Arm CEO Rene Haas Engineers $1.5 Trillion AI Pivot From Smartphones to Silicon
Image: Reuters AI
Main Takeaway
Arm pivots from licensing to making its own AI chips, projecting $15B revenue by 2031 as CEO Haas takes broader SoftBank AI role.
Summary
The $15 billion chip that changed everything
Arm Holdings unveiled its first in-house silicon, the Arm AGI CPU, on March 25 at its Everywhere conference in San Francisco. CEO Rene Haas told CNBC the move will lift Arm from roughly $2.4 billion in 2025 revenue to $15 billion annually by 2031—more than a six-fold jump. Meta signed on as the launch customer, and Business Insider reports OpenAI is also partnering. The reveal sent the stock up 6 % in after-hours trading and has continued to rally as investors price in the licensing-to-manufacturing shift.
Why smartphones are no longer enough
Arm’s mobile IP still ships in 99 % of the world’s phones, but that market is mature. Haas told Bloomberg’s Tom Mackenzie that cloud and data-center silicon will dominate Arm’s business “in the years ahead.” The company sees a $100 billion chip opportunity tied to generative AI demand and a broader $1.5 trillion addressable market for AI infrastructure. Smartphones simply don’t scale like racks of AI accelerators that burn megawatts and need constant upgrades.
Betting on energy efficiency as the moat
Power per watt is the new transistor count. Haas argues Arm’s heritage in battery-sipping mobile designs gives it an edge in the AI data-center gold rush where electricity is the hard constraint. At the Stratechery interview he framed it bluntly: “The only thing that matters is performance per watt per dollar.” That pitch has already landed Meta as a lead partner for the AGI CPU, and Haas hinted more hyperscalers are in late-stage trials.
SoftBank doubles down with Haas in charge
Masayoshi Son is handing Haas a wider brief. According to the Financial Times and confirmed by Reuters, Haas will keep the CEO title at Arm while also overseeing “much of SoftBank Group’s international business.” The elevation signals SoftBank’s intent to centralize its AI strategy around Arm’s silicon and IP, rather than funding a portfolio of competing chip startups. Investors cheered: Liontrust’s Clare Pleydell-Bouverie told Bloomberg the move puts Arm “in pole position for AI.”
What this means for the broader chip ecosystem
The pivot rattles the cozy ecosystem Arm built over three decades. Nvidia, Broadcom and Qualcomm all depend on Arm cores for their own AI chips; they now face a licensor that also competes directly. Yet the hyperscaler customers—Meta, OpenAI, Google, Microsoft—are cheering because an energy-efficient alternative to Nvidia’s GPUs lowers their cost curve. Expect more custom silicon from cloud giants, tighter margins for traditional chip houses, and a fresh wave of Arm-based AI startups funded by SoftBank’s Vision Fund.
What happens next
Volume production of the AGI CPU starts late 2026. Haas told Fox Business that early silicon is already back from TSMC at 3 nm. If yields hold, Arm could ship tens of thousands of units to Meta’s data centers by mid-2027, with broader availability opening in 2028. Meanwhile, SoftBank’s board will formalize Haas’s expanded role next quarter. Watch for acquisitions: insiders say Son has a shopping list of AI software startups to bolt onto Arm’s hardware stack.
Key Points
Arm unveiled its first proprietary chip, the Arm AGI CPU, targeting AI data centers with Meta and OpenAI as lead customers.
CEO Rene Haas forecasts revenue growing from ~$2.4B in 2025 to $15B annually by 2031 as the company shifts from IP licensing to direct silicon sales.
SoftBank is expanding Haas’s role to oversee much of its international AI operations, signaling deeper integration of Arm silicon across the group.
Arm positions energy efficiency as its competitive edge against Nvidia GPUs, aiming to capture share in the $1.5 trillion AI infrastructure market.
The pivot pressures traditional Arm licensees like Nvidia and Qualcomm while giving hyperscalers a lower-power alternative for AI workloads.
FAQs
The Arm AGI CPU is Arm’s first fully in-house data-center processor optimized for AI workloads. It leverages Arm’s power-efficient architecture to deliver high performance per watt, making it attractive to hyperscalers facing rising electricity costs.
For 30+ years Arm licensed CPU blueprints to companies like Qualcomm and Apple. Now it will also sell finished chips directly, competing with some of its own customers while capturing a larger slice of AI infrastructure spend.
Meta operates some of the world’s largest AI training clusters and is under pressure to cut power bills. Arm’s energy-efficient cores align with Meta’s sustainability targets and reduce total cost of ownership compared to traditional GPUs.
In the short term, minimal—Nvidia still dominates high-end AI training. Longer term, if Arm’s chips scale and software matures, they could siphon off inference workloads, shaving Nvidia’s data-center growth and margins.
Limited quantities ship to Meta and select partners in mid-2027. Broader commercial availability is slated for 2028, pending TSMC 3 nm yield improvements and customer qualification cycles.
No. Licensing remains the core business today and funds R&D. Haas emphasizes that selling chips is additive, although some licensees worry about future conflicts of interest.
Source Reliability
42% of sources are highly trusted · Avg reliability: 70
Go deeper with Organic Intel
Our AI for Your Business systems give you practical, step-by-step guides based on stories like this.
Explore ai for your business systems
