The Post-AGI Economy:
What the World Probably Looks Like in 2035

The modern economy runs on a single loop that has operated, with minor variations, since the industrial revolution: businesses make things, humans work at businesses to earn wages, humans use those wages to buy things from businesses. Consumer spending — the B2C economy — accounts for roughly 70% of American GDP and similar proportions across developed economies. Everything else — investment, government spending, net exports — orbits around this central engine.

AGI doesn't disrupt this loop. It fundamentally breaks it. The causality is straightforward and, once you trace it, inescapable:

A single GPU cluster running an AGI model can replicate the cognitive output of 10,000 white-collar knowledge workers — lawyers, analysts, coders, marketers, radiologists, accountants, designers — at a fraction of the cost and at 100x the speed. The corporation replacing those 10,000 workers saves $800M in annual payroll. Those workers lose their income. When they lose their income, they stop buying. When 50 million, then 100 million workers lose their income over the space of a few years, consumer spending fades dramatically structurally, permanently, and in a way that no interest rate cut or stimulus check can reverse.

The GDP doesn't shrink — it grows, potentially doubling. But the composition transforms beyond recognition. Where 70 cents of every GDP dollar once flowed through human wallets, that figure shrinks dramatically — to anywhere between 30 cents (with aggressive redistribution) and 4 cents (without it). The remainder circulates between corporations, between corporations and government, and through the digital welfare system. Ghost GDP — economic output that generates no human wages — becomes the economy. Government response determines how much of that ghost GDP gets redirected back to citizens via CBDC tokens, but the underlying dynamic is the same: earned income collapses, and the state decides what you receive.

As B2C contracts, what replaces it? The answer is B2B and B2G — businesses selling to businesses and businesses selling to governments. But the texture of this commerce is alien to anything that currently exists.

Every major corporation runs an army of AI agents — procurement agents, pricing agents, logistics agents, negotiation agents — that trade with the agents of other corporations autonomously, continuously, at speeds that make today's algorithmic trading look artisanal. 850 billion transactions per day. Roughly 10 million per second. The total gross volume: $180 trillion annually, eclipsing the combined GDP of every nation on earth.

Notice what becomes the bottleneck in this agentic economy. Software agents are abundant — every corporation runs millions. What constrains them is energy (you cannot conjure megawatts from code), compute hardware (TSMC fabs take 3 years to build), and unique data (no amount of intelligence substitutes for proprietary training signal). The companies and assets that sit at these chokepoints — power generation, grid infrastructure, semiconductor fabs, data center campuses, and holders of irreplaceable datasets — become the new toll bridges of the economy.

Consider how a single industry transforms end to end. Moderna uses AGI to discover a novel cancer biologic — the AI identifies the target, designs the molecule, predicts toxicity, and simulates Phase I-III trials in 14 months instead of the traditional 12 years. The biologic works. Who buys it?

In the old economy, the patient bought it (via insurance, via out-of-pocket). In the new economy, the patient has no job, no insurance, and limited CBDC health tokens. So the government buys it — on behalf of the entire population. Moderna signs a B2G contract: 350 million doses at $X per dose, paid from the federal health procurement budget. The government distributes access via CBDC health tokens that citizens use to "purchase" the vaccine from approved providers.

An alternative model — already being piloted — gives each citizen 100 CBDC health tokens over 10 years. They can choose which treatments to buy: Moderna's biologic, Pfizer's competing version, Biocon's biosimilar. The government controls spending through the token budget, not through price negotiation. The market still functions — companies compete for token-share — but the customer is a government-provisioned wallet, not an income-earning individual.

The government's function undergoes a fundamental transformation. Its primary job becomes converting machine-generated corporate profits into human sustenance. UBI consumes ~38% of the budget. B2G procurement — buying healthcare, infrastructure, defense from corporations on behalf of citizens — takes another ~32%. Together, redistribution accounts for 70% of all federal spending. The remaining 30% covers legacy debt service, government operations (themselves largely AI-run), and research funding.

Can this budget balance? At the model defaults — moderate AI adoption, modest policy response — the deficit is small enough to be manageable. Push the Government Response slider higher — implying stronger taxation, less corporate capture — and the budget moves into surplus. Push it lower and the deficit spirals, requiring either debt monetization or cuts to UBI that would cause social unrest at an unprecedented scale.

