Abundance and Its Owners

IV. Political Economy

When the Masses Are No Longer Needed

There is an idea hidden here that goes beyond economics. Throughout history, the power of ordinary people rested on the fact that the economy needed workers. That is why strikes worked, why states took their populations into account (they needed soldiers and taxpayers), and why business needed a mass customer. Imagine a world where the economy stops needing the ordinary person both as a worker and, in the limit, as a consumer. Then these levers of influence disappear.

In such a world, handing out money turns from an economic necessity into an act of goodwill by the owners. Musk is talking not about a modest “basic income” for survival but about a “high income” in a world where most people do not need to work; in January 2026 he even added that retirement savings would become “irrelevant.” It sounds generous, but it means that the well-being of billions will depend on the willingness of the owners of technology to share.

There is a counterargument: votes in elections are still counted by heads, not by capital, and social stability is in the owners’ interest, so they will probably share. But that is politics, not the market. And criticism comes from the other side as well: economist Sanjeev Sanyal, an adviser to the Indian government, warns that handing out money without withdrawing it through taxes (for example, a “robot tax”) would push demand ahead of supply and lead not to deflation but to hyperinflation. It is no accident that even OpenAI has released a paper proposing a sovereign wealth fund, a robot tax and a move to a four-day week with no loss of pay. In essence, the dispute is about one thing: how to share.

Signs of this logic are already emerging, not as intent but as a quiet retreat of care. Where a person becomes less needed, the incentives to invest in their well-being weaken too. In early 2025 the US agency USAID was effectively dismantled; according to an estimate in The Lancet, continued cuts could result in 9 to 14 million additional deaths worldwide by 2030, including millions of children under five. Within the US itself, Medicaid and food assistance are being cut, and epidemic surveillance is weakening even as the H5N1 virus spreads, which the UN already calls a “global zoonotic pandemic among animals.”

It is tempting to read this as a deliberate plan to reduce “surplus people.” But the facts point the other way: the loudest voices among the elites fear not overpopulation but extinction. Elon Musk calls falling birth rates a risk “bigger than global warming”; the administration is introducing “baby bonuses” and IVF subsidies, even as the US fertility rate fell in 2025 to a record low of about 53 births per thousand women, while support for raising those same children is being cut at the same time. Nor are the elites themselves monolithic: the tough restrictions even on labor migration, which capital opposes, reveal a clash of interests rather than a single headquarters. The picture that emerges is not extermination but neglect: where a person ceases to be economically necessary, investment in them quietly dries up, and no one’s malice is required for that. This is the truly disturbing version: there is no villain to be stopped, only a distributional logic that operates by default.

V. A Live Experiment

The Electricity Bill

So that this does not seem abstract, just look at the electricity bill, a perfect illustration of the fact that scarce energy does not get cheaper but more expensive.

AI requires colossal amounts of energy. According to Bloom Energy, total US data center demand will almost double from 2025 to 2028, from 80 to 150 gigawatts. The Federal Reserve Bank of Dallas allows that wholesale electricity prices could rise by up to 50% as a result. This is no longer a forecast: one Virginia resident received a bill of $281 in January 2026 instead of the usual hundred, and in the PJM region, home to 67 million people, household bills are rising by about 15% on average in 2026.

Consider the arrangement. The technology is built and owned by the richest companies in the world. Yet much of the cost of the energy it consumes is passed on to ordinary families through their bills. As one analyst noted on NPR, under the current market design it is almost inevitable that ordinary consumers subsidize the richest industry on the planet. A scarce resource does not get cheaper but more expensive, and it is paid for not by capital but by people without capital. The simple narrative that “robots will make everything cheaper” is already stumbling over this reality.

By autumn 2026 this thesis had been confirmed on the scale of the entire economy. After renewed clashes between the US and Iran in the Persian Gulf, US WTI crude rose from under $70 in early July to more than $100, and on September 11 the average price of a gallon of diesel in the US topped $6 for the first time in history, to the despair of farmers and truckers. Energy turned out to be not a side detail but the lever that swung the entire macroeconomy from deflation to inflation.

