A low-carbon computing platform from your retired phones

Original Article ↗ Hacker News Discussion ↗

Google-backed research into turning retired smartphones—especially Pixel devices—into low‑carbon data center nodes has reignited interest in reusing consumer hardware as clustered compute. Commenters like the environmental and efficiency angle but flag major obstacles: locked bootloaders and proprietary firmware that limit secure long‑term reuse, unclear lifecycle carbon benefits versus purpose‑built servers, storage and reliability concerns, and the labor cost of harvesting phone motherboards at scale. Many argue that mandatory unlockable bootloaders or time‑limited source releases for firmware would be needed to make this kind of “junkyard computing” broadly viable beyond research and hobby projects.

Overall Reception

  • Many are excited about reusing old smartphones as cluster nodes, likening it to Raspberry Pi or PS3 clusters and home lab projects.
  • Others see it mainly as an interesting research/demo rather than something likely to become commercially mainstream.

Locked Bootloaders & Firmware Blobs

  • A major theme: retired phones are often e‑waste because of locked bootloaders, proprietary blobs, and short OEM support.
  • Even if the OS is replaced, low‑level firmware (baseband, Wi‑Fi, Bluetooth) often remains unpatchable, especially pre‑Treble devices, raising long‑term security concerns.
  • Some argue these blobs don’t matter for a pure compute node if radios are disabled and only USB/ethernet is used; others insist outdated firmware still poses risk in production settings.

Security & Networking Concerns

  • Several posters argue old phones should not be on internet‑exposed networks due to unpatched vulnerabilities.
  • Counterpoint: for trusted, non‑exposed workloads (e.g., local compute/NAS) risk may be acceptable if there is no kernel‑level RCE.
  • How data and power are plumbed (USB‑C, internal buses, carrier boards) is noted as unclear.

Practicality & Economics

  • Questions about cost: disassembly, testing, custom racking, and ongoing heterogeneity may outweigh benefits versus buying conventional servers.
  • Some think it could work if paired with higher‑value services (e.g., gaming, LLM harnesses) or as a hedge against future hardware shortages.
  • Concerns about storage endurance (eMMC/flash wear) and battery issues (“spicy pillows”) are raised; others report very long lifetimes in practice.

Environmental Impact

  • Debate over whether this is a real carbon win.
  • One view: embodied emissions in manufacturing dominate; extending device life is highly beneficial, possibly needing ~25‑year lifetimes to beat frequent replacement.
  • Another view: the key question is whether retrofitting existing phones has lower carbon cost than manufacturing new datacenter‑class hardware; thread deems this unresolved.

Regulation, Openness & Prior Art

  • Strong support for regulations forcing unlockable bootloaders and eventual source release (e.g., after 7–20 years).
  • Frustration that mobile platforms, especially non‑Pixel Android and iOS, remain locked down despite such reuse potential.
  • Multiple references to existing efforts: PostmarketOS, Termux servers, prior “junkyard computing” and phone‑cluster research, decentralized compute networks, and laptop‑motherboard clusters.