Could Smartphones Become the Unexpected Foundation for AI's Future?
Projects exploring decentralized AI infrastructure aim to harness ordinary mobile devices as alternatives to massive data centers — though critics remain unconvinced.
The enormous data facilities supporting artificial intelligence development have earned a reputation comparable to asbestos-laden children's playthings.
Activists across the globe point to the tremendous quantities of electricity and water these facilities consume, while financial analysts question whether revenues can justify the $38 billion price tag of constructing a single one-gigawatt AI data center.
Yet what if an alternative solution for driving AI advancement is already nestled in your pocket?
By October 2026, mobile connections globally reached approximately 7.68 billion, with countless smartphones currently sitting idle in various locations — tucked away in pockets, resting on work surfaces, or forgotten in storage.
Every device contains processing capabilities, storage capacity and network access, collectively representing an enormous reservoir of computational resources awaiting deployment.
The question emerges: might networks composed of ordinary smartphones feasibly drive artificial intelligence at meaningful scale?
Putting smartphones to work
New decentralized AI platforms are appearing with ambitions to transform common mobile devices into distributed computational infrastructure, decreasing dependence on expensive, centralized data facilities.
According to Acurast, over 335,000 smartphones have joined its platform. Alessandro De Carli, the company's co-founder, indicates that approximately 40,000 devices maintain full activity on typical days, delivering computational capacity for applications spanning from data extraction and security operations to AI inference and hosting agents.
According to De Carli, mobile devices demonstrate particular effectiveness for specific computing demands. Their integrated hardware security capabilities "excel" in domains such as confidential computing or providing an "isolated environment" capable of installing and executing programs. They also prove well-matched to tasks that are "CPU intensive," or demand broad distribution and "unfiltered access" to online resources.
"Servers are better at multi-threading, but smartphones offer unmatched compute efficiency," De Carli tells Magazine.
Although standard smartphones possess constrained memory, and large servers maintain superiority when managing concurrent tasks in parallel, numerous computing jobs can still operate within available smartphone memory, he explains.
Collectively, the engaged devices across the platform "match the compute capacity of about 2,340 high-end data-center servers," according to his assessment, measured against the ASUS RS720A-E11, a high-density server built for artificial intelligence and high-performance computing (HPC).
A separate initiative, Phase, introduced Phase Proxy during October, enabling smartphone users to supply internet connectivity and blockchain validation for AI agents. According to founder Kirill Eves, "websites routinely block or rate-limit automated traffic originating from data centers," prompting the system to utilize users' mobile and residential connections as a proxy instead. Phase plans to charge 0.5 USDT per 1,000 requests and 1GB of traffic.
The subsequent development stage is Phase Router which will direct simpler requests to models running on phones in the network, or route them instead to major external providers like OpenAI and Anthropic, based on "task complexity, model capabilities, performance and cost."
Beyond smartphones, the network might ultimately incorporate desktops, servers and GPUs, according to Eves.
However, Markus Levin, co-founder of decentralized infrastructure provider XYO, suggests decentralized networks might prove more suitable for gathering and validating real-world data rather than running AI models.
Mobile devices and sensors can supply localized information absent from centralized datasets, according to Levin, though only when the data maintains accuracy and resists manipulation.
"A useful decentralized AI network should be judged by whether it solves a problem customers will pay to fix, not by its token price or node count," he says.
Coordination is the tricky part
Among the most significant obstacles facing mobile phone networks is orchestrating the devices to deliver consistent hardware, power and connectivity, matching what data centers provide.
Research conducted in 2025 examined small smartphone groups performing image recognition tasks and discovered that phone clusters could achieve competitive performance under specific circumstances, though the findings emerged from a controlled testing environment.
A Google-backed project at the University of California, San Diego found that a cluster of 20 retired smartphones could handle the peak computing workload of a class of more than 75 students, with grading times faster than the university's standard cloud setup. However, those devices were also organized in a controlled environment, not scattered across the world.
Under actual operating conditions, a distributed network must manage phones that lose connection, experience overheating, exhaust battery power, switch between networks, and possess different levels of processing capability and storage.
In a blockchain network like Ethereum for example, the capacity is currently limited by how long it takes to communicate across the network, and the processing abilities of the lowest spec devices.
Likewise, when computing tasks get divided among multiple phones, sluggish connections or disconnected devices can throw a spanner in the works. Distributing separate, independent tasks across available phones is generally more straightforward.
Can the economics work?
Regardless of whether technical obstacles can be addressed, networks must demonstrate economic viability for both customers and device owners alike.
Leveraging existing smartphone hardware avoids the substantial capital expenditures associated with data center construction, yet expenses encompass elements such as connectivity, coordination and incentives for people on the network.
Regarding compensation, De Carli explains there is "no fixed figure" because rewards for Acurast participants vary by "device uptime, hardware specifications (benchmark performance), total ACU staked and active compute demand" from developers. He says operators compete for a pool of inflation-funded rewards, rather than receiving direct payments from customers.
Phase network rewards depend on uptime, participation and customer demand.
"Phase will use a portion of its service revenue to purchase PHASE on the open market and distribute it to node operators based on their contributions to the network," Eves says.
Getting an AI model to run properly on a phone is no longer the main challenge — getting the economics to stack up while coordinating tens of thousands of phones to behave like a reliable cloud service is the real trick.