The global push to monetise idle computing power is gaining momentum. FAR Labs has opened node registrations for FAR AI, its distributed inference network, following the completion of core development and the start of a closed testing phase on March 30.
The company estimates more than 3 billion GPUs currently sit unused across consumer and enterprise environments. FAR AI is designed to bring a portion of that dormant hardware online for AI inference workloads, offering GPU owners a way to earn income from machines that would otherwise sit idle.
What is FAR AI and How Does It Work?
At its core, FAR AI is a distributed network built on Dizzaract technology. It connects GPUs from individual owners and enterprise operators into a unified infrastructure capable of handling inference requests at scale. The platform intelligently routes each workload to the node best suited to handle it, balancing performance and reliability across the network.
Operators who register early will receive priority onboarding access. Developers are expected to gain API access in the near future.
Ilman Shazhaev, founder and CEO of Dizzaract, said the platform was built with accessibility in mind. “Useful compute already exists in places far beyond traditional data centers. FAR AI brings that capacity online in a way that is practical for operators and immediately usable for developers.”
Opening Up the Supply Side
Registration is handled through the FAR AI website, where prospective operators can specify their GPU model and view estimated monthly earnings. The platform is compatible with most GPU types, which the company says should support a broad and diverse supply base as the network scales.
Rather than competing with hyperscale data centres directly, FAR AI positions itself as complementary infrastructure, drawing on hardware that already exists rather than requiring new capital investment.
The Technical Challenge: Memory and Scale
Inference, the process by which a trained AI model produces outputs in response to inputs, is FAR AI’s primary focus. One of the more technically significant features of the platform is its ability to pool multiple nodes together when a workload requires more memory than a single machine can provide. This makes the network viable for heavier inference tasks that would otherwise be out of reach for consumer-grade hardware.
Security is handled through isolated execution environments and encrypted communication channels. A proof-of-compute mechanism verifies that workloads have been processed as claimed, keeping the network accountable at scale.
What Comes Next
Closed testing is currently underway with a selected group of partners. FAR Labs says the focus at this stage is on stress-testing orchestration and refining developer workflows before the broader rollout begins.
Public API access for developers is being developed. The longer-term ambition is to establish FAR AI as a reliable infrastructure option for startups and software teams that need distributed GPU compute without the overhead of building and managing their own network. Closed testing is currently underway with a selected group of partners. FAR Labs says the focus at this stage is on stress-testing orchestration and refining developer workflows before the broader rollout begins.
