The AI Oligopoly: Joseph Lubin and the Dangerous Consolidation of Intelligence

In an interview published 2026-04-18, Ethereum co-founder Joseph Lubin issued what may be the most direct warning yet from a major figure in decentralized technology: artificial intelligence systems increasingly concentrated in the hands of a handful of corporations pose fundamental risks to societal autonomy. The statement, delivered during a wide-ranging conversation with CoinDesk covering Ethereum's ecosystem, stablecoins, and quantum computing, carries particular weight given Lubin's position at the intersection of cryptocurrency infrastructure and broader technological development. His remarks arrive at a moment when the global AI arms race has accelerated beyond regulatory capacity, raising urgent questions about the governance of machine intelligence and the structural power asymmetries embedded within it.

Lubin's warning crystallizes what scholars in the field of AI political economy have long theorized: the emergence of a technologically mediated form of power that bypasses traditional state-centric frameworks. As he noted in the interview, the dangers of AI controlled by "a few big tech firms" extend beyond competitive concerns into the realm of democratic legitimacy itself. This observation aligns with what historian of technology Edward Tufekci has termed the emergence of "machine intelligence" as a new form of infrastructure—one that, like roads or electrical grids in previous eras, shapes the possibilities of social organization while remaining largely outside democratic accountability. The question of who controls this infrastructure is, therefore, not merely a commercial matter but a foundational question about the distribution of power in the twenty-first century.

Immediate Context: The Consolidation Wave

The warnings come against a backdrop of aggressive consolidation within the AI sector that has accelerated since the release of advanced language models in the early 2020s. A 2024 investigation by the Financial Times documented how the three largest cloud providers—entities with roots in American e-commerce and technology platforms—now control an estimated seventy percent of AI training infrastructure globally. This concentration has proceeded with remarkable speed, aided by the capital-intensive nature of large-scale model development that effectively excludes all but the wealthiest actors. Lubin's Ethereum background offers a particular lens here: decentralized protocols emerged partly as a response to precisely this kind of financial infrastructure capture, and the parallels to AI consolidation have not gone unnoticed in the crypto community. MetaMask, the wallet infrastructure Lubin cited as central to Ethereum's evolution, represents one attempt to create non-custodial alternatives to centralized financial services; the question becomes whether analogous approaches can meaningful exist in the domain of AI infrastructure.

Counter-narratives to alarmist readings of AI concentration do exist within industry discourse. Proponents argue that the open-source movement has democratized access to powerful models, pointing to the proliferation of accessible AI systems that can be run on consumer hardware. Some analysts contend that the "centralization" narrative overstates the coherence and coordination among major players, noting that fierce competition between technology giants could serve as a check on any single actor's power. These arguments carry some weight—competition in AI development remains intense, and the open-source ecosystem has indeed produced remarkable innovations. Yet such reassurances often elide the deeper structural dynamics at play, focusing on competitive dynamics within the existing architecture rather than questioning the architecture itself.

Structural Frame: platform data extraction Meets AI Political Economy

AI concentration sits inside a larger dynamic: the platform business model extracts behavioural data for predictive purposes and feeds that back into AI capabilities, which in turn enhance the extraction potential through more sophisticated pattern recognition and prediction. The result is a system that concentrates not merely computational resources but the capacity to model, influence, and ultimately direct human behaviour at unprecedented scale.

When AI infrastructure concentrates in a handful of corporate entities, what concentrates alongside it is the capacity to encode particular visions of social ordering into the technical substrate of daily life. AI systems embody the values, assumptions, and power relations of their creators and deployers. The implications for Global South nations are particularly stark: technological dependencies create new forms of structural subordination that replicate older colonial patterns in novel technical form. A country that lacks the infrastructure to develop or meaningfully regulate AI systems becomes dependent on technologies whose governance it cannot influence — a dependency that shapes everything from economic policy to military capabilities.

Counter-Narrative: Decentralization and the Limits of Concentration

Critics of the centralization thesis point to developments that complicate any monolithic narrative. The cryptocurrency ecosystem itself, within which Lubin operates, offers one counter-example: despite early fears of crypto as a tool for illicit finance, the technology has evolved into a diverse ecosystem with multiple competing chains, consensus mechanisms, and governance models. Some argue that AI development may follow similar patterns, with specialized models and distributed approaches providing resilience against monopolistic capture. The emergence of inference-focused startups, specialized AI applications, and industry-specific deployments suggests that the AI landscape may prove more pluralistic than the infrastructure concentration figures imply.

Furthermore, state actors are not passive observers to corporate consolidation. China's state-directed AI development, the European Union's regulatory interventions, and the Indian government's technology sovereignty initiatives represent attempts to construct alternative power centres within the global AI ecosystem. These are balancing moves: states responding to concentrated power that threatens their interests by developing independent capabilities. Whether such efforts can meaningfully contest American corporate dominance remains deeply contested; the capital and talent requirements for frontier AI development create barriers that even well-resourced states struggle to overcome.

Stakes and Forward View

The stakes of AI concentration extend far beyond commercial competition or even the broad concerns about surveillance and behavioural control. At issue is the distribution of capabilities that constrains what is politically and economically possible. A world in which AI infrastructure concentrates in a handful of corporate entities, operating primarily within the legal and normative frameworks of a single geopolitical bloc, is a world with fundamentally constrained option spaces for billions of people. The implications for climate policy, development pathways, conflict management, and democratic governance cannot be overstated.

Lubin's warning, arriving from an unexpected quarter—someone whose career has been built on the premise that decentralized protocols can contest concentrated power—carries a particular credibility. His skepticism about quantum computing as an imminent threat, framed as a manageable long-term challenge, suggests someone accustomed to distinguishing genuine risks from speculative anxieties. The AI concentration problem, by contrast, presents immediate and concrete dangers that resist technological quick fixes. Whether decentralized technologies like Ethereum can provide meaningful alternatives to the emerging AI oligopoly, or whether the structural dynamics of AI development make such alternatives inherently marginal, remains to be seen. What is clear is that the question cannot be deferred much longer.

The Monexus desk framed this story around Lubin's specific warnings rather than the broader Ethereum ecosystem updates that dominated wire coverage. His remarks on AI centralization represent a notable intervention from someone with credibility in decentralized systems, and we prioritized the structural analysis of why such warnings carry particular weight in 2026.