Arm Holdings signaled a major shift this week. Its data center operations stand poised to eclipse every other part of the business before long. The news comes as hyperscalers pour hundreds of billions into AI infrastructure, creating both opportunity and strain across the supply chain.
During the company’s fiscal fourth-quarter earnings call, CEO Rene Haas left little doubt. “Soon, the datacenter will be Arm’s largest business,” he said. “The direction is clear. Customers want Arm at the center of the AI datacenter.” The Register captured those remarks in detail shortly after the Wednesday disclosure.
Haas pointed to strong uptake of the firm’s new AGI CPU, unveiled in March for agentic AI workloads. Customer demand now exceeds $2 billion across fiscal 2027 and 2028. That figure more than doubles the initial target shared at launch. “Customer response to the Arm AGI CPU has been very strong,” Haas noted. One thing seems certain. The company has under-called CPU demand in this transition.
The AGI design packs 136 dedicated cores suited to running AI agents. Hyperscalers appear ready to deploy racks of these CPUs alongside racks of GPUs. Amazon and Microsoft already rely on Arm-based silicon for their cloud data centers. Now others signal interest too. Someone besides Meta has committed to roughly $1 billion worth of the new chips.
Yet execution hurdles remain. Arm has not fully lined up supply to meet the $2 billion in orders. Haas acknowledged the gap but said the company works to close it. Cash from these sales should begin flowing more meaningfully in fiscal 2027-28. For now the focus stays on scaling manufacturing partnerships.
Investors reacted with typical volatility. Shares climbed 10 percent above the prior close of $237 before retreating to $222. The mixed move reflects confidence in demand set against worries about timely delivery.
Arm’s latest results offered a solid base. Quarterly revenue reached $1.49 billion, up 20 percent from a year earlier. Full-year sales hit $4.9 billion, a 22.8 percent gain. The firm guided to $1.25 billion for the current period. Licensing and royalty streams both advanced, though the real story sits in future AI infrastructure projections.
Management reiterated its target of $15 billion in annual AI infrastructure revenue. CFO Jason Child added that overall IP licensing sales could double to $10 billion by 2031. Datacenter products will drive most of that expansion. The numbers point to a company in the midst of transformation. Once known for low-power mobile designs, Arm now eyes the heart of the world’s most power-hungry computing environments.
This pivot arrives at a charged moment for the broader industry. Big Tech’s capital spending has reached record levels. Google, Amazon, Microsoft and Meta reported more than $130 billion in combined quarterly capital expenditures during the first three months of 2026. They plan roughly $700 billion for the full year. The New York Times detailed those figures and noted the sum dwarfs many historic infrastructure efforts.
Analysts describe an arms-race dynamic. Four members of the Magnificent Seven alone could spend $650 billion on AI infrastructure in 2026. That includes chips, data centers, power delivery and networking. A Fortune article highlighted the short useful life of some hardware. Much of it may become outdated within three years, turning traditional capital expenditure into something closer to rapid consumption.
Power shortages complicate the picture. Roughly half of planned U.S. data center projects face delays or cancellations this year. The bottlenecks stem from limited electrical equipment, grid capacity and components still sourced partly from China. Tom’s Hardware reported on those constraints, citing transformer and switchgear shortages that slow everything down.
Yet demand signals keep intensifying. The total addressable market for data-center CPUs could reach $100 billion annually within a few years. Agentic AI, where autonomous agents handle complex tasks, drives much of the CPU revival. A Wall Street Journal story from March observed that CPUs once left for dead now sit at the center of new workloads. “That’s all CPU work,” one executive remarked in the piece.
AMD also posted strong data center results recently. Its segment grew 57 percent year-over-year, fueled by EPYC processors and Instinct GPUs. CEO Lisa Su credited accelerating AI infrastructure demand. The company’s shares jumped on the news. Such performance underscores how the CPU layer gains relevance even as GPUs handle the heaviest training lifts.
Arm’s own forecast calls for revenue to roughly quintuple to around $25 billion over the next five years. That exceeds some analyst expectations. The bet rests on Arm architecture becoming standard in AI servers. Early design wins with cloud providers provide a foundation. But scaling silicon production at the required pace will test the firm and its partners.
Supply chain realities extend beyond chips. Data center build times stretch long. Available power through 2026 is largely spoken for in many markets. Spot prices for Nvidia GPUs in the cloud have climbed sharply. One provider called the capacity crunch unlike anything seen in years. These pressures suggest the infrastructure wave will unfold unevenly, with winners and laggards determined by who secures resources first.
Geopolitical angles add complexity. Projects in the Middle East face new security calculations after reported drone incidents. Some hyperscalers now weigh fortified designs and diversified locations. Meanwhile, nations compete to attract these facilities with land, energy policy and incentives. The United States remains the focal point, but Abu Dhabi, Thailand and others push their own multi-gigawatt plans.
Longer term, projections grow staggering. Global investment in AI-ready data centers could approach $5.2 trillion by 2030. Power demand for new facilities in the U.S. alone might triple by then. Such scale explains why utilities, real estate developers and equipment makers all scramble to keep up.
Arm sits at an interesting intersection. Its energy-efficient designs, honed in mobile, now appeal to operators worried about electricity bills. The 136-core AGI part targets exactly the inference and agent workloads expected to dominate future data centers. If the company can convert that $2 billion demand into actual shipments, the revenue mix will tilt decisively toward the enterprise segment.
Challenges persist. Competition from Intel, AMD and custom silicon efforts remains fierce. Hyperscalers continue to explore in-house chips. Execution on supply must improve quickly. Still, the momentum feels unmistakable. Haas and his team project clarity in direction even if the precise timeline holds some uncertainty.
The earnings call and subsequent coverage paint a picture of measured optimism. Demand runs ahead of supply. AI infrastructure spending shows no signs of slowing. And Arm believes its architecture belongs at the core of what comes next. Whether that belief translates into sustained market leadership will unfold over the coming quarters. For an industry consuming ever more power and capital, those quarters promise to be eventful.
Recent reporting reinforces the trend. A Wall Street Journal article published hours after the results noted Arm’s doubled demand forecast for the new chips while highlighting the supply shortfall. The piece captured the tension investors felt as shares reversed course.
So the race continues. Billions flow into facilities that may need replacement sooner than traditional assets. CPUs regain prominence in an agent-driven future. And one British chip designer, once defined by its phone pedigree, now sets its sights on dominating the data center. The numbers support the ambition. The hard work of delivery lies ahead.
WebProNews is an iEntry Publication