Arm Holdings CEO Rene Haas told Bloomberg on Tuesday that the company may reach its $15 billion AI chip revenue target ahead of schedule, a sign that AI infrastructure demand is still moving faster than chipmakers expected.
Arm is no longer talking about its data center ambitions as a distant licensing story. When the company unveiled its AGI CPU on March 24, it set a target of $15 billion in annual chip revenue by fiscal 2031. That was a big step for a business known for licensing chip architecture rather than selling finished silicon. Now, barely two months later, Haas is already saying the timeline could move faster.
The reason is simple enough: hyperscalers are still trying to secure more efficient compute for AI inference, and they are doing it before the market has settled into a normal supply rhythm. According to Bloomberg, Haas said demand for Arm's new AI-focused CPU is running ahead of the company's original expectations. That matters because inference is not a one-time training job. It is the continuous work behind chatbots, coding agents, search tools, recommendation systems, and the growing set of AI services that need to stay on all day.
The AGI CPU is the pivot point. Arm says the chip uses up to 136 Neoverse V3 cores, supports high-bandwidth DDR5 memory, and is designed for AI data centers where CPUs have to coordinate accelerators and keep latency under control. Its product brief lists configurable thermal design power rather than one fixed number, with top configurations reaching into the 400-watt range. In practical terms, the pitch is not that Arm replaces the GPU. It is that the CPU layer becomes more valuable as AI workloads spread across larger and more specialized systems.
Meta is the lead partner and first major customer, while OpenAI, Cloudflare, Cerebras, F5, Positron, Rebellions, SAP, and SK Telecom have been named among commercial customers or early adopters. Arm has also pointed to support across a broader ecosystem that includes AWS, Google, Microsoft, Nvidia, Broadcom, and TSMC. That roster is important because it shows the company is not trying to win by standing outside the existing AI supply chain. It is trying to sit deeper inside it.
Before the AGI CPU entered the picture, Arm's core business was already accelerating. For fiscal 2026, which ended March 31, the company reported record revenue of $4.92 billion, up 23% from a year earlier. Royalty revenue rose 21% to $2.61 billion, while licensing and other revenue rose 25% to $2.31 billion. In the fourth quarter, data center royalty revenue more than doubled year over year. That is the underlying business, before production chip sales become a major contributor, already reflecting the AI infrastructure buildout.
On Arm's May earnings call, management said it had line of sight to more than $2 billion in AGI CPU demand across fiscal 2027 and fiscal 2028, more than double the level discussed at launch. It also kept its near-term revenue outlook more cautious because production is still constrained by supply chain capacity. That distinction matters. The demand signal is strong, but the revenue curve still depends on wafers, packaging, memory, testing, and the unglamorous mechanics of getting hardware into racks.
What this means for the Nvidia narrative
Arm's move into production silicon is sometimes framed as a direct threat to Nvidia, but the real dynamic is more complicated. Nvidia's Vera CPU uses Arm cores. AWS Graviton, Google Axion, and Microsoft Cobalt are also Arm-based. The company gets paid when more Arm-based chips ship, whether the logo belongs to a cloud provider, an accelerator vendor, or Arm itself.
That makes Arm less like a single challenger and more like a toll collector across a widening set of AI infrastructure choices. If hyperscalers keep designing their own silicon, Arm benefits. If Nvidia ships more CPU and GPU platforms that use Arm architecture, Arm benefits. If cloud customers push for better performance per watt and more custom server designs, Arm is likely to be part of that conversation too.
Haas has argued that data centers will need more than four times today's CPU capacity by the end of the decade. That forecast is aggressive, but it fits the direction of the market. Training gets the attention because the models are expensive and the numbers are large. Inference may become the more durable business because it repeats constantly, across millions of users and devices.
The valuation question founders should watch
Arm's stock has surged in 2026, and valuation now carries a heavy amount of expectation. Recent market trackers put the shares at well over 100 times forward earnings, with some measures of trailing valuation far higher. That does not make the business weak. It does mean investors are already pricing in a lot of the AI upside that Haas is describing.
For founders and operators, the cleaner signal is not the stock multiple. It is the customer behavior behind the numbers. Hyperscalers are committing to CPU capacity earlier, custom silicon is becoming more central to AI cost control, and Arm's architecture is gaining leverage in parts of the data center that used to look like a slower-moving server market.
The next thing to watch is whether Arm can turn that leverage into better economics, not just bigger revenue. Its licensing model gives it room to raise royalty rates as customers use more advanced designs, while selling its own CPU gives it a new way to capture value directly. That is a different business from the smartphone-heavy Arm many investors first understood.
For anyone planning AI infrastructure products or trying to forecast compute costs, Arm's revenue curve is now one of the better indicators to follow. It touches cloud platforms, chip designers, networking hardware, and edge devices. If Haas is right that the $15 billion target arrives early, the message is clear: the AI buildout is still being pulled forward, and the CPU layer is becoming harder to ignore.
Also read: Bernie Sanders wants the government to own half of OpenAI and Anthropic and the AI industry is already pricing in the risk • SpaceX is carving out 5% of its IPO for insiders and waiving lock-up rules for them while everyone else waits • GoPro is running out of road as the AI memory crunch turns a camera company into collateral damage