VI. Timeline

Three Layers of Automation

No one has an exact date, but the process can usefully be broken into layers. The first, intellectual office work, is already under way. Goldman Sachs estimates that AI is reducing US employment by about 16,000 jobs a month. Recent graduates aged 22–27 are finding it ever harder to get hired: their unemployment rate is about 5.6% versus 4.2% for the country as a whole. Anthropic CEO Dario Amodei warns of a possible “white-collar bloodbath”: the disappearance of up to half of entry-level office positions within five years and unemployment rising to 10–20%. The insidious part is that “entry” positions, the ones careers used to start from, disappear first; the blow falls not so much on those who have jobs as on the very road to them.

The economics of a humanoid

The second layer is physical robots. Progress here is slower: hardware is harder than software. Tesla promises to unveil a new-generation Optimus this year, and the economics promise to be disruptive: in mass production such a robot is estimated at $20–25 thousand, and renting it at about $1.5–2 an hour, dozens of times cheaper than human labor. But broad penetration into the economy is more likely a story of the 2030s. The bottleneck is not the robot’s “brain” but energy, materials, manufacturing capacity and the sheer speed at which technology spreads through everyday life.

The third layer is civilizational “abundance.” If it arrives, it will most likely be somewhere between 2035 and 2050 or later, with a huge range of uncertainty. Faster if AI starts improving itself; slower if we run into limits of energy, physics and regulation. History has a cooling effect: both electricity and computers reshaped the economy over decades, not years; this is the well-known “Solow paradox.” In autumn 2026 one more brake was added: the cost of capital. The construction of abundance now has to be financed with debt at the highest rates in twenty years.

VII. The Resource Question

Russia in a World That Needs Its Resources

And this is where the logic of the essay meets geopolitics. If energy and raw materials are exactly what does not get cheaper in a world of robots but more expensive, then those who own them acquire special importance. Russia is one of the largest holders of the scarce physical world.

The scale is impressive. It is the second-largest gas exporter and one of the largest oil producers, and the world’s third-largest country by volume of mineral extraction. But the more interesting story is not oil but the metals of the future: Russia accounts for about 41% of the world’s refined palladium, about 44% of global uranium enrichment capacity (through Rosatom), almost a fifth of high-grade battery nickel (Nornickel), and is the largest exporter of titanium. By Moscow’s own estimates, demand for critical raw materials will grow manifold by 2050: twofold for uranium and titanium, threefold for niobium, and dozens of times for lithium. In other words, the coming economy of chips, batteries and magnets will need exactly what lies in Russia’s subsoil.

The paradox of abundance. Yet Russia is a textbook case of the "resource curse": a rich country with a poor people. For decades, resource rent has flowed not into diversification but into the pockets of a narrow group and into showcase projects. On Transparency International's Corruption Perceptions Index, Russia scores only about 22 out of 100, near the bottom of the global ranking. Inequality is extreme: the top 10% capture almost half of national income, and a substantial share of capital is estimated to be hidden offshore. In 2026 the IMF cut its growth forecast to 0.8%, companies began talking about "tactical poverty" (wage freezes and bonus cuts), and the budget runs a deficit amid falling oil and gas revenues (which dropped by roughly a quarter in 2025). That is the essence of the curse: the rent does not reach the population.

The world will need Russian rocks, but not Russian people.

The nuclear shield. Overlaying this is a circumstance that changes every calculation: nuclear weapons. They make forcible regime change from outside unthinkable. Russia cannot be coerced the way weak countries are; whatever government it has, the world has to trade with it rather than fight it. This locks the situation into a long, uncomfortable coexistence: the resources are needed, coercion is impossible, so it will have to be dealt with.

The pivot to the East. The sanctions imposed after 2022 did not collapse the economy (it is a “slow puncture,” not an explosion) but they reoriented it. Today China buys about half of Russia’s crude oil and India roughly a third (and its share is falling); more than half of seaborne oil is carried at a discount by the sanctioned “shadow fleet.” An asymmetry emerges: Russia is increasingly becoming Beijing’s junior raw-materials partner. Its rare-earth ambitions underline this: Moscow has announced plans to raise its share of global rare-earth supply from the current ~1.3% to 10% by 2030 and to launch the Siberian “Angara-Yenisei Valley” cluster (about $9 billion, overseen by Sergei Shoigu), and First Deputy Prime Minister Denis Manturov promised at the St. Petersburg forum a full production cycle by 2028. But processing is still dominated by China (about 90% of the global total), and without it these plans hang in the air.

How the relationship may evolve. These forces add up to several scenarios, not mutually exclusive but more likely to flow into one another.

