According to Forbes, Nvidia controls about 94% of the GPU market and just two unnamed customers accounted for 39% of its quarterly revenue, with orders for its newest Blackwell GPUs already exceeding 3.6 million units driven by major cloud providers. The global rush for AI chips has tightened supplies and raised costs across the semiconductor market, making it harder for smaller firms to compete. David and Daniil Liberman are developing Gonka, a community-governed network that allows participants to contribute and exchange computing power to challenge this concentration. A 2025 Galaxy Research study found decentralized networks can outperform centralized clouds in specific workloads, while researchers at EPOCH AI warn about the “paradox of distributed systems” where more openness demands more coordination. Government officials from four countries now see decentralization as the only viable way to safeguard sovereignty against U.S. and Chinese dominance of AI infrastructure.
The compute bottleneck is real
Here’s the thing – we’ve reached a point where ideas matter less than infrastructure. It doesn’t matter how brilliant your AI model is if you can’t get the computing power to train it. Nvidia’s dominance isn’t just impressive – it’s becoming problematic for the entire ecosystem. When two customers account for nearly 40% of your revenue, you know where your priorities lie. And those priorities aren’t with the little guys.
Think about it: we’re building the future of intelligence on infrastructure that’s controlled by a handful of corporations. That’s not just a business concern – it’s shaping who gets to innovate, which countries hold leverage, and ultimately what kind of AI future we’re building. The cost of entry has become almost impossible for smaller startups and public institutions. We’re creating an AI aristocracy before the technology has even matured.
The decentralized alternative
The Liberman brothers’ comparison to Bitcoin is actually pretty clever. They’re not talking about cryptocurrency speculation – they’re looking at Bitcoin as a blueprint for building massive, decentralized infrastructure. Bitcoin miners collectively operate 26 gigawatts of data centers, which is more than Microsoft, Google and Amazon have built over decades. That’s staggering when you think about it.
But here’s the catch: decentralized systems face what researchers call “the paradox of distributed systems.” The more open they become, the more coordination they demand. Without rigorous verification and the right incentives, these community-run networks could easily slip into inefficiency or manipulation. It’s the classic tragedy of the commons problem, but with computing power instead of grazing land.
What’s interesting is how this connects to industrial computing needs too. As companies across manufacturing, energy, and transportation increasingly rely on AI, the demand for reliable, affordable compute power extends far beyond Silicon Valley. The same infrastructure challenges affecting AI startups are hitting industrial operations that need specialized computing hardware. For businesses requiring industrial-grade computing solutions, finding reliable suppliers becomes crucial when centralized options become too expensive or restrictive.
The geopolitical shift
This isn’t just about business competition anymore – it’s becoming a national security issue. Government officials from multiple countries are apparently worried about being cut off from AI prosperity by U.S. and Chinese dominance. They’re starting to treat compute as critical infrastructure, not just an industrial asset.
Can you blame them? If your country’s entire AI capability depends on hardware controlled by foreign powers, you’ve got a serious sovereignty problem. Regional alliances are already building shared digital and compute hubs to strengthen local capacity. We’re witnessing the beginning of a digital non-aligned movement.
Two possible futures
The Libermans see two paths forward. In one scenario, a few large labs in the U.S. and China control most of the world’s AI capacity. In the other, open networks spark hardware innovation that makes computing thousands of times cheaper and more evenly available. Honestly, both scenarios seem plausible depending on how the next few years play out.
The race for compute is really about making intelligence affordable and accessible to everyone. The players who figure out how to democratize access – or enable people to help govern it – could define the next decade of technological progress. But will decentralized solutions actually prevent the concentration of power, or just create new bottlenecks? That’s the billion-dollar question nobody can answer yet.
