In an unconventional twist on data center design, researchers at the University of California San Diego, backed by Google, have found a way to repurpose old smartphones into functioning data centers. The project, dubbed “phone cluster computing,” is building toward a full deployment of 2,000 retired Pixel smartphones working in tandem as a single computing system.
Google introduced the initiative through its research blog in a post titled “A low-carbon computing platform from your retired phones,” laying out early results from a modest but promising test run. The team started small, with just 20 phones networked together, and that tiny cluster was already capable of handling cloud computing tasks for more than 75 students at UC San Diego.
The science behind the idea comes down to processing power. Single-threaded performance on modern smartphone chips now rivals, or even beats, that of multicore server processors. The real gap lies in scale: a server packs dozens of powerful, multithreaded cores and vast memory reserves, while a smartphone makes do with a handful of heterogeneous cores and 8 to 12GB of memory.
That gap is exactly what researchers are working around. According to the Google team, the central challenge is identifying, or adapting, applications that can fit within a smartphone’s more modest capacity rather than expecting phones to match server-grade specs outright.
This Is Samsung’s Biggest AI Leap With Google Gemini
The next step is significant. Expanding from the initial 20-phone trial to a full 2,000-phone cluster would open the platform to hundreds of students at once. Google describes the scaled-up system as delivering “50 server-equivalents worth of compute at a fraction of the usual cost,” with the complete setup expected to go live in Fall 2026.
Why Repurpose Old Phones At All?
The answer, unsurprisingly, is AI. Demand for AI services has surged over the past year, prompting Google to commit a staggering $80 billion toward scaling its infrastructure. Google is far from alone in this gamble: other major players have poured hundreds of billions of dollars into data centers, betting heavily that mainstream AI adoption will eventually justify the spend.
That spending pressure cuts both ways. Google needs more paid subscriptions to its AI and Gemini products to help offset the costs, and it has responded in part by lowering prices on its base AI tier to attract new subscribers.
Against that backdrop, cheap, low-carbon computing built from discarded phones offers a potentially clever release valve, a way to shoulder part of the computing load while hardware manufacturers work to keep pace with AI’s relentless demand for power.