recent low-teens growth rate but above Arm's long-term plan. RPO decreased slightly as the company delivered products, releasing revenue from the backlog.
Child emphasized Arm's strategy of investing in R&D to maximize future revenue, which resulted in a 9% year-over-year increase in operating costs, reaching US$522 million for the fiscal third quarter. Despite the higher spending, Arm achieved near-record non-GAAP operating profit ofUS$442 million. Looking ahead, Arm plans to continue balancing long-term investments in R&D with short-term profitability.
Arm predicts surging demand for AI computing
Haas highlighted two significant projects that underscore Arm's central role in the AI ecosystem. The first is Stargate, an AI infrastructure project in partnership with SoftBank, OpenAI, Oracle, Microsoft, and Nvidia. The project will initially see a US$100 billion investment, with plans for US$500 billion over time. Arm will serve as the CPU of choice for the platform, alongside Nvidia's Blackwell CPU and Grace.
The second project, Crystal Intelligence, focuses on developing AI agents for knowledge work, further enhancing Arm's position in enterprise AI. Haas believes advancements in AI, both for training and inference, will drive demand for computing power across the AI cloud, with Arm solutions addressing needs from the cloud to the edge for years to come.
DeepSeek's rise benefits Arm
Haas discussed the impact of DeepSeek's advancements, particularly their work on the V3 and R1 models. DeepSeek's models have built on existing frontier models, improving inference efficiency. Haas praised these innovations for two main reasons: they enhance industry efficiency by reducing costs, and they expand overall compute demand.
He emphasized that this is beneficial for Arm because AI workloads need to run everywhere, from smartphones and earbuds to cars, where power and computing resources are constrained. While Arm's Grace Blackwell chips are excellent but too large for such devices, Arm's efficiency in computing is a key advantage in these smaller, power-sensitive applications.
Haas pointed out that with major tech companies investing heavily in AI, there's still much room for growth, and DeepSeek's improvements will increase compute demand, benefiting Arm due to its efficiency-focused products.