CapaCloud is a peer‑to‑peer marketplace that lets GPU owners rent out their hardware to AI, rendering, and research users via Solana smart contracts. Users pay hourly in USDT, receive isolated Docker containers with root SSH access, and benefit from load‑balanced, carbon‑neutral compute without intermediary cloud providers.
Funding
Funding not disclosed
Founders
Product
Problem
Access to high‑performance GPU compute is constrained by centralized cloud providers that require long‑term contracts, incur high hourly rates, and operate energy‑intensive data centers, limiting flexibility for AI developers, rendering studios, and researchers.
Solution
CapaCloud offers a peer‑to‑peer neocloud platform that connects GPU owners with compute users through Solana‑based smart contracts. Users rent GPUs on a pay‑per‑use basis, paying hourly rates in USDT via wallet authentication, eliminating intermediaries and billing overhead. Each rental provisions an isolated Docker container with SSH root access, allowing full installation of drivers, frameworks (e.g., PyTorch, TensorFlow) and custom tooling. The marketplace leverages existing consumer GPUs distributed across a decentralized network, delivering a carbon‑neutral alternative to traditional data‑center clouds while preserving privacy—no personal data is stored and all sessions are automatically cleared after use. Real‑time load balancing and monitoring ensure reliable performance across the distributed fleet.
Target Audience
Primary customers are AI/ML engineers, data scientists, and development teams requiring on‑demand GPU power, as well as creative studios and research groups that run rendering, simulation, or high‑frequency trading workloads.
Features
- Solana blockchain smart contracts enforce trustless, usage‑based billing in USDT with automated escrow and immutable transaction logs.
- Wallet‑based authentication (Phantom, Solflare, etc.) provides password‑less login and per‑rental SSH key generation.
- Isolated Docker containers with network namespaces and resource quotas guarantee security and prevent cross‑tenant interference.
- Full root access via SSH enables installation of any GPU driver, AI/ML framework, rendering engine, or mining software.
- Distributed load balancer and scheduler allocate workloads across a global pool of peer‑hosted GPUs, optimizing latency and utilization.
- End‑to‑end TLS encryption and automatic data deletion ensure privacy‑first compute with no persistent user data.
- Carbon‑neutral compute model reuses existing consumer GPUs, achieving up to 99 % lower carbon footprint compared to conventional cloud GPU services.
- Web‑based dashboard and API expose real‑time monitoring metrics, usage reports, and integration hooks for CI/CD pipelines.