We tend to ask the future of artificial intelligence as if humanity will receive a single answer. Will AI create utopia or dystopia? Will it liberate human beings from scarcity, or give institutions unprecedented power over them?
Those questions may be too simple. There may not be one AI future. Different people, companies, regions and countries could experience very different versions of the same technological transition. The crucial variable may not be whether AI becomes extraordinarily capable. It may be who can access the most capable systems, and how much physical infrastructure stands behind them.
What if AI creates a utopia—but only for the people who can afford it?
1. The Two Scarce Resources of Civilization
For most of human history, civilization has been constrained by the availability of physical resources and human cognitive capacity. Energy determines how much physical work can be done. Intelligence determines how effectively we can discover, design, coordinate and solve. Both have always been scarce, although in very different ways.
The Industrial Revolution loosened one of those constraints. Machines allowed societies to command vastly more energy and physical work than human muscles could provide. Computing then made calculation extraordinarily cheap. The internet reduced the cost of moving information across distance. Each transition changed what society could do by making a previously scarce capability far more abundant.
AI may be the next step in that sequence because it addresses something more fundamental: the scarcity of cognitive labour itself. A human expert can think only so many hours a day. An organization can employ only so many experts. AI systems can potentially be replicated, specialized and deployed at scale, allowing cognitive work to become much more like an industrial input than a uniquely human bottleneck.
That does not mean intelligence becomes literally free, or that every form of human judgment becomes commoditized. It means the amount of deployable cognitive capacity available to an individual or organization could rise dramatically.
FIGURE 1 — ORIGINAL RESEARCH FRAMEWORK
Original research graphic: Energy Abundance vs. Intelligence Abundance. Retain the original graphic and provide its verified source attribution here.
A useful framework for thinking about this comes from Alex Wissner-Gross, whose recent work on the relationship between energy, computation and intelligence treats abundant intelligence as a major civilizational possibility. In that framing, the familiar science-fiction worlds plotted on the chart are not simply stories about technology; they become examples of different combinations of resource and cognitive abundance.
The important question is therefore not simply whether intelligence becomes abundant. It is abundant for whom?
2. Four Possible AI Futures
The original energy–intelligence framework gives us two dimensions. We can extend it analytically by adding a question that becomes critical once AI enters the picture: how broadly are those resources accessible?
| Intelligence scarce | Intelligence abundant | |
|---|---|---|
| Resources abundant | Controlled Utopia | Universal Utopia |
| Resources scarce | Dystopia | Selective Utopia |
This is an analytical extension, not a claim that the original research graphic labels these four AI-specific categories. It helps us ask a different question: what happens when intelligence becomes abundant faster than the rest of the economy does?
Universal Utopia is the obvious optimistic case. Intelligence becomes abundant, energy and physical resources become increasingly plentiful, and access is broad. AI systems, robotics, cheap energy and automated production reinforce one another. The gains are not confined to a narrow ownership class because the underlying productive capacity is widely available.
Dystopia is the opposite possibility. Scarcity persists, while AI makes institutions better at surveillance, manipulation, competition and control. The problem would not necessarily be autonomous weapons or humanoid machines hunting people. A quieter dystopia could emerge if AI works extremely well—but primarily for institutions that already possess disproportionate power.
Controlled Utopia describes a world in which material abundance improves living standards but advanced intelligence remains restricted. People might enjoy cheap goods, energy and services while the most powerful cognitive systems are tightly controlled by governments, corporations or other institutions. Material comfort and cognitive freedom would no longer be the same thing.
Then there is Selective Utopia: intelligence becomes abundant, but the most powerful combination of AI, compute, energy, robotics, proprietary data and capital remains concentrated. The technology works. Productivity rises. Extraordinary capabilities exist. But the benefits are distributed unevenly.
That scenario deserves more attention because AI can potentially move us horizontally before it moves us vertically. We could acquire enormous amounts of artificial intelligence while electricity, minerals, housing, factories, robots and other physical resources remain constrained.
Brains may become abundant before bodies do.
FIGURE 2 — FOUR POSSIBLE AI FUTURES
Four possible AI futures: Universal Utopia, Controlled Utopia, Dystopia and Selective Utopia.
3. The Intelligence Divide
Imagine two people living in the same economy at roughly the same moment. Both have access to AI. Person A has a capable personal assistant that helps with writing, research, planning and everyday decisions. Person B has access to thousands of specialized AI agents, large amounts of compute, proprietary data, autonomous software systems, robotics and enough capital to act on what those systems discover.
Both people have AI. They do not have the same amount of deployable intelligence. The distinction is important because the economic value of intelligence depends not only on the quality of a model, but on how much cognitive capacity can be commanded, how continuously it can operate, what information it can access, and whether someone has the resources to turn its conclusions into action.
This creates what we might call the Intelligence Divide. It is not a claim that wealthy people are intrinsically more intelligent. It is a claim that some people may be able to command vastly more artificial cognitive capacity than others.
That difference could matter at the level of firms as much as individuals. A company with a large fleet of specialized agents could research markets, write software, analyze contracts, model products, monitor competitors and coordinate operations continuously. The organization itself becomes more intelligent without growing proportionally in human headcount.
