{"slug": "arm-holdings-shifts-strategy-to-sell-its-own-data-center-chips-targeting-15b-in", "title": "Arm Holdings shifts strategy to sell its own data center chips, targeting $15B in annual revenue", "summary": "Arm Holdings, a SoftBank subsidiary, announced at its 'Arm Everywhere' event in San Francisco its first production silicon, the AGI CPU, a 136-core Neoverse V3 chip for AI data centers, breaking a 35-year tradition of only licensing designs. The company targets $15 billion in annual chip revenue by around 2031, with Meta Platforms as lead development partner and first customer, and customer commitments exceeding $2 billion through fiscal 2028. The move positions Arm as a direct competitor to its own licensees like Amazon, Google, and Qualcomm in the data center market.", "body_md": "Via 247wallst.com\n\n# Arm Holdings shifts strategy to sell its own data center chips, targeting $15B in annual revenue\n\nThe chip architecture giant breaks a 35-year tradition by launching its first production silicon, developed with Meta for AI workloads\n\nFor 35 years, Arm Holdings had one job: design processor blueprints and license them to everyone else. Now it wants to sell the finished product, too.\n\nAt its “Arm Everywhere” event in San Francisco, the SoftBank subsidiary unveiled the AGI CPU, its first-ever production silicon chip, purpose-built for AI data centers. The move represents the most significant strategic shift in Arm’s history, turning the company from a behind-the-scenes architecture licensor into a direct competitor to the very customers it has long served.\n\n## What Arm is actually building\n\nThe AGI CPU packs 136 Neoverse V3 cores into a chip with a 300-watt thermal design power. TSMC will manufacture it on its 3nm process.\n\nMeta Platforms signed on as the lead development partner and first customer. The chip is engineered for agentic AI workloads, the kind of inference tasks where autonomous AI systems need to reason, plan, and execute multi-step operations rather than just spit out a single response.\n\nArm claims these workloads demand up to four times more CPU cores than previous AI inference processes. The AGI CPU is designed to deliver over twice the performance per rack compared to x86 alternatives, with potential capital expenditure savings of up to $10B per gigawatt of data center capacity deployed.\n\nMore than 50 companies are supporting the AGI CPU platform, and Arm says customer commitments already exceed $2B through fiscal 2028.\n\n## The business math behind the pivot\n\nArm’s long-term revenue target tells the whole story. The company is aiming for $25B in total annual revenue, split between $10B from its traditional IP licensing business and $15B from direct chip sales. That $15B target for chip revenue is expected to materialize by around 2031.\n\nMeaningful revenue from the AGI CPU is expected to start flowing in the second half of 2026, with more substantial contributions arriving in fiscal year 2028. Production ramp-up begins in the latter half of this year.\n\nAfter some early market apprehension about Arm competing with its own licensees, shares rebounded roughly 16-20% following the announcement.\n\nThat risk is real, though. Amazon designs its own Arm-based Graviton chips for AWS. Google builds custom Arm-based processors for its cloud infrastructure. Qualcomm, MediaTek, and dozens of other companies depend on Arm’s architecture for their own products. All of them are now, at least in the data center market, competing with their IP supplier.\n\n## Why Arm couldn’t stay on the sidelines\n\nThe answer is straightforward: the AI infrastructure buildout is too massive and too fast for Arm’s licensing model to capture the full opportunity. Licensing fees are typically a small percentage of a chip’s selling price. When hyperscalers are spending hundreds of billions on data center capacity, selling the actual silicon is where the real money lives.\n\nArm’s architecture already dominates smartphones, powering virtually every mobile processor on the market. But in data centers, x86 chips from Intel and AMD have historically held the advantage. That dominance has been eroding steadily as Arm-based server chips from Amazon, Ampere, and others have gained traction on the back of superior power efficiency.\n\n## What this means for the competitive landscape\n\nThe data center CPU market just got considerably more crowded. Intel is fighting to regain relevance after years of manufacturing stumbles. AMD has been steadily gaining server market share with its EPYC lineup. Nvidia dominates AI training with its GPUs but has also been expanding into CPU territory with its Grace processor.\n\nThe execution risks are substantial. Arm has never manufactured and sold chips at scale. Managing a hardware supply chain through TSMC is fundamentally different from licensing IP.\n\nTSMC capacity allocation is another variable. With every major chip company competing for advanced node production, Arm’s ability to secure enough 3nm wafers to meet its $2B-plus in commitments will depend on its priority level relative to Apple, Nvidia, AMD, and others vying for the same fab lines.\n\nThe $10B per gigawatt in potential capex savings Arm is projecting would be significant for hyperscalers evaluating total cost of ownership across massive deployments.\n\n**Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our\n\n[Editorial Policy](https://cryptobriefing.com/editorial-policy/).", "url": "https://wpnews.pro/news/arm-holdings-shifts-strategy-to-sell-its-own-data-center-chips-targeting-15b-in", "canonical_source": "https://cryptobriefing.com/arm-holdings-data-center-chip-strategy/", "published_at": "2026-08-24 12:13:11+00:00", "updated_at": "2026-08-24 12:43:55.758153+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-chips", "ai-infrastructure", "ai-products"], "entities": ["Arm Holdings", "SoftBank", "Meta Platforms", "TSMC", "Amazon", "Google", "Qualcomm", "Intel"], "alternates": {"html": "https://wpnews.pro/news/arm-holdings-shifts-strategy-to-sell-its-own-data-center-chips-targeting-15b-in", "markdown": "https://wpnews.pro/news/arm-holdings-shifts-strategy-to-sell-its-own-data-center-chips-targeting-15b-in.md", "text": "https://wpnews.pro/news/arm-holdings-shifts-strategy-to-sell-its-own-data-center-chips-targeting-15b-in.txt", "jsonld": "https://wpnews.pro/news/arm-holdings-shifts-strategy-to-sell-its-own-data-center-chips-targeting-15b-in.jsonld"}}