A Million Tokens Won’t Pay Your Rent

Universal Basic Compute is an interesting idea, but like income and services it falls short on its own

Some believe AI is already eating jobs.

There’s evidence it’s at least exacerbating unemployment trends, especially for new graduates in white-collar professions. Blue-collar jobs may feel safer now, but improvements in robot hardware and models could put them under pressure too.

We don’t know whether this disruption will be temporary—as many, though not all, job losses were in prior industrial revolutions—or permanent, should we achieve human-level AI. But most reasonable people agree that it will happen fast and need addressing.

What’s harder to figure out is what, exactly, we should do.

Universal Basic Income and its limitations

One common answer is Universal Basic Income (UBI). The idea is simple: as AI and automation reduce the need for human labor, governments could provide everyone with a modest, unconditional cash stipend. Just enough to meet basic needs. Not luxury, but security.

UBI has been tested in pilots around the world, and results are generally positive. People spend more on food and education, stress levels drop, and most recipients continue to work. But UBI isn’t perfect, and obstacles to large-scale implementation are real.

Culturally, UBI carries baggage. In many countries—especially those with strong individualist values—citizens resent the idea of "free money," particularly if they see themselves as the ones paying for it through taxes. Politically, the redistribution required to fund UBI would be a hard sell. The money would likely have to come from taxing income, capital, corporations, and consumption, all of which provoke backlash. And if only a handful of countries adopt it, companies may shift operations to lower-tax jurisdictions.

UBI also introduces tricky questions about meaning and agency. If everyone receives money unconditionally, what happens to motivation? For some, a guaranteed income might unlock creativity, entrepreneurship, or caregiving. For others, it could lead to passivity or disengagement. These risks may be overstated by UBI opponents (after all, retirees and billionaires find meaningful things to do), but we need to consider them, especially if the change from full employment to widespread UBI is abrupt.

Perhaps most fundamentally, UBI assumes governments have the capacity to enact and administer it. COVID-era stimulus payments were a rare exception, and even then, the system showed cracks: delayed payments, fraud, inflation, political gridlock. There isn’t strong evidence that a full-scale UBI would roll out smoothly.

Universal Basic Services: A better alternative?

An alternative to UBI is Universal Basic Services (UBS). Instead of giving people cash to spend how they choose, governments could provide essential goods directly—like healthcare, housing, food, transportation, and even internet access.

This approach has some advantages. It targets basic needs. It can use public procurement to keep costs low. And it avoids some of the psychological and political baggage that comes with handing out free money.

But it has its own problems.

For one thing, it doesn’t seem aligned with current trends. The US, for example, still lacks universal healthcare. Food bank use is rising, not falling. Housing shortages are worsening in many cities. UBS requires a strong and competent public sector—something many governments, especially at the local level, struggle to maintain.

It also risks sapping innovation. When services are centrally provided and not subject to competition, there’s little incentive to improve them. UBI, for all its flaws, retains capitalism’s dynamism: individuals allocate resources, businesses compete to attract their dollars. In a UBS model, that mechanism breaks down.

Enter Universal Basic Compute

Sam Altman, CEO of OpenAI, has floated a third option, Universal Basic Compute (UBC): instead of giving people cash or services, give them access to AI. Specifically, we can give people a fixed allocation of AI tokens, which are units of AI output people currently buy from API endpoints, or receive as part of AI product subscriptions.

So, instead of getting a monthly check or services, everyone would get a ration of digital intelligence. You could use your tokens to earn money (send an AI agent off to a job), save money (get a service from AI that would cost a lot from a human), and save time (automate tasks). Perhaps you could also save or sell your tokens, turning them into a new kind of currency.

The advantage of this approach is that it preserves agency and market competition. Rather than passively receiving money, people would choose how to use their compute. They could spend it productively, creatively, or frivolously, but it would be a resource they controlled. It would also avoid centralizing service procurement and provision in a way that inhibits innovation via competition.