The UBI is not delivered in dollars or rupees. It arrives as programmable CBDC tokens — digital currency with rules embedded in the money itself. Food tokens can only be spent at food vendors. Health tokens only at approved healthcare providers. Housing tokens auto-debit to your landlord. You cannot save them, invest them, gamble them, or transfer them to your children. They expire monthly.

This is not speculative fiction. India's Reserve Bank already piloted programmable digital rupees for the Public Distribution System in Gujarat in February 2025 — digital food coupons credited to beneficiaries' CBDC wallets, redeemable only at Fair Price Shops via QR code, with real-time tracking and commodity binding. HDFC Bank introduced user-level programmability on the digital rupee in August 2024. IndusInd Bank used programmable e-Rupees to compensate Maharashtra farmers for carbon credits in April 2024. As of March 2025, India's e-Rupee pilot covers 17 banks and 6 million users. Globally, 137 countries representing 98% of world GDP are exploring CBDCs, with 49 active pilots.

The infrastructure for programmable welfare distribution is not a 2035 fantasy. It exists today. What changes is the scale: from 6 million users to 350 million adults in the US, 1.2 billion in India, and comparable proportions worldwide.

There is a genuine upside to all of this, and it would be dishonest to ignore it. AGI compresses decades of scientific progress into months. Drug candidates that took 12 years from target to approval now take 18 months. Cancer 5-year survival rates jump from 68% to 92%. Life expectancy gains 10-15 years. Solar falls below $10/MWh. Nutritionally complete meals cost $2-4. Material science breakthroughs yield room-temperature superconductors, carbon-negative concrete, batteries with 10x energy density.

Consumer credit exists because humans have future income to pledge against. When you take a mortgage, the bank is betting that you'll earn wages for the next 30 years. When you swipe a credit card, Visa is betting you'll have income next month to pay the bill. When you take a student loan, the lender is betting your degree will generate career earnings.

In the UBI-CBDC economy, there is no future income to lend against. UBI tokens are non-transferable, expiring, and purpose-locked. You cannot pledge them as collateral. The entire $17.5 trillion US consumer credit market — $13T in mortgages, $1.75T in student loans, $1.7T in auto loans, $1.1T in credit cards — winds down. The credit infrastructure that powered the 20th century economy becomes as relevant as the telegraph.

In the US, bank count drops from ~4,000 to ~120. Without consumer lending, retail banking, or mortgage origination, thousands of banks are simply redundant. The only growth vertical is corporate credit — and that market is dominated by 10 mega-banks serving 500 mega-corporations.

Venture capital has operated for 60 years on a single bet: that hungry, motivated individuals outside existing corporations can identify opportunities, build products, and create disruptions faster than incumbents can react. The PayPal Mafia, the Bangalore startup ecosystem, Y Combinator's entire thesis — all built on this assumption.

AGI inverts it. A corporation with $1 billion in compute and an AGI with the objective function "maximize economic profit" can evaluate every market opportunity on earth continuously, 24/7. It can prototype, test-market, and iterate at machine speed. When Reliance Jio's AGI can ideate, build, and ship any consumer product in 72 hours, what exactly is the founder's edge? Passion? The machine doesn't sleep. Market intuition? The machine has data the founder cannot access. Speed of execution? The machine is 1000x faster.

The game theory is unforgiving. Consider a VC evaluating a deal in 2033. The expected return of funding the startup must exceed the risk-free rate. But the probability that the startup survives 5 years against an AGI-backed incumbent running continuous optimization is negligibly small. The rational VC stops funding. 18,500 deals per year shrink to 120. $345 billion in deployed capital to $1.5 billion. The large corporations of 2025 become permanent. Schumpeter's creative disruption — the engine that made capitalism dynamic, that allowed outsiders to topple insiders, that gave the economy its regenerative quality — goes quiet.

For all of recorded history — feudal, mercantile, industrial, informational — capital needed labor as a factor of production. The feudal lord needed peasants to work the fields. The factory owner needed workers to operate machines. The tech CEO needed engineers to write code. This mutual dependence is the foundation of every social contract, every labor movement, every democratic norm. The balance of power between capital and labor — imperfect, often deeply unfair, but real — existed because capital could not produce without labor.

AGI combined with advanced robotics breaks this for the first time in human history. Capital genuinely, structurally, permanently does not need human labor. A labor-less capitalism is, in a perverse sense, the theoretical "perfection" of capitalism — the system finally achieves what it always wanted: production without the messy, unionizing, sick-leave-taking, wage-demanding human element. Of course it is also the dystopian endpoint that Marx warned about but never imagined would literally arrive.