1. Transactional coexistence. Sanctions “leak” through intermediaries and the shadow fleet; the world quietly buys Russian raw materials at a discount, keeping a moral distance but not refusing. This is the path of least resistance, especially in an energy shortage: the cheap molecule beats principles.

2. Vassalage to Beijing. The tighter the isolation from the West, the stronger the dependence on the single large buyer. Russia risks turning into China’s raw-materials appendage, with asymmetric prices, settlements in yuan and a loss of strategic autonomy.

3. A world split into blocs. Russia entrenches itself as the raw-materials anchor of the non-Western camp (China, India, part of the Global South), while the West “decouples” at great cost and with difficulty. Raw materials become a bargaining tool: Moscow is already trying to play the rare-earth card in negotiations, and elements of “resources in exchange for peace” even seeped into the US plan for Ukraine discussed in late 2025.

4. Internal fragility without an external solution. Demographics, the outflow of technology, stagnation and military spending are eroding the system from within. But the nuclear umbrella rules out external intervention, which means any change can only come from within, and that is a slow and unpredictable process.

The link to the main thesis. Here Russia becomes a macro-model of the entire essay. In the AI economy the world will need Russia’s natural resources more and more, and Russia’s people less and less. Resource rent goes to a rent-seeking elite, while the population, like the “person without capital” of the previous sections, becomes increasingly economically unnecessary, both to its own state and to the outside world. The wealth of the country and the poverty of its citizens cease to be a contradiction: they are two sides of the same distributional logic. If there is any place where the thesis “abundance does not guarantee a share of the abundance” is visible to the naked eye, it is here.

VIII. The Third Pattern

China: The Workshop That Exports Deflation

If Russia is the junior raw-materials partner, China is the senior one: not a rentier trading its subsoil, but the workshop of the world, which itself builds the machines that replace human labor. Its case is a third, entirely separate pattern, in which strength and weakness grow from the same root.

The engine of “good” deflation. It is China that is making the world cheaper. Its excess capacity and ferocious domestic competition, what Beijing calls “involution,” spill over into price wars and a flood of cheap electric vehicles, solar panels and goods spreading across the planet; its giant trade surplus, meanwhile, is not shrinking but growing. The cheap half of abundance from the opening of this essay is largely produced here, through market mechanisms but under the guiding hand of the state and of local authorities protecting their own.

The owner of the “new means of production.” China does not merely assemble other people’s gadgets; it seeks to own automation itself. The country dominates the links on which the robot economy rests: rare-earth processing, the production of permanent magnets and lithium-ion batteries, and the manufacture of solar modules.

Beijing has already turned this dominance into leverage: in 2025–2026 it imposed export restrictions on rare earths and magnets modeled on the US “foreign direct product rule,” and in early 2026 it applied them selectively against Japan to cool its support for Taiwan. European prices for certain rare-earth materials reached six times Chinese levels. An alternative supply chain can be built, but it takes years and tens of billions: the issue is not the ore but the processing, which China has refined over decades.

Energy solved: a contrast with America. Here China strikes at the main bottleneck of this essay. While the American AI build-out runs up against rising electricity prices and overloaded grids, China drowns the problem in capacity: in 2025 total installed capacity reached almost 3,900 GW, and wind and solar overtook coal in capacity for the first time. Solar generation is the cheapest (about $30–40 per MWh versus $50–150 for coal and gas), and modules have fallen to ~$0.1 per watt. There is so much energy that the state pays coal plants and batteries simply for standing by in reserve (around $47 per kW per year), and part of the output goes unused. In an era when energy is becoming a scarce resource, China is removing its own scarcity.

Robots instead of people, but according to plan. And this is where the thesis that “people are no longer needed” shows itself in its purest, state-directed form. China’s labor force is huge, but it is rapidly aging and shrinking: the birth rate in 2025 was the lowest since 1949, the working-age share fell to about 61%, and according to Barclays their number will decline by almost 37 million over the decade. Beijing’s answer is not to bring in people but to replace them with machines: more than 140 companies make humanoid robots on state subsidies, the goal is global leadership in their production by 2027, and the Chinese humanoid’s production cost (~$35 thousand, falling to ~$17 thousand by 2030) is already four to five times lower than that of its Western counterpart. Robots are planned for factories and even for elderly care. The state is openly betting on replacing the scarce worker, not on giving that worker more.

