Researchers at the University of California San Diego have shown how retired smartphones can be wired together into low-cost computing clusters. These small data centers built from devices that would otherwise become waste are taking new shape into something the team calls "phone cluster computing."
The process starts with stripping down the phone to its motherboard, removing the display, battery, and cameras — partly because some components, like batteries, aren't rated for a data-center environment. What's left gets networked into clusters that run real workloads.
The reason it works is that a phone most people call obsolete still has roughly half the compute of a modern server. Because of this, the UC San Diego team hopes to scale this up to a data center built from about 2,000 retired Pixel phones, giving hundreds of students and researchers low-cost, low-carbon cloud computing, with the full system expected to launch in Fall 2026.
Why now
The amount of environmental disruption from Data Centers is astronomical, which makes the timing of this research is particularly interesting.
Demand for AI services is fueling an unprecedented boom in data center construction. Companies including Google, Microsoft, OpenAI and Amazon are investing heavily in new infrastructure to support increasingly powerful AI models.
At the same time, electronic waste continues to grow worldwide. Many consumers replace smartphones every two to three years, even though the underlying hardware often remains functional. Researchers argue that extending the useful life of those devices could reduce both electronic waste and the carbon footprint associated with manufacturing new computing equipment.
Google specifically highlighted the role of "embodied carbon" — the emissions generated during the manufacturing process. By reusing existing hardware rather than building new systems from scratch, researchers hope to reduce the environmental impact of computing infrastructure.
Not every prompt needs a billion-dollar data center
To be clear, nobody is suggesting that clusters of old smartphones will replace the massive facilities used to train frontier AI models. Training systems like ChatGPT or Gemini requires enormous amounts of specialized hardware, including advanced GPUs and purpose-built AI accelerators.
But researchers believe repurposed smartphones could be useful for smaller workloads, educational research, edge computing projects and localized data processing.
Previous experiments from researchers at the University of Tartu demonstrated that old smartphones can be repurposed into tiny data centers for applications ranging from marine-life monitoring to collecting real-time transportation data. Those prototypes reportedly cost roughly 8 euros per device to assemble, making them an inexpensive alternative for certain computing tasks.
My take
What I find most interesting about this research isn't the technology itself. The tech industry often assumes that progress means constantly replacing older devices with newer ones. Yet the processors inside many discarded smartphones remain remarkably capable.
As AI drives demand for ever-larger data centers, researchers are exploring the opposite idea of simply using old hardware more efficiently. This research suggests far more of them might make a difference than we realize.
Follow Amanda Caswell and stay ahead of the AI curve
