📡 Breaking news
Analyzing latest trends...
AI Text-to-Speech.

Google & UCSD Build Low-Carbon Datacenters Using Old Smartphone Motherboards.

Google & UCSD Build Low-Carbon Datacenters Using Old Smartphone Motherboards.
Turning Electronic Waste into Cloud Infrastructure: Google and UC San Diego Build Server Clusters from Retired Smartphones

In a creative bid to tackle the tech industry's mounting electronic waste crisis, Google Research has partnered with engineers at the University of California San Diego (UCSD) to pioneer an eco-friendly computing paradigm: "Phone Cluster Computing." The collaborative project demonstrates how hundreds of decommissioned consumer smartphones can be extracted from junk drawers and wired together into fully functional, low-carbon cloud data centers.

The Architecture: Stripping Down to Raw Silicon

To transform a consumer handheld device into enterprise-grade infrastructure, researchers had to ruthlessly eliminate components unnecessary for a server environment.

  • Hardware Decoupling: The team stripped the smartphones down entirely, discarding the screens, batteries, cameras, and speakers. Only the core motherboards which contain the system-on-chip (SoC), memory, and integrated storage were kept and linked together.

  • Operating System & Orchestration Migration: Because Android's built-in background task-killers are unsuited for continuous enterprise workloads, the engineers wiped the mobile OS, replaced it with a general-purpose Linux distribution, and deployed Kubernetes to manage and distribute containerized computing tasks seamlessly across the hardware array.

Performance Benchmarks: Single-Core Muscle vs. Multi-Threaded Scale

The underlying motivation for the project relies on a surprising architectural reality: the single-threaded performance of a three-year-old mobile processor core is often on-par with, or even superior to, a single core inside a massive, power-hungry datacenter server CPU.

However, mobile chips fall short when handling massive multi-threaded tasks due to their limited core counts and lower overall RAM capacities (8GB to 12GB per phone).

MetricSmartphone Cluster NodeTraditional Enterprise Server
Single-Core Architecture🚀 High single-threaded efficiencyStandard per-core balance
Multi-Threaded Workloads⚠️ Limited by mobile core counts🏆 Massive scalability (Dozens of cores)
System Equivalence25 to 50 combined motherboards1 modern dual-socket server CPU

Despite these limitations, the prototype clusters are proving highly capable for lightweight, decentralized tasks. In localized campus field tests:

  • A modest cluster of 20 recycled phones successfully hosted Jupyter Notebooks and web applications for a university classroom.

  • The system supported 75+ students simultaneously submitting workloads, delivering grading latencies that actually clocked in below a default AWS cloud backend.

The term "Embodied Carbon," when we talk about "eco-friendly data centers," often focuses only on operational carbon, or finding clean energy to run the servers. However, the most alarming carbon emissions originate from the "mining of rare earth minerals and the chip manufacturing process" even before the devices are put into use. Consumer behavior of replacing mobile phones every 3-4 years, even when the chips aren't broken, represents a massive waste of resources. Recycling old smartphone motherboards into servers therefore cuts through the new hardware production cycle, truly saving carbon at the source.

This project hints at a major truth in the IT infrastructure industry: many cloud clients today don't need supercomputer-level processing power on a single machine. Instead, they simply need many small CPUs and just enough RAM to run short tasks (microservices). Using Kubernetes to manage old mobile phone motherboards proves that modern software can enable "cheap second-hand hardware" to seamlessly integrate and function like a high-end server.

This innovation is about to take another step forward. Because in Fall 2026, a team of researchers from UC San Diego is preparing to launch a large cluster consisting of up to 2,000 Google Pixel motherboards to run high-performance computing tasks, such as parallel processing physics simulators. If successful, it will become a model for universities and organizations worldwide to build their own "private cloud labs" at a low cost and in an environmentally friendly way.

 

 

Why curl is Banning Public Security Reports for a Month. 

 

Source: Google Research 

💬 AI Content Assistant

Ask me anything about this article. No data is stored for your question.

Comments

Popular posts from this blog

Moonshot AI Reportedly Secures 20,000 NVIDIA Blackwell GPUs via Alibaba to Scale Kimi K3.

Google Pulls Google Earth Nano Banana AI Feature Over Misinformation and War Disinformation Risks.

WhatsApp Rollouts Native Web Calling: Host Group Voice and Video Calls Directly in Your Browser.

Xbox Disc2Digital Feature Imminent Leaked Docs Reveal Disc-to-Digital Conversion Rollout This Month.

Apple Rumor RAM Squeezes MacBook Air Supply While Future Smart Glasses Target Health Tracking.

DeepSeek Releases DeepSeek-V4-Flash-0731 Matches Claude Sonnet 5 Benchmarks at a Fraction of the Cost.

Xbox CEO Asha Sharma Outlines Turnaround Roadmap Amid Q2 2026 Gaming Revenue Slide.