China exports cheapness to the whole world, and suffers from it itself.

A weakness that is this essay’s thesis turned inward. China’s main vulnerability is precisely the “bad” deflation the essay warned about. For decades the country suppressed its own consumption so hard (high savings, weak demand, a burst property market) that the surplus went not to households but to investment, the state and exports. Consumption’s share of the economy remains low, and now that is coming back to bite: consumer prices in early 2026 hover around zero and below, the inflation target of “around 2%” is the lowest in decades, and the growth target of 4.5–5% is the most modest since the early 1990s. Tellingly, the authorities deliberately prefer to prop up production rather than “aggressively” stimulate demand. This is a living illustration of the question “who gets the surplus”: when the answer is “the state and the factories, but not the people,” the country itself falls into underconsumption.

Political economy and risks. Redistribution here is also a political choice, but it is made not by the electorate but by the party: with an uncompetitive political system and opaque finances (debts of local governments and developers, stories like Vanke), Beijing chooses capacity over consumer well-being. Add to this the geopolitical fault line: the scarcest asset of the AI era, advanced chips, is made in Taiwan, and Beijing’s drive to reclaim the island is, among other things, a struggle for a non-reproducible resource of the highest order; access to Russia’s cheap resources and the penetration of Chinese goods into developing and Western markets complete the structure. Strength and fragility grow from the same root: a model brilliantly tuned for production has never learned to let people consume what is produced.

IX. The World Map

Three More Trajectories: Europe, India, Japan

The US, Russia and China mark out the extremes, but three more cases complete the picture, and each highlights a different facet of the same question: who is needed in the new world.

Europe: a generous redistributor without an engine of its own. The EU is better than anyone at sharing: a well-developed social safety net softens the blow of automation. But it owns almost none of the “new means of production.” Mario Draghi’s 2024 report called the situation an “existential challenge”: Europe lags the US by about a third in income per capita, invests half as much in research, has not produced its own tech giants and AI champions, depends on China for raw materials and suffers from expensive energy. It needs about €800 billion of investment a year (4–5% of GDP), yet within a year only about a tenth of the recommendations had been implemented. Europe’s risk is to become a museum: a comfortable place that consumes other people’s technologies and distributes dwindling wealth while losing its levers of influence. The Economist’s autumn calculations add a fiscal dimension: it is precisely the large European budgets that live on labor taxes (Italy, France, Germany) that would suffer most if AI cuts labor’s share of income. And from Wall Street, as the magazine caustically remarked, Europe imported private capital and aggressive debt restructuring, but not venture capital or securitization, its most useful parts.

India: a demographic dividend in an era that does not need labor. India has the largest youth population in the world, about 367 million people aged 15–29, but the “dividend” window is already closing: the oldest cohort passes its numerical peak in 2026. The trouble is that the path that lifted East Asia (cheap labor → factories and services) is being blocked exactly when Indian youth are entering the market: the country leapfrogged labor-intensive industrialization, and AI is now undercutting its mainstay, IT outsourcing and entry-level office roles. Unemployment among young graduates stays around 40%, real wages of the young have not grown since 2011, and the world’s lowest “employment elasticity” means the economy grows while creating almost no jobs. The bet is on “sovereign AI” and reskilling, but the scale of the response is still modest relative to the size of the workforce.

Japan: the future that has already arrived. Japan reached this path before anyone else: about 30% of the population is over 65, the population has been shrinking since 2010, the labor force is declining by millions, and there is almost no immigration. The answer is robots: one of the highest industrial robot densities in the world, robot exports three times domestic shipments, and a goal of up to 30% of the global “physical AI” market by 2040. And Japan is the “deflation veteran” from the opening of this essay: after three “lost decades” it has finally escaped (prices have been above the 2% target for three and a half years, wages are rising at historic rates; the 10-year government bond yield has approached 3% for the first time in 30 years, and the Bank of Japan is preparing to raise rates). The lesson is twofold: automation softens demographic decline, and a wealthy creditor society with huge foreign assets can age and shrink while remaining stable, but this took decades and rested on relative social cohesion.

X. Capital and People

Elevators That Are Closing

One question runs through the entire essay: who gets the surplus. It is time to answer it directly, and to map who has the money, because this is the boundary that is becoming sharpest.

