A research project at the University of California San Diego is exploring whether thousands of retired smartphones can be transformed into a practical computing platform, potentially reducing both electronic waste and the need for newly manufactured server hardware.
Working with support from Google, the team plans to deploy a computing environment built from approximately 2,000 retired Pixel smartphones. Rather than using the devices in their original form, researchers remove components unnecessary for data center operations, including displays, batteries, cameras, speakers, and outer casings. The remaining motherboard, which contains the processor, memory, storage, and other computing elements, becomes the foundation of the computing platform.
The unusual approach addresses a major environmental problem associated with computing infrastructure. While the tech sector focuses on reducing the energy consumed by data centers, a substantial portion of technology’s environmental impact comes from manufacturing new hardware. Researchers believe extending the useful life of existing devices, even basic smartphones, could help reduce that burden.
Comparable to Servers
The project is based on a finding that surprised many observers: processor cores in relatively recent smartphones can deliver single-threaded benchmark performance comparable to, and in some cases higher than, individual cores in server processors.
Servers, of course, maintain major advantages in memory capacity, scalability, and throughput, but the research results suggest older phones remain capable of supporting a range of computing workloads.
To make the hardware suitable for cloud-style deployment, researchers replace Android with a general-purpose Linux OS. The switch eliminates mobile-specific software limitations and allows the phones to run standard infrastructure tools. Kubernetes is then used to coordinate workloads across large groups of devices, enabling clusters of smartphones to function as a unified computing platform.
Benchmark testing indicates that between 25 and 50 smartphones can provide computing performance comparable to a modern server CPU for certain applications. Researchers organize the devices into clusters of that size, allowing software workloads to be distributed across multiple phones.
The initial focus is on educational and research use cases rather than demanding commercial workloads. Early testing showed that a cluster of roughly 20 smartphones could handle peak assignment submission activity for a university course with more than 75 students. Still, researchers reported performance that was lower than the AWS-based backend previously used for the course.
At full scale, the planned 2,000-phone deployment is expected to support roughly one hundred similar courses simultaneously. The system will be used for computer science programs including systems programming and parallel computing, while also serving as a research platform for studying smartphone-based infrastructure.
Beyond cost savings, the project offers a chance to examine how consumer-grade hardware performs under continuous operation. Smartphones are designed for personal use rather than around-the-clock data center workloads. Researchers will evaluate reliability and maintenance needs as the system operates over longer periods.
The motherboard-focused approach is important because researchers estimate the motherboard accounts for roughly half of a smartphone’s manufacturing-related carbon footprint. Reusing the motherboard preserves a large portion of the embedded resources and emissions associated with the device.
This new use of smartphones is not meant to replace data centers or the specialized infrastructure used for AI training. Hyperscalers require dense, highly reliable hardware optimized for massive workloads. But researchers believe smartphone clusters may offer a practical alternative for educational institutions and organizations with limited budgets.
The full 2,000-device platform is expected to enter operation in 2026.
