When your phone is idle, its screen may be off, but the chip is still ticking away. Most people only use their phones to scroll through short videos and reply to messages, leaving most of that computing power wasted. As AI moves increasingly to the edge, many tasks—like making decisions, classifying data, and running scheduled jobs—don’t need to keep calling on a data center. Phones can handle some of that work themselves. Acurast connects these scattered phones around the world into a network that can actually be put to work, rather than building yet another data center.
Once you install Processor, your phone becomes a compute node in the network. When developers submit tasks, the system matches them to devices based on factors like CPU, memory, and storage—not just whether they’re sitting idle. More importantly, it uses a trusted execution environment (TEE) to protect the computation, supporting both privacy and verification afterward. That means it can do more than run ordinary scripts: it can handle lightweight AI inference, automation tasks, and use cases involving more sensitive data. People are already running the open-source System One decision-making model on it for quick tasks like game decisions and email classification. Latency can be kept under one second, with everything running on real phone CPUs, and the blockchain can verify which device did the work.
For people building Agents, the hassle often isn’t writing code—it’s having to manually find computing power, pay for it, and deploy it afterward. Acurast connects these steps: an Agent can request a quote, then pay with USDC on Base through x402. Once payment is confirmed, the task is submitted directly to the network and a task hash is returned. Set the budget and permissions in advance, and the program can handle the rest of the process on its own. Some projects have already integrated this workflow, with payment and execution records corresponding to each other, so the books actually reconcile.
It also addresses the finer details. Previously, when using phones to scrape webpages or take screenshots, differences in screen sizes across models could throw off the layout, making it a pain to check everything. The latest update adds controls for rendering dimensions, letting you preset the width, height, and pixel density so different phones render to roughly the same specifications and cause fewer device-related headaches. It sounds like a minor change, but it really does make things easier for automation tasks that need to collect web data in bulk.
The network is now connected to more than 300,000 phones across over 100 countries, with on-chain transaction volume in the billions and the number of deployments continuing to rise. These figures show that the supply side and orchestration layer are expanding, but what will really determine whether it can go the distance is whether developers are willing to keep paying to use this computing power. $ACU is more than a token for participation rewards: deployment-related fees are consumed, and the token is also tied to staking and governance. Inflation has mechanisms to offset it, but in the end, it all comes down to whether real-world usage can keep pace.
The shift toward lightweight, edge-based AI is a broad trend. Phones are already distributed around the world, offer hardware-level isolation, and use less energy than traditional servers. Acurast is positioned right here, organizing devices that already exist so Agents and developers can call on them as needed. What’s worth watching next is how many of these deployments are actually running real workloads and generating repeat orders, rather than simply piling up in number. If idle old phones can reliably handle these tasks, those devices in the drawer won’t just be taking up space anymore.