Once intelligence becomes a productive resource, ownership of that resource can compound. More capital can buy more AI capacity. More AI capacity can increase productivity. Higher productivity can generate more capital. The feedback loop is not guaranteed, but the possibility is economically important.
This is where the AI story becomes paradoxical. AI could be the greatest democratizer of expertise in history: a person with a laptop may suddenly access capabilities that previously required lawyers, programmers, analysts, teachers or consultants. At the same time, frontier AI could become one of the greatest concentrators of productive power if compute, energy, data, robotics and capital remain tightly held.
The important question may no longer be who owns the most machines, but who can command the most minds.
FIGURE 3 — THE SELECTIVE UTOPIA
Different levels of AI access can create very different amounts of deployable intelligence—and potentially reinforce differences in productive capacity.
4. The Race Between Intelligence and Access
There is another reason the outcome is unlikely to be determined by AI capability alone: AI is not weightless. Every artificial mind depends on a physical stack. Chips must be manufactured. Data centers must be built. Electricity must be generated and transmitted. Systems must be cooled. Robots need motors, batteries, sensors and materials. All of it depends on land, factories, logistics and investment.
Recent discussions around the intelligence–energy relationship, including work and commentary involving Wissner-Gross and Ramez Naam, have emphasized this physical constraint. Rapid growth in compute can translate into rapidly growing demand for electricity and infrastructure. The digital appearance of AI can obscure the very physical economy underneath it.
AI may make intelligence cheap before it makes the physical world cheap.
That creates an unusual transition. The world could become extraordinarily rich in cognitive capacity while remaining unevenly supplied with energy, machines, housing, factories and other physical necessities. A brilliant artificial system cannot build a house without materials, transport and machines. It cannot manufacture a million products without factories and energy. Intelligence expands what can be done; it does not erase the cost of doing it.
And this means the future may not be represented by one point on the original chart. Different countries, companies and households could occupy different positions. One region might have abundant energy but limited AI infrastructure. Another might have exceptional compute but expensive energy. A wealthy organization could combine both and then add robotics and capital, effectively creating a private pocket of abundance inside a more constrained economy.
This is why access is the missing dimension. Energy abundance and intelligence abundance tell us what is technically available. Access tells us who can actually use it.
Two forces will push in opposite directions. Falling model costs, open knowledge, widely available tools and competition can democratize intelligence. High compute costs, proprietary data, scarce energy, expensive robotics, concentrated capital and control over infrastructure can concentrate it.
The timing matters. If early advantages compound for years before powerful systems become broadly accessible, the eventual abundance may arrive in an economy whose ownership patterns have already been reshaped by the transition.
The gap can widen before abundance arrives.
FIGURE 4 — THE RACE BETWEEN INTELLIGENCE AND ACCESS
[PLACE OUR SECTION 4 RACE-BETWEEN-INTELLIGENCE-AND-ACCESS INFOGRAPHIC HERE]
AI depends on physical infrastructure, while access determines whether abundance becomes broadly shared or concentrated.
Summary
| Future | Intelligence | Resources | Access | Likely character |
|---|---|---|---|---|
| Universal Utopia | Abundant | Abundant | Broad | Abundance shared widely |
| Controlled Utopia | Restricted | Abundant | Controlled | Material comfort, limited cognitive freedom |
| Dystopia | Scarce or uneven | Scarce | Unequal | AI intensifies competition and control |
| Selective Utopia | Abundant | Scarce or uneven | Concentrated | Exceptional prosperity for those with access |
The most important distinction is simple: a technology can be extraordinarily successful without being equally distributed. AI could succeed on its own terms—becoming cheap, capable, reliable and deeply embedded in the economy—while still producing very different outcomes for people with different levels of access.
The optimistic case therefore requires more than smarter models. It requires abundant energy, plentiful compute, capable robotics, open knowledge, competitive markets, broad ownership and institutions that prevent control over critical infrastructure from becoming permanently concentrated.
FIGURE 5 — WHO GETS TO LIVE IN WHICH FUTURE?
Final synthesis: intelligence abundance may be achievable, but the distribution of energy, infrastructure, capital and access will determine how many people experience it as abundance.
Conclusion
The deepest promise of AI is not that machines will become impressive. It is that intelligence itself could become abundant. That would be one of the most consequential changes in the history of civilization because so much of economic progress depends on our ability to understand problems, design solutions and coordinate action.
But intelligence abundance does not automatically produce universal prosperity. The history of technology gives us plenty of reasons to separate the existence of a capability from access to that capability. AI may democratize expertise at the same time that frontier intelligence becomes concentrated in the hands of those who control compute, energy, data, robotics and capital.
That is why the most interesting AI inequality may not be a difference in intelligence at all. It may be a difference in how much intelligence each person, company or country can deploy.
AI could become both the great democratizer and the great concentrator. Which one dominates will depend less on whether intelligence becomes abundant than on whether access to that abundance follows it.
The future may not be a choice between utopia and dystopia. It may be a question of who gets to live in which one.