The Compute Cartel: How a Few Giants Are Shaping the Future of AI

As social media concentration unfolds before our eyes, it’s easy to overlook a more insidious trend taking shape beneath our feet. In AI, the same dynamics are at play – but this time, the stakes are far higher than mere market share. A handful of tech giants have effectively cornered the market on computing infrastructure, setting in motion a chain reaction that will reshape not just industries but entire economies.

NVIDIA holds 85% of the data center GPU market, while three American cloud companies control 63% of global AI infrastructure. This concentration is no accident – it’s the result of a few behemoths investing heavily in compute capacity, leaving smaller players struggling to keep up. Countries without access to these powerful chips are left on the sidelines, their economies stunted by dependence on foreign powers.

The consequences extend beyond economic domination. As AI increasingly determines what gets built and how information flows, control of compute infrastructure has become a de facto chokepoint for global influence. The US’s recent export controls, for instance, have restricted China’s ability to develop certain AI capabilities. This is a clear example of how access to computing power can be wielded as a tool of geopolitics – and it’s a trend that will only accelerate in the coming years.

Investing in cutting-edge compute capacity may seem like a natural consequence of technological advancement, but scratch beneath the surface, and you’ll find a quieter form of inequality embedded in the technology itself. Large language models are trained overwhelmingly on English-language data, meaning users from other linguistic backgrounds pay more for lower-quality results – a stark example of how unequal access can be built into AI.

The United States has already demonstrated its control over compute infrastructure by setting terms for 191 countries – and those 191 countries are, to varying degrees, dependent on foreign powers. This raises pressing questions about our collective future: what happens when services become unavailable due to sanctions or regulatory shifts? When AI models can be degraded without notice, or quietly redirected to serve the interests of a few powerful actors?

The concentration of computing power is not just an economic issue – it’s a fundamental threat to global stability. The answer lies not in competing on top of this infrastructure, but in rebuilding the layer itself: creating open, decentralized networks for AI compute where anyone can participate as a contributor, not a customer.

This is precisely what we aimed to do with Gonka – our own attempt at building an alternative, community-governed network for AI inference. But it’s too late to simply compete; the next generation will be choosing between permissions, not just models. The future of AI hangs in the balance – and it’s time for us to confront this stark reality: whoever controls compute infrastructure decides what gets built on top, and what gets pushed out.

It’s a harsh truth, but one we can no longer afford to ignore. Whoever holds control over computing power will determine the course of global development – and that’s why it’s imperative that we take action now to build a more decentralized future for AI.