A new project at UC San Diego is transforming discarded smartphones into a massive, energy-efficient computing platform that could reshape how universities and organizations approach cloud computing infrastructure. Researchers supported by Google are building a datacenter from 2,000 Pixel smartphones, extracting the motherboards from retired devices and clustering them together to provide low-cost, low-carbon computing power. The initiative directly addresses embodied carbon, the emissions locked into hardware manufacturing, a challenge that traditional energy-efficiency efforts often overlook.

Why Does the Carbon Footprint of Old Phones Matter?

Computing's environmental impact comes from two major sources: operational carbon, which reflects energy consumed during use, and embodied carbon, which encompasses emissions from manufacturing hardware. While companies have made progress reducing operational carbon through cleaner energy and efficiency improvements, the manufacturing footprint remains a stubborn problem. When people replace their phones every four years on average, most discarded devices still have fully functional compute capabilities intact.

The motherboard alone accounts for approximately 50 percent of a smartphone's embodied carbon footprint, making it the most impactful component to repurpose. By extracting and redeploying these motherboards instead of manufacturing new servers, the project avoids the raw material extraction and production emissions associated with building fresh hardware from scratch.

How Does a Smartphone Stack Up Against a Traditional Server?

On paper, smartphones and servers seem like mismatched tools. A modern smartphone has a handful of heterogeneous processor cores and 8 to 12 gigabytes of memory, while a server contains dozens of powerful multithreaded cores and vastly more storage capacity. However, when researchers benchmarked the single-threaded performance of a 2023 Pixel Fold against a standard data center server using industry-standard testing, the smartphone's performance cores actually outperformed the server on most benchmarks.

The real difference lies in scale and specialization. A single smartphone cannot match a server's total computing power, but 25 to 50 smartphones clustered together can deliver equivalent performance to a modern server. The challenge, then, becomes identifying applications that fit within a smartphone's memory and processing constraints, and orchestrating work across dozens of devices simultaneously.

How to Deploy Smartphones as Cloud Computing Infrastructure

• Hardware Preparation: Remove all non-essential components like displays, batteries, cameras, and chassis from each smartphone, leaving only the motherboard containing the core compute functionality. This step is critical because batteries and other consumer-grade components are not rated for sustained datacenter environments.
• Operating System Conversion: Replace the mobile-oriented Android userspace with a general-purpose Linux distribution. This switch enables broader programmability and disables consumer protections like the "low memory killer" daemon that throttles memory-hungry applications, which are unnecessary in a cloud computing context.
• Cluster Orchestration: Use containerized applications managed by Kubernetes to coordinate jobs across the large number of devices. Smartphones are organized into self-managing clusters of 25 to 50 devices, allowing the system to distribute workloads and manage failures automatically.

What Applications Can Run on a Smartphone Cluster?

The UC San Diego deployment will initially support computer science classes such as Parallel Computation and Systems Programming. Early experiments with a 20-smartphone cluster demonstrated that even this modest setup could handle peak submission rates for a 75-plus student class, with grading latencies below those of typical cloud backends like AWS's t3.micro instances.

The full 2,000-phone deployment, expected to launch in fall 2026, will be capable of supporting roughly 100 such classes simultaneously. Beyond education, the project will serve as a testbed for smartphone-based computing at scale, investigating the reliability of consumer-grade hardware under sustained use and exploring which types of workloads are best suited to this architecture.

"The carbon footprint of computing is a key sustainability challenge. It is driven by two major sources: operational carbon reflects emissions from energy consumed during use, and embodied carbon encompasses emissions associated with hardware manufacturing," explained Jennifer Switzer, Visiting Postdoctoral Researcher at Google.

Jennifer Switzer, Visiting Postdoctoral Researcher, Google

What Makes This Approach Different From Other Green Computing Efforts?

Most sustainability initiatives in tech focus on reducing operational carbon through renewable energy and efficiency improvements. This project takes a different angle by targeting embodied carbon, the emissions baked into hardware before it ever powers up. By giving second life to devices that would otherwise be discarded or recycled, the approach avoids the need for newly manufactured hardware entirely.

The initiative also demonstrates that consumer-grade hardware, when properly refurbished and orchestrated, can deliver competitive performance for many real-world applications. This finding challenges the assumption that cloud computing requires purpose-built, energy-intensive server farms. For universities and organizations running educational workloads, grading systems, and other latency-tolerant applications, a smartphone cluster could provide significant cost and environmental savings.

The project is expected to provide 50 server-equivalents worth of computing power at a fraction of the usual cost, while simultaneously reducing the environmental footprint of cloud infrastructure. As the deployment scales and researchers gather data on reliability and performance, the model could influence how institutions approach computing infrastructure procurement and sustainability planning.