Gonka AI

Gonka AI operates a decentralized computing network for AI training and inference, designed as an alternative to centralized cloud providers. The platform allows anyone, from individuals with a single GPU to large-scale data centers, to contribute their hardware's computing power to the network. In return, these hardware providers, or 'Hosts', are rewarded with the network's native token, GNK. The system is structured to redirect the vast majority of GPU power toward productive AI tasks like model training and scientific simulations, rather than consuming it for blockchain consensus mechanisms.

The company was founded by brothers David and Daniil Liberman, serial entrepreneurs who previously founded Product Science Inc., Kernel AR (acquired by Snap Inc.), and Frank Money. Their background in software performance engineering influenced Gonka's design, which emphasizes a seamless developer experience through an OpenAI-compatible API. Gonka officially launched on September 15, 2025.

The business model is built on a Layer 1 blockchain that uses a novel Transformer-based Proof-of-Work mechanism called Sprint Consensus. This system allocates voting weight based on verified computational work, ensuring that incentives are aligned with providing useful AI computation. For developers and researchers, Gonka provides permissionless access to this distributed GPU capacity, aiming to lower the cost of AI development and break the dependency on a few major cloud and hardware companies. The network's GNK token serves as the currency for accessing compute resources, with early adoption incentivized through a temporary zero-cost inference period and token distribution.

Gonka has secured significant funding, including an $18 million round in 2023 from investors like Coatue Management and Slow Ventures, and a subsequent $12 million strategic investment from Bitfury Group in November 2025.

Keywords: decentralized AI, GPU computing, Proof-of-Work, AI infrastructure, distributed network, blockchain, AI model training, AI inference, compute-to-earn, GNK token, Layer 1, DePIN, high-performance computing, community-owned AI, cloud alternative, David Liberman, Daniil Liberman, Product Science, Bitfury, Coatue Management