This idea is appealing. It sidesteps political wrangling over taxation. It gives people a tool, not just cash. And it plays to the strengths of the private sector: companies like OpenAI could roll it out unilaterally, without waiting for legislative approval.

But the more you examine it, the trickier it becomes.

Four challenges of Universal Basic Compute

As I see it, there are at least four big challenges for this idea: Pricing, governance, physical needs, and economic uncertainty.

The pricing paradox

The core challenge with universal compute is that its value is bounded by whatever the market charges for compute access. If a ChatGPT Plus subscription gives you a million tokens for $20, then a universal allocation of a million tokens is worth—at most—$20. That’s not nothing, but it’s a long way from a basic income.

This creates a conundrum. OpenAI could raise subscription prices, increasing the value of distributed tokens, but then they’d reduce their customer base and therefore their means of subsidizing those tokens. Prices would be capped by customer price sensitivity and competition. If people could sell their freely received tokens, that competition would include recipients of UBC. Why, as a paying user, would you buy tokens via a more expensive subscription if you could buy cheaper tokens from resellers at a discount?

Governance and control

Who decides how many tokens you get? Who verifies your identity? Who prevents fraud or duplicate accounts? If it’s a corporation, that raises obvious issues of accountability and power. If it’s the government, we’re back to building new bureaucracies.

There’s also the problem of market structure. If OpenAI gives away compute, will Anthropic and Google be forced to match it, by consumer pressure or government regulation? Will tokens only be usable within a single provider’s ecosystem, like frequent flyer miles you can’t redeem elsewhere? That looks less like a public service than a private moat.

Physical needs still matter

Compute can do many things. It can write a business plan, generate a workout routine, optimize your grocery budget, or give you free legal advice. But it can’t build you a house. It can’t grow your food. It can’t do surgery. Even in a world of ubiquitous AI, physical goods and services remain essential, and they’re still constrained by scarcity. (Robots may help, but we’re further away from human-level robots than superhuman AI.)

So unless compute tokens can be exchanged for fiat currency, or used to earn money in a real economy, they won’t be enough on their own. And if they can be exchanged, then we’re back to the earlier pricing problem. How much real-world value can a million tokens command, if everyone gets a million for free, and a ChatGPT subscription is $20?

Economic weirdness ahead

There’s also the question of how this would evolve over time. AI is getting rapidly smarter and cheaper simultaneously. So what happens if you can both sell and save your granted tokens?

A token granted today would have more cash value today but more intelligence value in a year. This is an odd dynamic. It might push people to sell their tokens immediately, to maximize their cash value now. Conversely, it might push people to hoard their tokens, to maximize their intelligence value in future.

Maybe we get an equilibrium, with some people selling now if they need the cash, and others hoarding if they have big future plans. But how might we intervene if not, to avoid deflation if everyone’s selling their tokens immediately, or speculative bubbles if everyone’s hoarding? It’s unclear how you’d design a system that maximizes productive token use. Might you need some kind of token central bank?

A possible piece of the puzzle

I like the idea of Universal Basic Compute. But the more I think about it, the more I realize it’s nowhere near a panacea. It’s not going to soon replace income, nor solve housing or food insecurity. It could enable more productivity, creativity, and entrepreneurship. It could help bridge today’s world of scarcity to, hopefully, tomorrow’s abundance.

But it needs rigorous analysis. Who gets it? How is it allocated? Priced? Can it be saved? Exchanged? What happens when AI grows 10x more capable but 100X cheaper? Do we really want corporations redistributing wealth—taxation without representation? If not, is the government up to the task?

The likeliest answer is not UBI, or UBS, or UBC, but a combination. Some cash, some services, some AI. A foundation of support topped by tools for empowerment.

Giving everyone an allocation of AI tokens is a compelling idea. But right now, it’s more a sketch than a strategy. It deserves serious scrutiny before it becomes the default solution to a problem we’ve barely begun to confront.