Capital is concentrating, and is increasingly inherited. The numbers speak for themselves. The top 10% own about three-quarters of the world’s wealth, the top 1% about 40%, and the bottom half of humanity about 2%. At the very top, things are accelerating: some 60,000 people in the top 0.001% hold three times as much as the 4 billion in the bottom half, and their share grew from 3.7% in 1995 to 6.1% by 2025. In the US, the top 10% of households own 93% of all stocks, and labor’s share of national income is at a 78-year low. The gains from the stock market boom and the AI wave surface exactly where this essay predicted: with the owners of assets.

Neo-feudalism: the elevators go up, and then close. The intuition about “feudal times” is borne out by the data. Whereas almost 90% of those born in 1940 earned more than their parents, among those born in the 1980s only about half do; on social mobility the US ranks only 27th. About 60% of billionaire fortunes today are linked to inheritance, monopoly or proximity to power rather than entrepreneurship; economists call this “unearned wealth.” Wealth is passed on not only as money but also as connections, education and cultural capital, forming almost hereditary estates. The old “elevators” to prosperity (education, a good job, a home of one’s own, savings) are weakening one after another: there are more and more diplomas, but their value is falling; entry-level jobs are being eaten by AI; housing is rising faster than wages (the median US home has risen in price by about 55%); savings are losing to asset inflation. There is an objection (“creative destruction” still creates new rich people, and AI may make access to expertise cheaper), but the trend points toward the hardening of estates, not their erosion.

Who the main buyer is now. Since the wealth of owners is multiplying while wages stand still, the well-off are also becoming the main consumers. In the US, the top 10% account for about half of consumer spending (versus ~43% in 2020), and the top 20% for about 60%; spending by the top 10% grew 62% in five years. Hence the “K-shaped economy”: both luxury (luxury fashion sales up 8% in a year, business class) and discounters are flourishing, while the middle is being hollowed out. The economy “at the top” is becoming self-sufficient, circulating among AI, the stock market and the spending of the rich, and precisely because working hands are no longer needed, only talents and innovators, the line between those who own or create and everyone else is becoming starker.

The feudal analogy is accurate, but with a caveat that makes it more alarming. Back then the lord needed peasants to till the land; now the owning class may need neither the labor nor, beyond a certain point, even the demand of the majority. The economic pressure that in earlier “gilded ages” forced elites to share (the need for workers, soldiers, customers) is weakening. Two ways out remain: a political one, taxes on capital and “predistribution,” and a personal one: to end up on the owning side of the line, that is, to own assets or be a scarce talent, because the middle is being hollowed out. History also knows a way out of such a dead end; that, and why it may not work today, is the subject of Section XIII.

XI. The Price of Money

Abundance on Credit

The first edition of this essay warned that energy is a bottleneck that breaks the simple narrative of “everything will get cheaper.” By autumn 2026 the warning had come true faster than one might have expected. Instead of the promised deflation, the world got the opposite: an oil shock, sticky inflation and the sharpest rise in the price of money in two decades.

Energy turns the central banks around. The rise in oil and diesel (Section V) pulled prices up with it: US core inflation for August came in above forecasts. The central banks, which we described as fighters against deflation, turned in the opposite direction: on September 10 the ECB raised its rate from 2.25% to 2.5%, and markets expected hikes from the Fed and the Bank of Japan in the middle of the month. New Fed Chair Kevin Warsh promises “not to flinch” in bringing inflation back to 2%.

Bonds: the return of the “Macarena.” In 2017 San Francisco Fed President John Williams assured that high rates, “like the pager, the Walkman and the Macarena,” would not come back; during the pandemic the 10-year Treasury yield fell below 1%. On September 10, 2026 the 30-year Treasury yield approached 5.4%, its highest in almost twenty years, and the 10-year touched 4.98%, nearly matching the October 2023 peak. This is not just an American story.

There are three reasons, and two of them were anticipated above. The first is debt. The US budget deficit is about 6% of GDP, the highest outside of recessions and wars, and this with a strong economy; more than half of it consists of interest on past borrowing. From October through July, federal interest expenses reached $931 billion, three times the figure for the same period of 2020. Scott Bessent’s goal of cutting the deficit to 3% of GDP by 2028 looks like fantasy. The second is policy detached from macroeconomics. On September 9 Donald Trump promised to pay every adult American $5,000 if the Republicans keep control of Congress: $1.2 trillion, or 3.5% of GDP. Implementation is unlikely, but here, in miniature, is that same “universal income” from the opening of this essay, only announced not out of surplus but on credit.

