Note: This is a research note supplementing the book Unscarcity, now available for purchase. These notes expand on concepts from the main text. Start here or get the book.

Universal Basic Compute (UBC): Why the Future May Run on Compute, Not Cash

In May 2026 the man who bankrolled one of America’s largest basic-income studies said he no longer believes in the idea as much as he once did. What he wants in its place is stranger, and more powerful.

The Premise: What If Compute Becomes the Currency?

Universal Basic Income (UBI) gives people money. Universal Basic Compute (UBC) gives people processing power: a guaranteed quota of AI computation that every citizen can use, trade, or delegate. The difference sounds technical, but it carries civilizational weight.

UBI solves poverty by distributing the output of economic production (cash). UBC solves displacement by distributing the means of economic production (compute). UBI treats citizens as consumers, UBC treats them as producers.

Sam Altman, CEO of OpenAI (which filed confidentially for a public listing in June 2026, says it will list “in 2027 or sooner,” and is in early talks on a funding round at more than $1.2 trillion, up from the $852 billion set in March), floated the idea on a podcast in May 2024: “everybody gets like a slice of GPT-7’s compute,” to use, resell, or donate. Two years later it has hardened into a position. In May 2026 he told The Atlantic’s Nicholas Thompson, “I no longer believe in universal basic income as much as I once did,” and said he was more interested in “collective ownership that could be in compute or in equities or something else.” In a world where AI does most economically valuable work, owning a share of the AI’s capacity may matter more than owning a share of its revenue.

What Davos 2026 Made Clear

The World Economic Forum’s January 2026 meeting kept returning to the agentic economy, a world where autonomous AI systems (called “agents”) don’t just answer questions but take actions: booking flights, writing code, managing projects, negotiating deals. These agents execute multi-step workflows with minimal human supervision, displacing not just manual labor but cognitive work like legal research, financial analysis, and customer service. The old policy toolkit (retraining programs, minimum wage adjustments, tax credits) assumes humans will always be the workers.

The numbers presented there were blunt. In Mercer’s survey of 12,000 workers and executives, the share of employees who fear losing their job to AI rose from 28% in 2024 to 40% in 2026. Entry-level workers face the sharpest risk: an analysis published by the Forum found that organizations could already automate 30% of entry-level work hours. The binary framing of “jobs lost vs. jobs created” gave way to a more honest assessment: most jobs will change, and many workers will be left behind unless we act deliberately.

UBC offers a different premise: what if every human becomes a manager of AI agents?

The debate over UBC turns on three questions:

Access inequality: If compute is the new means of production, who controls it?
Pre-distribution vs. redistribution: Is it better to give people a stake in the AI economy before wealth concentrates, or redistribute after?
Inflation immunity: Unlike cash, compute quotas can’t be inflated away by governments printing more.

The compute wallet, as common as a bank account, is still a white-paper idea. The public compute it would draw on is not: since Davos, governments have started building it.

How UBC Works

The Compute Wallet

Every citizen receives a personal allocation of AI compute capacity. Just as electricity is measured in kilowatt-hours, AI processing is measured in units that represent how much thinking the AI can do for you: the more complex the task, the more units it consumes. This is not a one-time grant. It’s a recurring allocation, perhaps weekly or monthly, that enables ongoing participation in the AI economy.

It works like receiving a plot of fertile land in an agrarian economy, except the “land” is processing power.

Three Use Cases

Personal AI Agents: Use your compute quota to run personal AI assistants, agents that negotiate your bills, manage your schedule, create content, or run side businesses. A tireless assistant handles the administrative friction of modern life.
Pooled Compute Cooperatives: Small communities or cooperatives can pool their allocations to access more powerful models. Ten people with modest quotas can collectively run an agent that one person couldn’t afford alone, the way neighbors might share a powerful tool none could afford individually.
Delegation and Specialization: Citizens who don’t want to manage AI themselves can delegate their compute to organizations (Mission Guilds, professional communities) in exchange for services or reputation credits. Not everyone wants to be a tech power user, and that’s fine.

