The promise of artificial intelligence is seductive. Unlimited computational power, instant answers, personalized services available to everyone at virtually no cost. Tech evangelists paint a utopian vision where AI abundance redistributes wealth and democratizes access to previously scarce resources. But beneath this optimistic narrative lies a more sobering reality: whoever controls the underlying infrastructure—the energy sources, the computing factories, the data centers—effectively controls the terms of access, the distribution of benefits, and ultimately, user autonomy itself.
The Centralization Paradox
The infrastructure required to power modern AI systems demands unprecedented computational resources and energy consumption. Training large language models requires massive server farms consuming millions of kilowatt-hours. Inference—running these models to generate responses—continues consuming significant electricity at scale. This creates a fundamental economic problem that abundance rhetoric glosses over.
Those who control these energy resources and computing facilities become the gatekeepers of AI services. Rather than abundance trickling down freely, we see the emergence of highly concentrated power structures. A handful of companies—OpenAI, Google, Meta, Anthropic—already dominate the AI landscape. Their control extends beyond mere service provision; it encompasses the ability to set terms of service, determine pricing models, and decide who gets access to computational resources.
Energy as the Real Constraint
Energy represents the binding constraint in AI's promised abundance. While computing power has become cheaper thanks to Moore's Law and technological advances, energy remains fundamentally limited and costly. The electricity required to power AI infrastructure scales with computational demand, making energy the ultimate bottleneck.
This creates a predictable pattern: whoever secures long-term energy contracts, whether through fossil fuel partnerships, nuclear power agreements, or renewable energy sources, gains structural advantages in the AI economy. These aren't subtle market advantages—they translate directly into the ability to undercut competitors on pricing while maintaining profitability, or to restrict supply to maintain margins.
Consider the practical implications for developing nations or regions with less abundant or more expensive energy. The promise of AI abundance effectively excludes populations unable to afford energy-intensive infrastructure. The digital divide becomes an energy divide, and centralized control of both deepens existing inequalities rather than resolving them.
User Autonomy in the Age of Gatekeepers
When infrastructure remains centralized, user autonomy necessarily declines. Terms of service can change unilaterally. Access can be restricted, throttled, or revoked entirely. Content moderation decisions—whether appropriate or not—rest with infrastructure controllers rather than users. The illusion of abundance masks the reality of dependence.
This concern extends beyond traditional tech platforms into the cryptocurrency and blockchain space, where similar dynamics are beginning to emerge. As AI-enhanced services integrate with blockchain applications, we're seeing the same centralization patterns reassert themselves:
- API gateways: Access to AI services often flows through centralized API providers, creating single points of control and potential censorship
- Oracle problems: Blockchain applications relying on AI for data feeds depend on centralized sources of truth, reintroducing trusted intermediaries
- Computational nodes: Running AI models locally requires computational capacity most users cannot afford, creating dependence on cloud providers
- Data ownership: AI models train on user data but remain proprietary, concentrating information asymmetries and value extraction
The Distribution Question
Even if AI genuinely produces economic abundance, the distribution mechanism remains critically important. Abundance isn't valuable if only a privileged subset can access it. History suggests that technological abundance doesn't automatically distribute equitably. The internet promised to democratize information; instead, it concentrated power among search engines and social media platforms that monopolize attention and data.
AI abundance could follow the same pattern. Consider a scenario where AI productivity genuinely increases economic output by 10x or 100x. Who captures these gains? If infrastructure control remains centralized, the majority of benefits flow to infrastructure owners while workers face wage suppression from AI-powered automation. This isn't an unavoidable outcome—it's a policy choice reflected in how we structure ownership and governance.
The blockchain and crypto community initially emerged partly as a reaction to this centralization problem. Decentralized systems, theoretically, distribute decision-making power and value more equitably. Yet we're increasingly seeing AI and crypto converge, and the default trajectory replicates traditional centralization patterns unless deliberately designed otherwise.
Decentralized Alternatives and Their Challenges
Some projects attempt to address these concerns through decentralized AI infrastructure. Models like federated learning, where computation happens on distributed nodes rather than centralized servers, offer promise. Blockchain-based incentive structures could theoretically align AI service providers with user interests rather than corporate shareholders.
However, these alternatives face substantial obstacles. Decentralized AI systems typically underperform centralized equivalents due to coordination complexity and infrastructure redundancy. Energy efficiency generally declines when computation distributes across many participants. Users must accept performance trade-offs for autonomy gains.
More fundamentally, decentralization doesn't eliminate the energy constraint. Someone still must provide that electricity. The question becomes whether decentralized structures can negotiate better terms with energy providers, or whether they'll ultimately depend on the same energy monopolies that enable centralized services.
Looking Forward: Questions Without Answers
As AI becomes increasingly central to economic activity, the infrastructure control question demands urgent attention. Will regulatory frameworks emerge to prevent AI monopolies? Can decentralized alternatives achieve the efficiency necessary to compete? Will energy constraints force reconsideration of how we allocate computational resources?
The abundance that AI promises remains tantalizing. But abundance without autonomy, without equitable access, and without distributed control of the underlying infrastructure may simply represent a new form of dependence. The technology itself is neutral; the outcomes depend entirely on the structures we build around it.
