As AI models demand ever‑greater compute, developers are exploring alternatives to traditional data centers. Two projects highlighted by Cointelegraph—Acurast and Phase—are building decentralized networks that enlist idle smartphones to provide distributed processing power for tasks such as web scraping, cybersecurity, and AI inference.

Acurast’s growing phone pool

According to Cointelegraph, Acurast has onboarded more than 335,000 smartphones, with roughly 40,000 devices actively contributing compute each day. Co‑founder Alessandro De Carli claims that the active pool matches the capacity of about 2,340 high‑end data‑center servers when benchmarked against the ASUS RS720A‑E11 AI server. The network focuses on workloads that benefit from smartphones’ built‑in hardware security and CPU‑intensive tasks that require wide geographic distribution.

Phase’s mobile proxy and pricing model

Phase launched Phase Proxy in October 2026, allowing phone owners to provide internet connectivity and blockchain validation for AI agents. Founder Kirill Eves says the service helps bypass data‑center rate‑limiting by routing traffic through residential mobile connections. Phase charges 0.5 USDT per 1,000 requests and 1 GB of traffic, and plans to introduce Phase Router, which will dynamically direct simple requests to phone‑based models or to external providers such as OpenAI and Anthropic based on cost and performance.

Technical and economic challenges

Industry observers caution that coordinating a global fleet of phones presents hurdles not seen in centralized facilities. A 2025 study cited by Cointelegraph showed that small, controlled clusters of smartphones could match cloud performance on image‑recognition tasks, but real‑world deployments must contend with battery drain, network variability, and heterogeneous hardware. Markus Levin of XYO argues that decentralized networks may excel at gathering real‑world data rather than running large AI models, emphasizing that any viable solution must solve a paid problem rather than rely on token speculation. Acurast’s compensation scheme, which varies rewards based on device uptime, hardware specs, and demand, reflects the difficulty of creating a financially sustainable model.

What to watch next

Future coverage will track the scalability of Acurast’s and Phase’s networks as they expand beyond smartphones to desktops and GPUs, monitor adoption by AI developers seeking cost‑effective compute, and assess regulatory scrutiny around decentralized processing and data privacy.