UBC vs. UBI: The Critical Differences

Dimension	UBI (Cash)	UBC (Compute)
Nature	Passive consumption	Active capability
Inflation	Government can print more	Fixed by physical infrastructure
Agency	Dependent on what markets offer	Enables production, not just consumption
Meaning	Doesn’t solve the “what am I for?” problem - people with nothing to do still feel purposeless	Provides a tool for contribution - you can create value, not just consume it
Corruption	Can be captured by rent-seekers	Harder to extract (non-transferable options)

The key insight: UBI assumes humans are consumers who need money to buy things. UBC assumes humans are producers who need tools to create value.

The Unscarcity Framework Connection

In the Unscarcity model, UBC represents a transitional mechanism, a bridge between the old economy (work-for-wages) and the new one (The Foundation + The Frontier).

During transition: UBC ensures everyone can participate in the emerging AI economy, even as traditional jobs disappear.
Post-transition: Once the Foundation infrastructure is complete, the need for personal compute allocations diminishes, since survival is guaranteed regardless. But UBC principles inform how the Impact economy works: contribution, not consumption, drives access to the Frontier.

UBC also aligns with several Constitutional Core principles:

Law 2 (Truth Must Be Seen): Compute allocations are transparent and auditable.
Law 3 (Power Must Decay): Compute quotas can’t be hoarded indefinitely; they’re “use it or lose it” allocations.
Law 5 (Difference Sustains Life): Diverse uses of compute, from art to caregiving to research, are all valid.

The Objections

“This is just UBI with extra steps”

Not quite. UBI is a check; UBC is a capability. The psychological and economic dynamics differ. A UBI recipient is a passive beneficiary. A UBC holder is an economic actor who must decide what to build, who to collaborate with, and how to deploy their allocation.

“Most people won’t know how to use compute”

Less true every month. “Vibe coding,” describing what you want in natural language and letting AI build it, was Collins Dictionary’s word of the year for 2025. Non-programmers are shipping working software. Within a few years, “using compute” will be as intuitive as “using a smartphone.”

“This favors the tech-savvy”

Initially, yes. But the same was true of literacy when printing was invented. The answer is to distribute the capability while ensuring education and interface design make it accessible. The Civic Service program includes mandatory training on AI orchestration for exactly this reason.

“Who decides how much compute each person gets?”

This is a genuine governance challenge. Options include:

Equal per-capita allocation (simple but ignores varying needs)
Needs-based allocation (complex but potentially fairer)
Hybrid systems (baseline equality plus additional allocation for validated projects)

The Diversity Guard mechanism can adjudicate allocation policies, ensuring no single faction captures the distribution system.

The Geopolitical Dimension

If compute becomes the primary means of production, then control over compute infrastructure becomes the ultimate geopolitical asset. This is no longer theoretical. It is unfolding in real time:

United States: The NSF is transitioning NAIRR (National AI Research Resource) from a pilot program to permanent national infrastructure. Its Operations Center solicitation was a rare winner-take-all federal competition, a single award of up to $35 million over five years to whoever would run the national compute hub and build the unified access portal, and on September 1, 2026 NSF named the winner: the San Diego Supercomputer Center at UC San Diego, partnered with the Texas Advanced Computing Center. The pilot is over; the infrastructure is staying. The Department of Energy’s Genesis Mission is building an “American Science and Security Platform” integrating supercomputers, AI tools, and federal datasets. Meanwhile, OpenAI’s $122 billion funding round (March 2026, $852 billion valuation) was paid for in part by the companies that sell it compute: Amazon, Microsoft and Nvidia all invested, Nvidia alone $30 billion.
Canada: Budget 2025 proposed $925.6 million over five years for sovereign, large-scale public AI computing. In January 2026, Ottawa opened a one-month call for proposals for sovereign data centers exceeding 100 megawatts each, with selected builders negotiating support through memoranda of understanding.
European Union: Council Regulation (EU) 2026/150 entered into force on January 20, 2026, expanding the EuroHPC Joint Undertaking’s mandate to build AI “gigafactories” — public-private compute intended to let Europeans train advanced models without depending on American clouds. It is now a live procurement: the Commission has opened tenders for up to seven gigafactories, raised from an initial five after member-state demand, putting up €10 billion in EU and national money to pull in at least €20 billion private. Bids close November 12, 2026; awards land in early 2027. France’s Mistral borrowed $830 million in March 2026 to build a 44-megawatt data center south of Paris.
India and China: The nearest thing to a compute allowance already exists, though it goes to organizations and not yet to citizens. The IndiaAI Mission has pooled more than 38,000 GPUs from 14 providers and rents them to startups, researchers and students at a subsidized average of about 65 rupees per GPU-hour, well under a dollar. Chinese cities including Beijing, Shanghai, Shenzhen and Chengdu hand small companies “computing power vouchers” to spend at local data centers.
Semiconductors: US export controls, in force since October 2022 and separate from the CHIPS Act’s subsidies, require a license to ship advanced AI chips to China, and enforcement has reached the smugglers: on October 1, 2026, a California computer dealer was arrested on charges of routing more than $300 million of Nvidia servers to China through Malaysia and Singapore. Cloud providers are becoming strategic assets comparable to oil companies in the 20th century.

