Can Your Old Smartphone Become a Low‑Cost Edge Data Center?

Background

In the era of "AI Everywhere," computing power is a key resource, but large data centers are not suitable for every scenario. A recent study from the University of Tartu proposes repurposing old smartphones as "micro data centers" to provide distributed compute capability at minimal cost.

Experimental Findings

Even ten‑year‑old phones can efficiently perform image‑recognition and sensor‑data tasks. The overall retrofit cost is only about 8 EUR (≈ 66 CNY), far cheaper than a modern Raspberry Pi board.

Why Use Old Phones?

Smartphones such as the 2013 Google Nexus 5 feature quad‑core Snapdragon 800 CPUs, 2 GB RAM, and robust wireless interfaces, making them comparable to many IoT devices. Their design for high‑energy scenarios gives them good thermal performance and low power consumption.

Conversion Process

Remove the original battery and supply power externally via a regulated module.

Replace Android with a lightweight Linux‑based OS (e.g., PostmarketOS) to gain full hardware control.

Connect multiple phones in a master‑worker cluster: one device acts as the master, distributing tasks to three workers for parallel processing.

3D‑print a protective frame to hold the phones together.

Demonstration

The prototype was deployed in a waterproof housing at a 25‑meter depth off Madeira Island, where it performed continuous, fault‑free image‑recognition of marine life for eight hours.

Potential applications include city‑center foot‑traffic monitoring, edge processing for drones, and low‑cost compute platforms for startups needing data analysis and web hosting.

Limitations

Not all phones are suitable; models with modular hardware (e.g., Nexus series) are easier to convert. Newer phones tend toward sealed designs, making disassembly and OS replacement more difficult and costly.

Environmental Impact

With billions of phones discarded annually, repurposing them as reusable compute resources can mitigate e‑waste while providing affordable edge computing.