Author | Gonka.ai
Preface: Against the backdrop of the ongoing global AI discussion, industry focus often centers on model capabilities, technological breakthroughs, and regulatory frameworks. But beneath these discussions, a more fundamental question is gradually emerging: Who ultimately controls AI's computing infrastructure? In a conversation at the Unlockit Conference, Daniil and David Liberman, co-creators of the Gonka protocol, futurists, entrepreneurs, and investors, put forward a core argument: artificial intelligence is never a neutral technology; the computing infrastructure determines who AI ultimately serves. In their view, the future of AI is not just a technological race, but a long-term game revolving around the control of infrastructure.
The True Foundation of AI: Not Models, but Computing Power
Centralized AI infrastructure only seems like an inevitability when people don't question its underlying assumptions.
For a long time, most discussions about artificial intelligence have focused on models, ethics, or regulation. But beneath these lies an even more decisive layer—computing power. Who owns the computing power, who controls access to it, and under what conditions it can be used, these ultimately determine how AI operates and who it serves.
Once you view AI from this perspective, the current landscape is hard to ignore. OECD research and other public data indicate that advanced AI computing power is increasingly concentrated in the hands of a few cloud service providers and within a limited number of countries. This creates a widening "computing power gap"—the disparity between those who have access to the infrastructure and those who do not.
This concentration is not accidental. Today, access to advanced GPUs is controlled by a handful of providers and is increasingly influenced by national-level priorities. The result is expensive computing power, limited capacity, and uneven geographical distribution. And this is happening precisely as AI is becoming critical to science, industry, and social infrastructure.
At the same time, current decentralized systems do not automatically solve this problem. Many decentralized systems still consume significant computing power on consensus and security overhead, while incentive mechanisms often reward capital rather than genuine computational contribution. This discourages hardware providers and slows innovation at the infrastructure level.
It is here that our thinking diverges. We do not start from an ideological stance, nor do we choose decentralization to oppose centralized players. We start from a more practical question: What would AI infrastructure look like if efficiency, access, and contribution could be aligned, rather than being in conflict with each other?
This question ultimately leads us to a model: most computing power is used for real AI work, not system overhead; participation and governance rights are determined by verified computational contribution, not capital; access to global GPU resources is permissionless by design. In practice, these assumptions are also constantly stress-tested through ongoing open discussions, including real-time collaboration with GPU operators, developers, and researchers—for example, in our Discord community.
AI has never been just software. It has always been an infrastructure. And infrastructure choices often lock societies into development trajectories that last for decades. Placing this infrastructure under the jurisdiction of a few corporations or nations is not a neutral technological outcome, but a structural decision with long-term economic and geopolitical consequences. If intelligence itself is to become abundant, then the infrastructure supporting it must be designed for "abundance" from the very beginning.
The True Success Criteria for Decentralized AI
The difficulty lies mainly in the fact that you are not arguing with people, but with "default assumptions."
The mainstream tech community often optimizes for what is effective in the short term: speed, capital efficiency, centralized control, and scale through consolidation. These choices are locally rational, but once they become the default, people rarely question them again. When you challenge these default assumptions, it feels like speaking a different language—not because the ideas are extreme, but because they touch upon the incentive structures upon which many careers, companies, and strategies are built.
Even more difficult is the issue. Centralized systems often appear very successful before their long-term costs become apparent. While massive investments and infrastructure spending are already very visible, the deeper costs often manifest later, such as increased dependency, loss of flexibility, pricing power concentrated in the hands of a few providers, and the inability to change course once the system is deeply embedded.
For us, success does not mean winning an argument, nor does it mean replacing existing players. The picture of success is actually much quieter. Success is when decentralized infrastructure is no longer a manifesto but becomes mundane: when people use it not because they believe in decentralization, but because it is the most practical choice.
Ultimately, true success is when the entire discussion itself changes. When the question is no longer "Should intelligence be centralized?" but becomes "Why did we ever think it had to be centralized?" At that point, beliefs no longer need to be directly challenged; they will evolve naturally.
How Do Enterprises Decide Between a Centralized or Decentralized Path?
AI infrastructure is no longer just a technical layer; it is becoming a strategic dependency.
For enterprises, centralized AI infrastructure creates lock-in effects that are difficult to reverse. Once critical systems depend on a few providers, control gradually shifts from the user to the infrastructure owner. Over time, this affects price, access, the pace of innovation, and the range of viable strategic choices.
For enterprises, the issue is strategic flexibility. Centralized infrastructure may work well in the early stages, but it often solidifies into a long-term dependency. Costs become increasingly difficult to control, alternatives become harder to adopt, and changing architectural decisions at scale becomes more and more difficult.
The critical decision moment usually comes earlier than most people think. Infrastructure choices are often locked in before their consequences are apparent. Once AI moves from the experimental stage to everyday infrastructure, the cost of changing the underlying architecture increases exponentially. Therefore, the real decision moment is not when centralized systems fail, but while they still appear to be working well. Exploring decentralized options early preserves choice; waiting often means the choice has already been made.
If Already Dependent on Centralized Infrastructure, Is It Too Late?
