The project, reported by hardware specialized publication Tom's Hardware on 15/6 and funded by Google, aims to extend the lifespan of old phones. These devices are typically replaced by Americans every four years.

For this initiative, a team of scientists from the University of California San Diego (UCSD) extracted motherboards from 2,000 2023 Pixel Fold phones. These were then assembled into "servers" for a data center. Each server comprises a cluster of 25 to 50 motherboards, managed by a common application. Other components such as screens, batteries, frames, and peripheral hardware like cameras were removed due to their incompatibility with a server environment.

Expected to be completed this autumn, the project offers a low-cost data center solution that reduces carbon emissions. This is achieved by decreasing raw material extraction during manufacturing. Motherboards account for over 50% of carbon emissions in the production phase. Reusing this component, instead of dismantling it to extract precious metals, will have a significant impact on both cost and the environment.

Electronic devices generally emit from two primary sources: emissions during use (electricity consumption) and emissions from hardware manufacturing. Due to their short lifespan, carbon emissions from phone manufacturing constitute a large proportion. For instance, a Pixel 10 used for three years emits 82 kg CO2e, according to Google, with 87% of that carbon released during production.

Unlike operational emissions, the carbon footprint from manufacturing presents a complex challenge. Meanwhile, the rapidly increasing consumption rate generates a large volume of electronic waste.

According to the Global E-waste Monitor report published by the United Nations in 2024, the world generated 62 million tons of e-waste, including 5 million tons of laptops and mobile phones, with only 22% collected. For phones specifically, 60-70% of users tend to store old devices in drawers, meaning they are neither reused nor recycled.

Users often replace phones when screens break, they desire better cameras, or they follow trends, even if the processor still functions well. However, according to Jennifer Switzer, a postdoctoral researcher at UCSD and one of the two lead researchers for the project, phone processing performance is increasingly powerful, even surpassing the common Intel Core-i3 processors found in servers.

The research team announced that the single-thread performance (the ability to process a single task at one time) of the phone-based "servers" is equivalent to or better than Asus's multi-core servers, which are designed for AI tasks and big data analysis.

The biggest difference between the two server types is size. High-end servers contain dozens of powerful multi-threaded processing cores and massive memory capacity. In contrast, phone-based servers have only a few processing cores and 8-12 GB of memory.

This data center, built from 2,000 phones, will support computer science specialized courses, such as Parallel Computation and Systems Programming. Initial tests show that even a server cluster comprising 20 phones can support a class of 75 students submitting assignments with better performance than an AWS server. A center with 2,000 phones will be capable of supporting hundreds of similar classes simultaneously.

Detailed information on the cost comparison between the two server types in this project has not yet been released. In previous research, Jennifer Switzer experimented using a server built from a Pixel 3A phone to support hotel booking operations, replacing an Amazon EC2 C5 virtual server. This resulted in a 40-fold reduction in investment and operational costs over the first three years. Comparing both production and operational emissions, using old phones instead of servers reduced emissions 10 to 19 times per query, due to significantly lower power consumption from the old phones.

According to the research team, these results provide a testing foundation for reusing old phones in large-scale computing.

Bao Bao (according to Tom's Hardware, Google, UCSD)