Rene Haas posed half‑reclined on a couch in his San Jose office, a basketball in hand. He grimaced at the thought that photographers would spin it into a headline about him "sleeping on the job." He still made time for a long interview, then hopped on a call with Masayoshi Son, SoftBank’s chairman and Arm’s largest shareholder.

The plot twist: Arm builds a chip

Arm is known for licensing CPU architectures to the world. Apple, Samsung, Microsoft, Amazon, Nvidia, Qualcomm, and many others design chips that use Arm technology. For years Arm has made money from royalties and IP. Now the company is taking a major step: it is launching its own server CPU called the Arm AGI CPU. For a firm that rarely made physical silicon, this is a big strategic move.

A quick history lesson

  • Arm traces its DNA back to Acorn Computers and early RISC work from the late 1970s and 1980s.
  • In the 1990s the company pivoted to licensing designs instead of building chips itself.
  • SoftBank bought Arm in 2016. Nvidia tried to acquire Arm in 2020 but regulators blocked that deal.
  • Rene Haas joined Arm in 2013 from Nvidia, led the IP products group, became CEO in 2022, and later took Arm public while SoftBank still retains about 90 percent ownership.

Why build a server CPU now?

Haas says Arm has quietly evolved from an IP company into a compute platform company. That matters because software and hardware are tightly linked. When a platform company makes a physical product, it can move the ecosystem forward: think of a company shipping a reference device that encourages developers and partners to optimize for that platform.

Haas uses examples like Microsoft shipping Surface or Google shipping Pixel. Those products are a small slice of the market, but they drive software and optimizations that benefit many partners. Arm’s goal is the same: build something that nudges the broader Arm ecosystem forward.

What is the Arm AGI CPU?

  • Name: Arm AGI CPU. Yes, AGI as in artificial general intelligence.
  • Target: data center workloads, especially agentic AI tasks where lots of CPU work coordinates with GPUs and other accelerators.
  • Key strength: extreme power efficiency. Arm says its mobile roots give it an efficiency advantage that matters for energy hungry AI infrastructure.
  • Fabrication: TSMC is manufacturing the chips. Arm is also working with partners like Super Micro and Foxconn to deliver reference server racks.

Customers and market fit

Arm named Meta as its first customer. Other early partners include SK Hynix, Cisco, SAP, and Cloudflare. The pitch is that some customers want an Arm-based server solution but either do not want or cannot build it themselves. Cloudflare and others are examples of businesses that need more vendor choice.

Haas argues that this chip will not replace GPUs. Instead it will handle workloads that only CPUs can do, particularly in systems that run many software agents. GPUs still have a role, but they will need CPUs to manage and orchestrate a lot of the work.

Will this upset partners?

Yes, some companies might be annoyed. Haas is candid: he expects Intel and AMD to feel pressure because Arm’s server CPU competes with x86 offerings. He downplays the idea that Nvidia will be deeply hostile. He points out that more software and optimization for Arm benefits anyone building on Arm technology. Still, building a CPU inevitably shifts market dynamics.

Haas also noted that big cloud players like Amazon will likely keep their own chips. Arm did not build AGI CPU specifically for Amazon. Instead the company sees a large, underserved market for Arm server CPUs across many customers.

Operations, team, and risk

Moving from an IP licensing model to building and shipping silicon brings new operational challenges. Haas admits Arm will have to manage yields, volume forecasting, returns, and gross margins in ways it historically did not.

To prepare, Arm added roughly 2,000 engineers focused on backend design, implementation, and subsystems. The new CPU is built on compute subsystems that Arm has already developed and which partners have shipped in silicon. That gives Haas confidence the first generation can be a volume candidate, though he acknowledges chips often need multiple generations to reach full market impact.

Haas the CEO

Haas says the CEO sets the company tone. He favors a founder-style, risk-taking culture that embraces bold bets and fast learning. He speaks with SoftBank’s Masayoshi Son frequently. He mentioned he and Son talk often enough that Son knows his daily routine.

He is pragmatic about timing. Economic uncertainty, geopolitics, or supply shortages have not changed the plan. This is a long journey and the Arm AGI CPU is the first step.

Quick takes from Haas

  • On Intel: historic.
  • On RISC-V: nascent.
  • On Sam Altman: brilliant, long‑term thinker.

Bottom line

Arm is trying something bold. It wants to use a physical product to accelerate the Arm software ecosystem and capture a share of the fast‑growing AI data center market. That could help many partners, but it will also shift the competitive landscape and annoy some established players. Haas is confident, the company has built teams and partners to support the move, and the first Arm AGI CPU is meant to be just the beginning.