It is rarely truly "too late," but the difficulty increases exponentially over time.
Once most systems are built on centralized AI infrastructure, the challenge is no longer technical, but institutional. Workflows, incentives, budgets, compliance requirements, and even talent development paths gradually assume that centralization is "just how things work." By then, change is no longer just about migrating infrastructure; it requires unlearning habits, contract models, and ways of thinking that are deeply ingrained in the organization.
Research on infrastructure lock-in also reinforces this. Industry analysis consistently shows that after running in centralized cloud environments for several years, switching costs rise sharply, not linearly. This growth comes from long-term contracts, regulatory frameworks, deeply integrated internal processes, and a highly specialized workforce. OECD research also points out that countries and organizations that do not gain early access to AI computing power face accumulating disadvantages over time, not only losing competitiveness but also losing architectural freedom—the true ability to choose other infrastructure models.
At the same time, history shows that infrastructure transitions rarely happen all at once. They usually start at the edges. New application scenarios, new players, and new constraints create pressure points where centralized systems begin to be insufficient—perhaps too expensive, too slow, too restrictive, or too fragile. This is often where alternatives begin to matter.
Over time, what is truly eroded is "choice." The longer centralized infrastructure dominates, the fewer real choices exist.
Dependencies gradually solidify, and decentralization changes from an active design decision to a passive correction, which is always more expensive, more complex, and harder to control.
Therefore, the real risk is not that it is already too late. The real risk is waiting until decentralization is no longer a choice but a necessary measure forced by systemic failure. The earlier one explores, even if just in parallel with centralized solutions, the more space there is to proactively shape the outcome, rather than being forced to change under pressure.
For the Next Generation, AI Architecture Will Determine Opportunity Distribution
The next generation needs to understand that technology does not become neutral simply by becoming advanced.
Each generation inherits the infrastructure choices made before them, often without realizing that these choices were once deliberate decisions, not inevitabilities. For the next generation, AI will feel as natural as electricity or the internet does today. Precisely because of this, the underlying architecture is so important—it determines not only what is possible, but for whom it is possible.
The next generation needs to know that access to intelligence can be organized in fundamentally different ways. It can be treated as a shared foundation: open, abundant, and difficult to monopolize. Or it can be fenced off, priced, and controlled, even if it appears convenient and efficient on the surface. Both paths can produce impressive technology, but only one preserves long-term freedom, resilience, and genuine choice.
They should also understand that centralization often arrives quietly. Not through coercion, but through convenience. The initial trade-offs often seem small: slightly lower costs, faster deployment, simpler coordination. But the consequences appear later—when changing direction becomes expensive or even nearly impossible.
It is equally important to recognize that infrastructure directly impacts social mobility. Systems that seem technologically neutral can either reduce the unequal starting points between people and generations, or they can quietly lock these inequalities in place for decades. As you may know, this is also a topic we care deeply about. The younger generation already faces greater disadvantages at a similar stage than previous generations. The current implementation of AI does not solve this problem; it may even make it worse. In this sense, architectural choices determine not only efficiency, but also who truly has the opportunity to experiment, build, and shape the future.
Most importantly, the next generation needs to understand that these systems are still designed by people. Not determined by fate, not by the "market," and not by the machines themselves. Questioning default assumptions, asking who benefits from a particular architecture, and insisting on preserving choice is not a resistance to progress. It is precisely the way to keep progress open.
Why Decide to Share These Stories at Unlockit?
Unlockit seems to be a discussion space where conversations are not centered on hype, launches, or predictions, but on why people make certain choices. This is important to us. Our story is not really about a specific project or a specific technology, but about identifying structural patterns early and deciding not to accept them as inevitable.
For years, we have operated within mainstream systems: building companies, investing, collaborating with large organizations, and benefiting from centralized infrastructure. We understand from the inside how these systems work. At some point, we realized that repeating the same structures while hoping for different results usually doesn't produce anything truly new. Rather than staying silent or packaging it as another success story, we chose to share this realization openly.
At the same time, we came to Unlockit not only to reflect but also to share practical experiences that have real-world significance for the different groups present. For entrepreneurs, these issues involve infrastructure control, dependency on providers, and the ability to scale without losing flexibility. For investors, they involve long-term risks, infrastructure lock-in, and which models truly create lasting value. For corporate and technology leaders, it's about cost structures, reliability, regulatory constraints, and strategic freedom in a rapidly changing environment.
We wanted to share an alternative path that is already operating in practice—not as a universal answer, but as a different way of thinking: how to build AI infrastructure with fewer dependencies, higher transparency, and greater long-term choice. Equally important, we also wanted to hear feedback from those making real decisions at the business, capital, and institutional levels.
We also believe these discussions should not be confined to insiders. Once infrastructure decisions are no longer publicly debated, they quietly solidify into default choices. Unlockit provides a space for reflection before these choices become irreversible, which makes participating in this conversation meaningful.
Ultimately, attending Unlockit is not about explaining what we are doing, but about illustrating why questioning default assumptions remains important, especially in an era where technological progress seems rapid, powerful, and inevitable. At the same time, it is also about listening to the perspectives of those who are shaping the future of business, technology, and social systems.