The clearest sign that compute is becoming metered infrastructure is that the biggest players now rent it like it. SpaceX’s 2026 IPO filing revealed that Anthropic pays roughly $1.25 billion a month for about 325,000 GPUs at the Colossus sites, and Google about $920 million a month for a block of its own, on the order of $26 billion a year in compute rent between two firms alone. In late August Anthropic signed a further $45 billion rental agreement with Nscale, and its draft IPO prospectus, reported by Reuters in late September, lists at least $518 billion of computing commitments over the next decade, about 80% of it owed whether or not the capacity is used. Meta is standing up a “Meta Compute” business to sell its spare cycles. The rent even has a futures contract waiting: in August CME Group announced futures on the hourly price of Nvidia’s H100 and B200 chips, with one of its executives declaring that “compute has become the currency of the AI age.” Trading was due to start on October 5; The Information reported that the regulator has extended its review to November 9. When the most valuable companies on Earth buy processing power by the month the way factories buy electricity, the unit UBC proposes to hand every citizen has already become a priced commodity — the compute-rental economy is no longer hypothetical.

UBC policies will face international coordination challenges:

Can citizens “export” their compute quota?
How do we prevent “compute havens” where quotas are hoarded?
What happens when national allocations conflict?

The Sovereign EXIT Protocol addresses these questions by framing compute infrastructure as shared civilizational capability rather than national property.

Practical Steps Toward UBC

For Individuals

Learn basic AI orchestration. The barrier to entry is the price of a token, and it keeps falling: in September 2026 OpenAI launched its mid-range and small GPT-6 models at half their predecessors’ prices.
Experiment with multi-agent workflows now, and budget for the hardware side to cost more than it did. Silicon Data’s H100 index stood at about $2.53 per GPU-hour in mid-September, roughly 40% above its October 2025 low, as coding agents soaked up supply.
Join or form compute cooperatives to pool resources.

For Policymakers

Study the programs already underway: the US NAIRR now has a permanent operations center, awarded in September 2026. Canada is committing nearly $1 billion to sovereign compute. The EU has a €30 billion tender open for up to seven AI gigafactories. India rents pooled public GPUs to startups and students for under a dollar an hour. These are the early templates for UBC policy.
Consider pilot programs in municipalities or regions experiencing severe job displacement.
Develop frameworks for auditing and transparency in compute allocation. The OpenAI Foundation’s pledge of $1 billion in grants over a year (March 2026), part of it aimed at AI’s impact on jobs and the economy, shows even the private sector recognizes the need for intervention.

For Builders

Create user-friendly interfaces that abstract away technical complexity.
Build “compute cooperative” platforms that enable pooling and delegation.
Design systems that demonstrate diverse uses beyond productivity tools: creative, caregiving, and community applications.

Altman's UBC: Why Give Computing Power Instead of Cash?