The rapid advancement of artificial intelligence (AI) is leading to an unprecedented demand for computing power, yet many industry players are grappling with significant structural challenges in their underlying infrastructure. Centralized cloud providers hold a monopoly and impose high costs, while existing decentralized networks often exhibit considerable resource waste. This situation has created critical bottlenecks that hinder industry development. In response to these challenges, Gonka has emerged, aiming to build a truly efficient decentralized AI computing network.
Gonka’s vision is clear: to ensure that every unit of computing power is allocated to meaningful AI tasks rather than being squandered on pointless hashing or staking activities. Within this ecosystem, hosts can generate revenue by contributing their GPU computing power; developers can escape the limitations posed by centralized cloud services and proprietary large language models (LLMs); and users will help shape network demand through AI inference requests they initiate.
A core innovation of Gonka is its fundamental rethinking of the traditional proof-of-work (PoW) consensus mechanism. While traditional PoW, exemplified by Bitcoin, has historically been viewed as essential for network security, it typically consumes computing resources solely for that purpose, leading to inefficient resource utilization. Gonka’s “Sprint Consensus Protocol” aims to remedy this issue. During a designated “sprint” phase, hosts showcase their computational power by executing a Transformer model containing 20 billion parameters. The voting weight in this model is proportionate to the computational effort exerted, thus ensuring fairness under the principle of “one unit of computing power, one vote.”
This innovative approach sets Gonka apart from competing decentralized computing networks. For instance, in contrast to Bittensor—which allocates up to 60% of its rewards to stakers who do not contribute to network hashrate—Gonka’s model is designed to direct nearly all rewards to those who actively contribute computing power. This approach effectively eliminates the misallocation of resources that has plagued other models. Similarly, Akash Network, which distributes rewards mainly to stakers, fails to effectively subsidize AI computing infrastructure, creating further inefficiencies.
By adopting the economic incentives of Bitcoin’s PoW system, Gonka is able to slash AI computing costs by upwards of 90% through block subsidies. This model continuously encourages the GPU hardware ecosystem to strive for improved efficiency and performance. Additionally, Gonka incorporates probabilistic verification along with a reputation system, successfully minimizing the verification overload to a mere 1-10%, while still maintaining robust network security and eliminating unnecessary computing power waste.
Behind the Gonka project is a formidable team and industry ecosystem, significantly increasing its credibility. Incubated by Product Science Inc.—a US-based AI developer co-founded by the Liberman siblings, who are seasoned Web 2 industry veterans and former core product directors at Snap Inc.—Gonka is drawing upon a wealth of experience. Since its inception in 2021, Product Science has been a trailblazer in the AI sphere, serving clients including prominent names like Walmart, JPMorgan Chase, and Airbnb. In 2022, the company successfully secured $18 million in funding from a prestigious group of investors, including a16z founders Marc Andreessen and Chris Dixon, as well as figures like Arielle Zuckerberg and Notion founder Akshay Kothary.
The project has also enlisted early contributors from notable Web 2 and Web 3 domains, such as 6 Blocks, Hard Yaka, Gcore, and Bitfury. Enhancing its strategic development is an influential advisory team that features figures like Greg Kidd of Hard Yaka, Patrick Dai of Qtum, and Val Vavilov of Bitfury.
Gonka’s native token, GNK, represents a pivotal asset in the future of AI computing power. With a fixed total supply of one billion tokens and a Bitcoin-like issuance schedule that includes halvings approximately every four years, this scarcity-driven model aims to foster long-term sustainable value for the network. Adhering to the principles of a fair launch, GNK tokens are exclusively obtainable by contributing GPU computing power to the network or purchasing from the community pool. The Gonka team emphasizes that the direction of the network will be determined entirely by its community.
Currently, the Gonka testnet is live and expanding, with the total connected GPU computing power growing at a rate of 50% monthly. Additionally, the project has successfully passed a Certik audit, further solidifying its reliability and potential for transforming decentralized computing. With its unique model, Gonka aims to redefine the future of decentralized computing and provide a robust infrastructure for the next major leap forward in the AI industry. Leveraging an innovative Transformer-based PoW mechanism, Gonka seeks to ensure that 100% of GPU computing power is harnessed for impactful AI training and inference tasks, thus empowering developers to build freely while ensuring fair compensation for those contributing computing power.
