There’s a rather surreal phenomenon in the AI industry right now. On one side, tech giants are pouring massive amounts of money into building data centers and hoarding GPUs—compute power is becoming increasingly expensive, with companies racing to secure computing resources for years ahead. On the other side? Countless smartphones around the world, with perfectly capable performance, sit unused and gather dust in drawers. Recently, while exploring Acurast, I discovered it’s precisely targeting this overlooked resource. The platform currently reports over 336,000 connected smartphones, 1.1 billion on-chain transactions, and 687,000 compute deployments. Previously, I thought DePIN simply meant people contributing hardware in exchange for tokens—but Acurast has made me reconsider: does AI compute really need to be concentrated entirely in massive data centers? ① Smartphone compute power may have been vastly underestimated Modern smartphones boast capable CPUs, ample memory, and secure silicon chips—especially those older models still functioning well after being retired. Acurast’s approach is to connect these dispersed devices into a global, decentralized computing network. Developers can deploy APIs, automation tasks, and AI inference workloads onto this network, where qualified devices execute them. Of course, smartphones can’t replace high-end GPU clusters for training large models—but for lightweight inference and distributed tasks, this compute potential is absolutely viable. ② What truly fascinates me is TEE-based secure computation How can anyone confidently entrust their business data to tens of thousands of strangers’ phones? Acurast leverages smartphones’ built-in TEE (Trusted Execution Environment) to run compute tasks within hardware-protected zones and provides verifiable execution proofs. This is critical—especially as AI agents begin handling wallet operations, automated transactions, and sensitive data processing. Developers will increasingly demand secure, private execution environments. ③ Real-world AI applications are already emerging This year, Acurast integrated with Base, enabling developers to build AI agents, run confidential inference, and automate on-chain workflows using decentralized smartphone compute. It has also partnered with Pocket Network and NodeGhost to explore fully decentralized AI gateways—early tests successfully used phone nodes to relay AI inference and web search requests. These efforts are still far from mass commercialization, but they clearly point toward tangible use cases. ④ Where’s the real opportunity in AI × DePIN? What excites me most is the possibility that future compute supply won’t rely solely on a few large providers. Acurast’s Lite mode lets users contribute idle smartphone compute time; Core mode transforms eligible Android devices into dedicated compute nodes. Developers gain more flexible compute options; device owners can earn ACU rewards for providing resources. Of course, whether rewards offset electricity costs and device wear ultimately depends on individual calculations. I’m now increasingly excited about Acurast’s future trajectory. 336,000 phones are just a starting point—the real question is: how many AI developers will consistently use and pay for this compute? If this path succeeds, those old phones we casually toss into drawers might truly become part of AI’s foundational infrastructure. @Acurast