Acurast, a decentralized network using everyday smartphones as secure compute nodes, has officially activated a 225,000-node smartphone compute network on Base. It’s a big development in bringing confidential onchain artificial intelligence (AI) into mainstream Web3.

The integration with Base, an Ethereum Layer-2 chain designed to make decentralized applications faster, cheaper, and more scalable, enables developers to run confidential AI workloads directly onchain using millions of smartphones worldwide. 

Instead of relying on centralized infrastructure, this network uses Trusted Execution Environments (TEEs) built into mobile devices to execute sensitive tasks securely, preserving user privacy and maintaining verifiability.

Smartphones are the new cloud

Acurast has set out to leverage the billions of already-deployed smartphones around the globe to create a decentralized compute layer. Whereas traditional cloud providers have centralized servers that carry risks of censorship and data exposure, Acurast’s model distributes workloads across devices in over 140 countries, all running confidential AI inference tasks within secure hardware enclaves. 

Jesse Pollak, Creator of Base, said:

“Base is about giving builders the best place to bring new ideas on-chain. Acurast is expanding that surface area by introducing decentralized, confidential compute powered by smartphones. That makes it possible for developers to run AI workloads on Base that are secure, verifiable, and not dependent on centralized infrastructure. This is the kind of infrastructure that helps move autonomous, real-world applications fully on-chain.”

The network just went live on Base’s mainnet, following its token generation event, and already handles production workloads securely.

Acurast’s founder, Alessandro De Carli, said:

“AI agents cannot rely solely on centralized servers if they are tasked with managing real assets onchain. By utilizing smartphone-based TEEs, we’re enabling confidential AI that is verifiable, decentralized, and owned by the users who power it.”

Confidential AI, native payments

A key part of this integration is the payment mechanism for compute. 

Acurast now supports native USDC payments on Base’s network without the need for bridging or offchain settlement layers. By embracing the x402 payment standard (originally developed to enable instant, HTTP-native stablecoin payments), AI agents can autonomously pay for compute resources in real time.

This opens the door for a pay-per-request model in decentralized services, where AI agents can automatically settle fees in USDC as they process tasks. It’s a crucial building block for autonomous Web3 applications that interact with APIs, data services, and complex onchain logic without intermediaries.

A new layer for onchain AI workloads

Developers leveraging Acurast on Base can onboard devices and manage compute infrastructure via the Acurast Hub with a Base wallet. 

Within the Hub, builders can deploy secure, autonomous AI agents, such as bots that execute trades, manage assets, or perform on-chain reconciliations. This happens while inputs and outputs remain encrypted and unseen by node operators.

All AI inference runs inside smartphone TEEs, meaning neither the device owner nor external observers can access confidential data during processing, key for privacy-focused applications in finance, identity, and enterprise workflows.

Beyond data centers

This move comes on the heels of strong growth for Acurast. Indeed, the decentralized compute network has expanded rapidly throughout 2025, moving from tens of thousands to hundreds of thousands of phones powering Web3 workloads. 

Acurast is pushing forward the development of large-scale confidential computing, pulling together decentralized physical infrastructure (DePIN), onchain AI, and real-time machine-native payments.

With its native token now trading on major exchanges and the global network running live production jobs, Acurast aims to lay the foundation for a new class of onchain applications that are decentralized, verifiable, confidential, and autonomous by design.

Disclaimer