The third reason is AI itself. At first the tech giants paid for the data center build-out from their giant cash piles; now they are borrowing. Investment-grade borrowers will raise about $1.9 trillion in 2026, more than a third more than in 2025; according to UBS, this flow of issuance has added about 0.2 percentage points to 10- and 30-year Treasury yields. Norway’s sovereign wealth fund ($2.3 trillion) wants to cut the share of government bonds in its fixed-income portfolio from 70% to 50%; if parliament approves the plan, about $80 billion will move from Treasuries to private borrowers. In other words, building “abundance” itself raises the price of money for everyone, governments included.

“Now I am the casino.” Treasury Secretary Bessent, a former hedge fund trader, decided to take on the market. On August 19 the Treasury announced buybacks of long-dated government debt worth tens of billions of dollars, and on September 9, the day after the remark “Now I am the casino… You can bet against me,” it tripled the program to $6 billion. It did not work: yields kept rising. Even after the August announcement the dollar fell and gold soared, both signals of growing distrust of American assets. Bessent’s former boss, the legendary macro investor Stanley Druckenmiller, publicly urged him to “let the bond market speak.” Meanwhile, Fannie Mae and Freddie Mac are stepping up purchases of mortgage securities to push mortgage rates down, and a dollar swap line with the UAE is under discussion so that its sovereign fund does not sell Treasuries in a crisis. The Economist calls things by their name: these are steps toward “financial repression,” capping yields, pushing bonds onto domestic savers and quietly inflating away the debt, as the US did after the Second World War. For our subject this is the key turn: the most likely way to “share” abundance is not the generosity of elites but a slow inflation tax on those who hold cash and bonds.

The cheap money on which asset prices grew has run out.

Speedball capitalism. Meanwhile, the American financial machine is running at full throttle. Within a year the market has digested the largest IPO in history (SpaceX), the largest share offerings by a public company (Google) and by a foreign one (SK Hynix), a record private funding round (OpenAI) and a record private credit deal (Broadcom), most of the largest bond issue (Amazon), the first trillion-dollar ETF (Vanguard), Berkshire Hathaway’s record cash pile and Nvidia’s record buyback, the largest leveraged buyout (Electronic Arts) and record junk bond issuance. The Economist compares this to a “speedball,” a mix of heroin and cocaine: euphoria and stupor at the same time. Every player in the economy is defined by its position in the national bet on AI, and the balance sheets of the largest companies, as in a war economy, have been “requisitioned for the cause.”

The speedball in numbers

The scale is visible in adjacent industries too: GE Vernova’s gas turbine order backlog of $176 billion, record capital spending by energy companies, profits at the banks arranging data center financing. According to Stijn Van Nieuwerburgh of Columbia University, if 183 GW of new capacity is built by 2032, this boom will exceed the railway and dot-com booms combined as a share of the economy. And every American’s “patriotic duty” is to spend and to bet: thanks to the stock market rally, household net worth has reached a record ~830% of disposable income.

Where it may crack. The cracks are already visible. US banks have $327 billion of unrealized losses on bonds (in 2022 the figure reached $688 billion), and if yields keep rising, Silicon Valley Bank will come to mind. Riskier segments of life insurance are under pressure because of their investments in private credit, just when that capital is needed most. Oracle has reportedly invoked force majeure on a data center project in New Mexico; Oura has postponed its IPO, citing “uncertainty,” with the market near an all-time high. Rising yields hit expensive tech stocks and real estate first, that is, precisely the assets whose growth fuels the spending of the top decile. The “K-shaped economy” from Section X turns out to be reflexive: its upper branch rests on asset prices, and those rest on cheap money, which has run out.

XII. The Fiscal Bill

Who Will Pay for Abundance

Musk promised that AI would wipe out the US national debt in three years. The autumn calculations paint the opposite picture: the profit boom is already under way, yet the Treasury is getting almost nothing from it.

The fiscal arithmetic of AI

Profits, but no taxes. Over the year, pre-tax profits of US companies rose by $1.2 trillion, or 30%, while corporate tax receipts collapsed: the hugely profitable tech giants are writing off their huge capex against the tax base under the new accelerated depreciation rules. The scale of the drop surprised even budget analysts. This is the fiscal mirror of the “K”: there is a surplus, but neither workers nor the state have it.