The labor share of GDP tells the story: 64% in 1974, grinding slowly to 54% by 2024 — a 10-point decline over 50 years that economists debated endlessly. Then: 54% to 18% in seven years. What took half a century to erode by 10 points drops 36 points in less than a decade. The cliff is not gradual. It is a phase transition.

What replaces creative disruption — Schumpeter's insight that capitalism's dynamism came from outsiders toppling insiders — is what Yanis Varoufakis calls techno-feudalism. A static arrangement where a small number of corporations control the means of production (AI + compute + data), the state manages the population through UBI, and the dynamic churn that characterized capitalism for 250 years simply stops. The lords own the land. The serfs receive their allotment. The system perpetuates itself because the lords' AGI continuously optimizes to prevent any disruption to the arrangement.

There is a darker possibility that needs to be stated plainly, even though stating it feels alarmist. In a world where capital does not need labor, the 270 million Americans (or 1.2 billion Indians) living on UBI are, from capital's perspective, a pure cost center. They consume resources. They require healthcare, food, housing, entertainment, all funded from corporate profits. They produce nothing that capital cannot produce more cheaply with machines.

The game theory here is unsettling. Each corporation's AGI has an objective function that includes maximizing profit. UBI is funded by corporate taxes. Higher UBI = higher taxes = lower profit. Every corporation, individually, has a rational incentive to minimize the UBI population — or at least to minimize the per-capita cost of maintaining it. No CEO in a boardroom will say "let's reduce the population." But the AI optimizing corporate strategy will continuously find ways to reduce the tax burden, which means reducing the cost of human maintenance, which means lower-quality UBI, which means worse health outcomes, which means...

COVID may, in retrospect, have been an inadvertent proof-of-concept. Governments around the world demonstrated that they could confine entire populations to their homes, control their movement, monitor their compliance digitally, and maintain social order through screen-mediated subsistence — all while the economy continued to function via remote work and logistics automation. The infrastructure for managing a non-working population at scale was stress-tested, and it worked. That infrastructure does not need to be rebuilt for the AGI transition. It already exists.

The UBI recipient of 2035 is, by every material measure, better off than the median worker of 2025. They live longer (88-92 years vs 77.5). Their cancer is caught earlier and treated better. Their food is nutritionally optimized. Their housing is new and energy-efficient. Education is personalized and unlimited. Entertainment is infinite and tailored. Transport is autonomous and nearly free. No commute. No workplace stress. No performance anxiety.

And yet — no agency. No career to pursue. No business to start. No wealth to build. No inheritance to leave. No political leverage that comes from being economically necessary. Comfortable subjects, not empowered citizens.

Aldous Huxley understood this better than Orwell. The danger was never the boot on the face — that is easy to identify, easy to resist, easy to name as evil. The danger is the soma tablet, the pleasure hormone, the entertainment that is precisely calibrated to your dopamine receptors. Brave New World is the more accurate dystopia: a population that is too comfortable to revolt, too entertained to notice its captivity, too well-fed to question who controls the feeding.

Noah Hawley's Alien: Earth paints a version of where this leads. In 2120, Earth is governed not by nations but by five mega-corporations — Weyland-Yutani, Prodigy, Lynch, Dynamic, Threshold — each controlling vast territories, commanding private armies, treating citizens as corporate assets. "Your job is not just your livelihood," the show makes clear. "It is your oxygen, your housing, and the reason you are still breathing." The corporate dystopia of science fiction is not a metaphor. It is a business plan that AGI makes executable.

The most common objection to the post-AGI economy described above is psychological: people won't accept it. They'll resist, revolt, demand agency. History suggests otherwise. Society has been moving toward acceptance of managed sustenance for decades, and the data is unambiguous.

The scale of existing dependence

In the United States alone, 72.2 million people — nearly 1 in 5 — receive Medicaid. 66.8 million receive Social Security. 42 million receive SNAP (food stamps). 6.1 million receive housing assistance. Taken together, over 40% of American households receive some form of government transfer payment. The numbers have grown every decade since the 1960s, under both parties, through every ideological cycle. The direction has been one-way.

