Gonka — Linux for the AI era

In 1991, Linus Torvalds — a 22-year-old Finnish student — released his “hobby” for free and with open source code. Corporations scoffed: Microsoft called Linux a “cancer,” Sun sold Solaris for thousands of dollars, and Linux's market share was near zero. No one believed that a free OS, written by enthusiasts, could compete with billion-dollar corporations.

Key milestones of what happened next:

• 1999: Red Hat goes public — the first company to build a business on open-source software.
• 2008: Google launches Android on the Linux kernel — today, 90% of smartphones globally.
• 2013: Docker revolutionizes servers — containers run on Linux.
• 2025: 75% of servers, 100% of TOP500 supercomputers, all cloud infrastructure from AWS, Google Cloud, Azure — run on Linux.

Gonka replicates this pattern: open-source code on GitHub, anyone can connect their GPU, price is determined by the market, not a corporate price list. The difference is one: Linux took 20 years. In the age of crypto and AI, cycles accelerate — Bitcoin went from $0 to $100K in 15 years, and Gonka has already connected ~4,648 GPUs in the first months after launch.

Today, four corporations control most of the world's AI computations — and each builds its own "walled garden":

• OpenAI: ChatGPT, GPT-5.5, prices $5–30 per million tokens. Closed code, closed training data, complete control over what the model can and cannot answer.
• Google: Gemini, a monopoly on TPU — specialized chips that cannot be bought or rented outside of Google Cloud.
• Anthropic: Claude. One of the leaders in quality, but the API is only available through their own platform with full control over usage.
• Meta: Llama — a formally open model, but inference is still centralized: running Llama 400B+ costs tens of thousands of dollars a month on your own hardware.

The problem is not just prices. Four companies de facto determine the future of AI for all humanity. Each of them decides what the model can say (censorship), who to give access to (geo-blocking), and how much it costs (monopoly pricing). One board decision — and millions of users lose access to an intellectual resource. This repeats the situation of 2001: a few corporations control critical infrastructure, and alternatives seem unserious. Until someone proves otherwise.

All Gonka code is open on GitHub. Any GPU owner can join the network — no corporate permission is needed, no license, no subscription. The price of an AI request is determined by supply and demand in an open market, not by a board of directors. In the first few months, about 4,648 GPUs from ~113 participants have connected to the network — and that number is growing.

The key mechanism that Linux and Gonka share is evolution through competition. Over 15 years, Bitcoin went from mining on ordinary processors (CPU, 2009) to specialized ASIC chips (2013), increasing mining energy efficiency by 300,000x. This is no accident — it's a market mechanism: when thousands of independent participants compete for a reward, everyone looks for a way to make computing cheaper and faster.

The same mechanism works in Gonka. Hosts compete for AI tasks, optimizing their GPUs, software, and network infrastructure. Every efficiency improvement by one host results in lower inference prices for all users. The cost of AI requests via Gonka is already $0.003 per million tokens — about 830x cheaper than OpenAI's GPT. And this is while the network is at an early stage: as the number of GPUs grows and inferenced is optimized, prices will continue to fall.

Linux birthed a multi-trillion dollar ecosystem. Red Hat — acquired by IBM for $34B. Canonical (Ubuntu) — became the standard for servers and IoT. Android — an ecosystem worth over $1T, powered by the Linux kernel. Thousands of companies built businesses on open software without even touching the kernel source code — they built on top: tools, services, integrations.

Gonka can become the foundation of a similar ecosystem for AI. Business layers are already forming around the network:

• Gateways: proxy services that sell AI requests for fiat, e.g., joingonka.ai
• Pools: operators that bundle GPUs and sell shares — Gonka.Top, GonkaPool.ai, Ancapex, CloudMine
• Integrators: companies embedding Gonka inference into their products via an OpenAI-compatible API
• Infrastructure: GPU providers (Spheron), node monitoring, analytics tools

GNK is the "fuel" of this ecosystem, analogous to electricity for data centers: every AI request requires GNK; every ecosystem participant creates demand for the token. The project has raised $80M from Coatue, Bitfury, Insight Partners — institutional capital believes in this story. For an investor, GNK is a bet not on a single company, but on an entire open AI ecosystem, similar to investing in Linux in 2001.