A tax system built for a world where people are needed. The main problem runs deeper. In the US, about three-quarters of federal tax revenue comes from taxes on labor, income tax and social contributions; in OECD countries their share is high and rising. As Anton Korinek (Anthropic) and Lee Lockwood (University of Virginia) write, “as AI reduces labor’s share of income, the traditional tax base erodes.” AI may create value where taxes are low and destroy it where they are high. The Economist has done the math: if labor’s share of national income falls by 10 percentage points, countries with high taxes on capital, such as Australia and South Korea, will barely suffer, while most budgets that live on labor taxes will be “crippled.” Italy’s deficit would more than double as a share of GDP, with France and Germany not far behind; the hit to the US is smaller but significant.

Even the optimism here is conditional. In the model by Karen Dynan, Douglas Elmendorf and Louise Sheiner, AI accelerates growth and over 30 years reduces US debt by about 50% of GDP relative to the baseline, but the debt burden still keeps rising into the 2050s, only more slowly. That is a very long way from “wiping it out in three years.”

The scissors: spending rises along with need. Worse still, AI may make the state itself more expensive. This is the same “Baumol’s cost disease” from Section I, but seen from the Treasury’s side: in the industries where AI has the greatest effect wages rise, and the public sector has to pay more to retain people, so the cost of every service goes up. At the same time, demand for those services grows: retraining, benefits, perhaps even a basic income, “to keep the mob at bay,” as The Economist puts it. In its scenario, Italy would need an extra $200 billion a year to compensate the earnings of people displaced by AI, while tax revenue fell by $90 billion, a deterioration of about 10% of GDP. Here is the mechanics of the “neglect” from Section IV: it is not so much that the state does not want to care for people as that it loses the base for doing so.

Whom they will tax. Since money is moving from labor to capital, the tax collector will follow it. A direct “AI tax,” on compute or on every query to a model, is popular with the public, but the base is small: Americans spend about $700 billion a year on building and running AI versus a $16 trillion wage bill, and the tech giants’ capital spending is forecast to plateau from 2028. (Daron Acemoglu, Andrea Manera and Pascual Restrepo point out, meanwhile, that the current system itself encourages automation: labor is taxed heavily, software and machines lightly.) Wealth taxes are also on the rise: in November Californians vote on a 5% annual levy on fortunes above $1 billion; a bill has been introduced in Congress for 2% on fortunes above $50 million and a 1% surcharge above a billion. But experience shows that the rich leave and move their assets out, and private businesses are hard to value: in 1990, 11 rich countries levied a wealth tax; today four do. And there is a new detail: part of the AI fortunes is flowing into charitable foundations. According to The Economist, all seven co-founders of Anthropic, which is preparing an IPO at a valuation of about $2 trillion, have pledged to give away 80% of their wealth, so there may be fewer taxable fortunes than it seems.

That is why the consensus among economists is not new levies but a retuning of the old ones. First, bring the tax on capital gains closer to the tax on labor: California already taxes them at income tax rates, Belgium has introduced a 10% tax on income from financial assets, Australia is raising its tax on realized capital gains (adjusted for inflation), and the Netherlands is moving to a tax on actual rather than notional returns. Second, a tax on corporate “excess profits” generated by monopoly power, scarce land or unique data, that is, precisely on the “non-reproducible” this essay is about; a number of countries are already raising corporate rates, reversing the trend of the 2000s–2010s. Third, consumption taxes: they are hard to avoid, and the rich, spending more, pay the lion’s share. The Economist’s formula: the tax collector will have to look less at how money is earned and more at where it goes.

Abundance has not yet arrived, but the bill for it has already been issued.

A new countervailing force. In Section IV we wrote that the levers of the masses are weakening and that elites will share only out of goodwill. Autumn 2026 suggests a third mechanism, harsher than goodwill and more reliable than a strike: the fiscal hunger of the state itself, spurred on by the bond market. The Treasury needs money, there is nowhere left to take more of it except from capital, and a market that refuses to finance debt cheaply accelerates that choice. The sharing of abundance, it seems, will happen not through generosity and not through a utopian “high income,” but through three mundane channels: taxes on capital and consumption; financial repression with inflation; and, if states fail to cope, crises that will distribute the losses on their own.