ARM Chips Go From Smartphones to Networks
ARM Holdings Plc (ARM), whose chip designs are used in most smartphones, is set to power the networks that run them as it steps up competition with Intel Corp. in a $9 billion market. — With ARM already working with Hewlett-Packard Co. (HPQ) …
Context & Ripple Effects
ARM's push into networking extends the rivalry it opened back in October 2007, when The Register reported its plan for a multi-core chip aimed directly at Intel. What changes with this move is placement: instead of contesting Intel's PC franchise head-on, ARM is carrying its phone-proven, low-power design into the equipment that moves mobile traffic, with Hewlett-Packard already working with the company on network gear.
The story also travelled fast — PC World picked it up twice alongside Bloomberg's original — which suggests editors read it less as an HP supply arrangement than as the next front in ARM versus Intel, this time in a $9 billion market.
First-order effects
- Intel faces a credible architecture rival in the $9 billion networking-chip segment, where ARM's power-efficient cores compete on energy cost rather than raw clock speed.
- Hewlett-Packard gains a second processor option for its network equipment, loosening single-architecture dependence for gear that runs continuously.
Second-order effects
- Other network-equipment makers acquire negotiating cover to demand ARM-based options or better pricing from their existing silicon suppliers, compressing margins across the vendor base.
- ARM's licensee ecosystem wins a second volume market beyond handsets, shifting the royalty mix toward infrastructure silicon and giving chip designers a reason to optimize cores for always-on workloads.
Third-order effects
- If energy-per-workload keeps winning sockets in always-on infrastructure, the industry splits structurally between Intel's integrated design-and-fabricate model and ARM's licensed-cores model, giving large buyers genuine dual-source leverage.
- Networking would serve as the beachhead from which ARM-based designs contest progressively larger compute tiers — plausible given the 2007-to-2012 arc, but unproven beyond edge equipment as of early 2012.
The trend: Licensed low-power chip architectures are expanding out of handsets and into fixed network infrastructure, eroding Intel's architectural hold one workload class at a time.