The revealed preference

What these programs reveal is not laziness or dependency — that framing misses the point entirely. They reveal a revealed preference for material security over abstract agency. Given the choice between uncertain income from labor and guaranteed sustenance from the state, a large and growing share of every population chooses security. This isn't a moral failing. It is a rational response to a world where labor markets are increasingly precarious, where healthcare costs can bankrupt a family, where housing consumes 40% of income.

The post-AGI economy doesn't create this preference. It merely fulfills it at scale. UBI on CBDC tokens is the logical terminus of a trajectory that began with Bismarck's social insurance in 1889, accelerated through Roosevelt's New Deal, expanded with Johnson's Great Society, and now encompasses some form of direct transfer in virtually every nation on earth.

The screen as soma

One data point crystallizes this. Average daily screen time in the US: 7 hours 4 minutes (2024). In India: 7 hours 3 minutes. Among 18-24 year olds: over 9 hours. Humans are already spending the majority of their waking lives in digitally mediated consumption — entertainment, social media, short-form video — voluntarily, enthusiastically, and at accelerating rates. The infrastructure of contentment does not need to be built. It exists. TikTok, YouTube, Instagram, AI-generated content in 2035 — these are not tools of oppression. They are tools of willing absorption. The UBI economy doesn't need to force people indoors. People are already there.

US UBI: $17,400/year. India: $2,400. The 7x gap reflects a simple reality: AGI-generated profits are taxed where the AGI infrastructure sits. Nvidia, Google, Anthropic, Microsoft — these companies sit in the United States. India's IT services revenue — the golden goose that built Bangalore and funded a generation of middle-class prosperity — contracts sharply when global clients no longer need TCS and Infosys to provide human talent. They have AGI.

India has the plumbing. JAM (Jan Dhan-Aadhaar-Mobile) links 500M+ bank accounts to biometric identity. UPI processes 12 billion+ transactions per month. The Digital Rupee pilot is the world's second-largest CBDC experiment. The direct benefit transfer infrastructure is arguably the best on Earth. The rails for distributing UBI exist and work.

What India lacks is the tax base. You cannot redistribute wealth that is generated in American data centers by American-headquartered AI companies. Control of AI infrastructure is the new control of trade routes. The nations that host the GPUs — the US, and to a lesser extent China — capture the economic surplus. The nations that provided cheap human labor (India, Philippines, Vietnam) lose their comparative advantage overnight, because the entire point of cheap labor was that it was cheaper than the alternative. AGI is the alternative, and it is cheaper than any human anywhere.

None of this is inevitable. There are interventions — sovereign AI funds that distribute ownership broadly, democratic governance of AI infrastructure, wealth taxes that prevent terminal concentration, new international frameworks that share the AGI dividend across borders. These are real policy options that real governments could implement.

The window in which they can be implemented is narrow. Once the AGI-powered corporations reach sufficient scale and the UBI population reaches sufficient dependence, the system becomes self-reinforcing. The corporations use their AGI to optimize lobbying, regulatory capture, and narrative control. The UBI population, materially comfortable and politically atomized, lacks the collective leverage to demand structural change. The state, dependent on corporate tax revenue to fund UBI, cannot afford to antagonize the corporations that generate that revenue.

Seven years. Maybe less. The decisions made between 2025 and 2032 — on AI governance, on taxation, on asset ownership, on democratic institutions — determine which equilibrium we land in. After that, equilibrium is exactly the right word: a stable state that resists perturbation.

This is the highest probability future. I desperately hope I am wrong.

A necessary caveat before the punchline. Nothing above is a prediction of what stocks will do in 2025, 2026, or 2027. Markets in the near term are driven by sentiment, flows, momentum, and narratives that have little to do with structural economic shifts a decade away. A company whose terminal value is zero in 2035 can double in 2026 on hype alone. This section is about terminal value — the long-run steady-state worth of different businesses and asset classes in a post-AGI economy — which is inherently nebulous, ever-evolving, and impossible to time. Treat it as a directional compass, not a trading signal.

The organizing principle: value accrues to scarcity.

Every economic era is defined by its bottleneck. In the industrial age, the bottleneck was capital (machines, factories). In the information age, the bottleneck was software talent and distribution. Post-AGI, software becomes abundant — any corporation can generate unlimited code, analysis, content, design at near-zero marginal cost. The bottleneck shifts decisively to what cannot be generated: energy, physical infrastructure, atoms, unique data, and the capacity to turn intelligence into physical reality.

