Decentralized Networks Eye Smartphones to Challenge AI Data Centers

Turning billions of smartphones into an AI computing layer

Decentralized AI networks are working to transform everyday smartphones into a global computing system. By using the processors and memory already in people's pockets, these projects hope to create an alternative to traditional, centralized data centers that power artificial intelligence.

Building massive data centers is expensive and requires significant amounts of water and energy. Using existing hardware from billions of unused or active mobile phones could lower these costs and provide a distributed way to handle digital tasks.

Key figures on mobile computing potential

• As of late 2026, there are approximately 7.68 billion mobile connections globally.
• The Acurast network has registered over 335,000 smartphones.
• Research indicates a cluster of 20 retired phones can handle the computing needs of dozens of users simultaneously.
• One network project plans to charge 0.5 USDT for specific internet connectivity tasks. USDT is a stablecoin, a type of crypto token tied to the value of the U.S. dollar.

How projects like Acurast and Phase work

Acurast co-founder Alessandro De Carli reports that about 40,000 phones are active on the network daily. These devices handle tasks like cybersecurity and hosting AI agents—automated programs that perform digital work. De Carli claims the total power of these active phones matches that of over 2,300 high-end servers.

Another project called Phase recently launched a service that uses smartphones to provide internet connectivity for AI. This helps AI agents avoid being blocked by websites that limit traffic coming from traditional data centers. Phase plans to eventually route complex tasks to large providers while keeping simpler tasks on the mobile network. The network rewards users for their participation and hardware uptime.

What has been proven in testing

Studies have confirmed that small groups of smartphones can perform tasks like image recognition. A Google-backed project at the University of California, San Diego, demonstrated that retired phones could grade student work faster than some standard cloud systems. Smartphones are also recognized for having built-in hardware security features that help protect sensitive data during processing.

Coordination and hardware challenges

While the hardware exists, making thousands of phones work together like a single reliable computer is difficult. In the real world, phones frequently disconnect, overheat, or run out of battery power. Their internet speeds vary, and they have less memory than the powerful servers used in traditional data centers.

Experts like Markus Levin from XYO suggest that these networks might be more effective at collecting and verifying real-world data rather than running heavy AI models. For these systems to succeed, they must solve practical problems that customers are willing to pay for, rather than just focusing on the number of connected devices.

Why decentralization offers an alternative to data centers

Using a decentralized network—a system where control is spread across many users instead of one company—could reduce the environmental and financial impact of AI. This approach avoids the massive costs associated with building new infrastructure. It also allows individuals to earn rewards by contributing their device's extra computing power to the network.

Sources

• Cointelegraph