Lumino

Lumino is a decentralized compute protocol that connects users needing AI model training with idle GPU resources. Its core business is a marketplace where GPU providers (data centers or individuals) rent out computing power, while AI developers access it through a simple SDK to train or fine-tune open-source models.

Market pain point: AI training requires massive, expensive GPU clusters, but traditional cloud services are costly, scarce, and often controlled by a few large companies. Meanwhile, many GPUs sit idle.

Solution and edge: By pooling idle GPUs, Lumino claims to cut training costs by up to 80%. The platform offers pre-built templates, real-time monitoring of GPU/CPU/memory, and strong data privacy controls. Users pay only for what they use, and GPU owners can earn income. It supports NVIDIA (H100, A100, V100) and AMD GPUs.

Recent development (Feb–Aug 2026): Lumino launched its devnet in late 2024, allowing fine-tuning of models like Llama 3.1 8B on decentralized infrastructure. In early 2025, the team participated in community events (e.g., Unchained Summit) and podcasts discussing decentralized AI. No major negative events were reported in the past two years, though the project remains early-stage with limited public adoption metrics.