Sectors facing dramatic value compression

Pure software / SaaS. If your moat is code, your moat is gone. AGI generates equivalent software in hours. SaaS multiples — which depend on sticky recurring revenue — erode as switching costs approach zero. Any workflow tool, CRM, ERP, or analytics platform that doesn't own proprietary data or physical infrastructure faces commoditization. Salesforce, ServiceNow, Workday, and their Indian equivalents become utilities priced at near-zero margin.

IT services. The most exposed sector on Earth. TCS, Infosys, Wipro, HCL, Cognizant, Accenture — the entire model of selling human hours to perform cognitive work becomes structurally obsolete. Clients don't need 50,000 engineers in Bangalore when AGI handles the work. Revenue doesn't decline gracefully; it falls off a cliff once a critical mass of clients adopt AGI internally. The Indian IT services sector accounts for ~$250B in annual revenue and employs 5M+ people. Most of this goes to zero or near-zero over a decade.

Traditional financial intermediation. Lending, brokerage, insurance underwriting based on human income and behavior patterns — all structurally impaired. Retail banking, consumer lending, and insurance that prices human risk face dramatic contraction as the human economic footprint shrinks.

Commercial real estate. Offices, retail malls, co-working spaces — all depend on humans commuting to work and shopping in person. The demand base erodes permanently.

Sectors where value concentrates

Unique data & network lock-in. The platforms that survive and thrive are those sitting on irreplaceable data that AGI needs but cannot generate. YouTube (the world's largest video dataset — trillions of hours of human behavior, speech, interaction), LinkedIn (the professional graph — who knows whom, who's done what, the only comprehensive map of human professional capability), X/Twitter (real-time global sentiment, breaking news signal, public discourse patterns). In India: Jio (500M+ users, telecom data, e-commerce behavior, content consumption — the most comprehensive dataset on Indian consumer behavior ever assembled), Reliance Retail (physical-digital purchase data across 18,000+ stores). These companies become more valuable post-AGI, not less, because their data is the training signal that makes AGI useful in their domains. The data is the moat. The network effect that generated the data is the moat behind the moat.

Energy infrastructure — the ultimate bottleneck. AGI runs on compute. Compute runs on electricity. No amount of software brilliance generates a megawatt. Global AI-related electricity demand is projected at 1,000-1,500 TWh by 2030 (up from ~100 TWh in 2024), and post-AGI demand is essentially unbounded. Every layer of the energy value chain benefits: solar manufacturing (the cheapest electrons win, and solar is cheapest), BESS / battery energy storage (intermittent solar needs storage — the grid bottleneck of the 2030s), gas turbines (bridge fuel for data center baseload — GE Vernova, Siemens Energy), grid infrastructure (transformers, switchgear, transmission lines — ABB, Siemens, in India: NTPC, Power Grid Corp, Adani Energy, Tata Power), nuclear (for 24/7 baseload where geography permits). Energy is the new oil. Whoever controls electrons controls AGI throughput.

Data centers & compute proxies. The physical real estate of the AI economy. Equinix, Digital Realty, Nxtra (Airtel), Yotta (Adani), CtrlS, STT GDC — these are the factories of the post-AGI world. Data center REITs may become the most valuable real estate on Earth, even as office REITs approach zero. The supply chain feeding data centers — cooling systems, UPS, fiber optics, precision construction — inherits the value.

Metals & mining — the atoms beneath the infrastructure. Every solar panel, battery, data center, robot, and power line is built from physical materials. The bottleneck behind the bottleneck is the metals themselves:

Hard assets: Gold, silver, and the logic of physical scarcity

In a world of digital abundance — where software, content, analysis, design are generated at zero marginal cost — the assets that cannot be digitally replicated become stores of value by default. Gold and silver have served this function for 5,000 years, not because of tradition but because of physics: finite supply, costly extraction, impossible to counterfeit, no counterparty risk. In a post-AGI economy where fiat currencies may be partially supplanted by corporate tokens and CBDCs, and where trust in institutions is strained by the concentration of power, hard assets become the neutral reserve. Gold is the wealth preservation layer. Silver is the industrial-monetary hybrid. Land (in jurisdictions with strong property rights) is the physical asset that AGI cannot replicate. These are not speculative bets — they are the mathematical consequence of abundance everywhere else creating scarcity value in what remains finite.

The framework in one sentence

If the asset is built from bits, its value compresses toward zero. If the asset is built from atoms that AI needs but cannot create, its value compounds. Position accordingly — on a 10-year horizon, not a 10-week one.