The battle for computing power sovereignty: How will decentralization reshape the future landscape of AI?

Original author: Kolawole Samuel Adebayo

Original translation: Gonka.ai

Currently, the competition for the future of artificial intelligence is focused on a key area: computing power. Startups, research institutions, and Fortune 500 companies are all vying for limited GPU resources. This clearly indicates that the next race in AI has shifted from algorithms to the acquisition of computing power. What was once a debate about concepts has now evolved into a substantive struggle for computing resources.

Computing power: The "new power grid" of the AI era

If we compare AI to the electricity of the new era, then computing power is the power grid that supports its operation. However, this "power grid" that determines the lifeline of innovation is controlled by a few giants. They not only determine the allocation and pricing of GPUs, but also, to a large extent, which innovative projects can survive.

According to Reuters, orders for Nvidia's next-generation Blackwell GPUs have surpassed 3.6 million, with the majority going to major cloud service providers. The barrier to entry for smaller startups and public institutions is becoming increasingly high.

Power Structure: Nvidia's Hidden Empire

Nvidia currently controls approximately 94% of the GPU market, making it an indispensable infrastructure behind almost all modern AI systems. More notably, the company recently disclosed that just two "unnamed" direct customers contributed 39% of its quarterly revenue.

This centralization has transcended the commercial realm, reshaping the entire innovation ecosystem. It determines who can participate in innovation, the speed at which costs decrease, and which countries will gain an advantage in the future AI economy. The global frenzy for AI chips has led to supply shortages and soaring prices, further narrowing the survival space for small businesses.

The solution: The rise of decentralized computing networks

Faced with this dilemma, a growing number of researchers and entrepreneurs are rethinking how computing power is allocated. David Liberman, co-founder of the Gonka protocol, points out: "In an efficient market, all products tend to become commoditized, driving down profits and prices to a minimum sustainable level. To enable AI to achieve this, we can draw inspiration from Bitcoin—not as a financial asset, but as a blueprint for building large-scale decentralized infrastructure."

This analogy is quite insightful: "Today, Bitcoin miners collectively operate 26 gigawatts of data centers, a scale exceeding the combined construction of Microsoft, Google, and Amazon over decades. Meanwhile, advancements in Bitcoin mining hardware have reduced computing power costs by hundreds of thousands of times. If the same transformation could be achieved for AI computing power, AI could become truly accessible and affordable for everyone on Earth."

Practical Challenges: The Paradox of Distributed Systems

However, the road to decentralization is not without its challenges. A 2025 study by Galaxy Research showed that, under certain workloads, decentralized networks can outperform centralized cloud services, but verification and reliability remain significant challenges.

Researchers at the nonprofit research organization EPOCH AI call this phenomenon the "paradox of distributed systems": the more open a system is, the more coordination is needed. Without rigorous verification mechanisms and performance-linked incentives, community-run networks can become inefficient or manipulated.

Governance dilemma: Will power be recentralized?

History suggests that decentralized systems may gradually revert to centralization, as power tends to gravitate towards areas with concentrated capital and production capacity. The Liberman brothers acknowledge that even decentralized systems can unintentionally favor large participants.

"No one can unilaterally change the rules of Bitcoin or Ethereum; any change requires broad consensus," they explained. "Some design rules do give mining pools an advantage, leading to a concentration of power. Therefore, when building the Gonka protocol, we deliberately avoided mechanisms such as delegation."

Geopolitical Dimension: Political Considerations of Computing Power Sovereignty

The issue of computing power has evolved into a significant geopolitical issue. Liberman revealed, "Our discussions with government officials in four countries revealed that they increasingly view decentralization as the only viable path to defending their national sovereignty in a context of relying on global AI infrastructure."

"Their concern isn't about control itself, but rather the monopolistic positions of the US and China—monopolies that could isolate them from the prosperity brought by AI. Decentralization is the only way to ensure their citizens equally enjoy the full benefits of AI."

Future Vision: Two Possible AI Worlds

When asked about future development directions, the Liberman brothers described two possible futures: one is that a few large laboratories in China and the United States control most of the world's AI computing power; the other is that open networks will ignite a new wave of hardware innovation, reducing computing power costs by thousands of times and distributing it more evenly around the world.

"In the decentralized future, large cloud companies will still have a place, but they will no longer be able to charge such high premiums for computing power access," they added.

Conclusion: The Reshaping of the Right to Innovation

This battle for computing power is essentially about the right to innovation and the inclusiveness of technology. As AI becomes the core driving force of the future, ensuring the openness and fairness of computing networks will determine the innovation landscape and wealth distribution of the next decade. Those participants who can solve the problem of democratizing computing power will not only achieve commercial success but also shape the direction of the entire artificial intelligence era.

In this race to determine the future, decentralized computing networks are emerging as a vital hope for breaking monopolies and achieving technological inclusion. Their ultimate success depends not only on technological innovation but also, and perhaps more importantly, on the design of governance mechanisms and the building of global community consensus.

Original link