George vs. today
Each Book of Progress and Poverty, its 1879 example set beside the compute economy, and where the comparison holds, bends or breaks.
George built his argument out of things he had seen: whaling crews, gold diggings, a settler on open prairie, a redwood forest held off the market. These notes take each Book’s main example, find the nearest thing in the compute economy, and then say where the comparison fails. Some of them fit well. Some barely fit at all, and I say so.
Two rules hold throughout. The first is the one from the essays: blame the game, not the players. Nobody here is a villain for collecting what the rules allow. The second is that Georgism should be judged by what it measurably does. I don’t accept the defense that it has never really been tried. Parts of it have been tried, in Pittsburgh and other Pennsylvania towns, and the results are modest and measurable, which is more useful than a perfect idea nobody can check.
Page numbers are those of this site’s PDF edition, which are also the printed page numbers in it. Figures are as of October 2026 and most come from Who Collects the Compute Rent?, where each one is sourced and dated. Each note links to its session in the reading group guide.
The puzzle: poverty beside progress
George's example, 1879
George opens with a thought experiment. Show a man of the eighteenth century, a Franklin or a Priestley, the steamship, the railroad, the reaping machine and the factory, and he would expect poverty to vanish. It hadn’t. George’s puzzle is that want was deepest where wealth was greatest, in the big cities and the old countries.
“This association of poverty with progress is the great enigma of our times.”
The compute era
The productive leap of our time is easy to name. Epoch AI estimates that the performance per dollar of new AI chips has grown about 49% a year, and that the price of reaching a fixed level of model performance has fallen between 9 and 900 times a year, depending on the task. George’s question carries over intact: when the power to produce grows that fast, who ends up with the gain?
Where it's weak
George’s puzzle was that poverty got worse as progress went on. On a world scale that is not what has happened since. The World Bank counts the share of people in extreme poverty falling from 38% in 1990 to 8.5% in 2024, at $2.15 a day. The honest version of his question today is about shares and about places left behind, not about progress making most people poorer.
Book I: wages come from the work, not from capital
George's example, 1879
To show that wages are not advanced out of a stock of capital, George takes the American whaler, whose crew are paid a “lay,” a share of the catch. The oil and bone in the hold are the crew’s wages and the owners’ profits at once. He adds the egg gatherers on the Farallone Islands, paid in coin for eggs already sold.
“she carries in her hold the wages of her crew, as well as the profits of her owners”
The compute era
The compute economy runs on the opposite timing. Engineers at a new AI lab can be paid for years before the models they build earn anything, out of money raised from investors. The cloud builders do the same with machines: in the second quarter of 2026 Alphabet spent $44.9 billion on property and equipment against $39.1 billion of operating cash flow, and Amazon spent $54.2 billion against $45.4 billion.
Where it's weak
This is where the parallel breaks, and it breaks against George. His examples work for a short voyage or an egg season. They work much less well when the voyage lasts years and the catch is uncertain. That was Taussig’s objection in 1896, and the compute economy is close to a pure case of it. I think Book I is best read as clearing away a bad theory of its day, not as a description of how a frontier lab pays its staff.
Book II: scarcity blamed on nature
George's example, 1879
George tests Malthus on California. After the gold rush the rich diggings ran out and wages fell, which looks like nature growing stingy as population grew. George argues the reverse: roads, wharves, railroads, telegraphs and machinery had made each worker far more productive than in 1849, so falling wages needed another explanation.
“Instead of the wealth-producing power of labor being less in California in 1879 than in 1849, I am convinced that it is greater.”
The compute era
The Malthus of compute is the claim that there simply isn’t enough power to go around. Part of that scarcity is made by rules, not by nature. Lawrence Berkeley National Laboratory found that the median power project built in 2025 took five years from interconnection request to operation, against under two years for projects built from 2004 to 2008. Generators were not five years harder to build. The queue was longer.
Where it's weak
Some of the scarcity is physical. Power plants, transmission lines and chip factories take years to build no matter who owns the ground, and I can’t say how much of the queue is paperwork and how much is real engineering. George’s own evidence here was a single state and his own observation, which is thin by any modern standard.
Book III: rent is set at the margin
George's example, 1879
George adopts Ricardo’s law of rent and makes it the key to his whole system. The poorest land anyone needs to use sets the baseline. What better land yields above that baseline goes to its owner as rent, whether the owner works it or not.
“The rent of land is determined by the excess of its produce over that which the same application can secure from the least productive land in use.”
The compute era
The compute version of the margin is the worst capacity anyone still has to use. Above it sit very large margins. Amazon Web Services earned a 39.4% operating margin in the quarter ended June 30, 2026, and Nvidia reported a 75.0% gross margin on $96.2 billion of revenue. Those are the kind of returns above the margin that George taught readers to look for.
Where it's weak
George’s own categories cut against an easy reading. In Book III, Chapter IV he separates rent from interest, from the wages of management and from the profits of monopoly, and a cloud margin mixes all four. Company accounts don’t split them, so I can’t tell you how much of a 39.4% margin is rent in George’s sense. Anyone who says they can is guessing.
Book IV: the settler on the savannah
George's example, 1879
George’s best known story. A settler stops on an empty, endless prairie where every acre is as good as every other. As neighbors arrive, then a village, then a town, his land is worth more and more, though it grows no more wheat than before. The value comes from the people around it. George adds that it makes no difference what the land is like; what counts is that people gather there.
“a filled-in marsh like that of St. Petersburg, or a sandy waste like the greater part of San Francisco.”
The compute era
Northern Virginia is the savannah settled. The land under the cloud is mostly ordinary ground, and its value comes from the fiber, the substations and the other data centers already there. The home builders’ trade group tallied a sale of 97 acres in Loudoun County for $615 million, about $6.3 million an acre, against a county median near $125,000. CBRE counted vacancy of 0.2% in Northern Virginia in the first half of 2026. That is the settler’s story with power lines in place of neighbors.
Where it's weak
Two cautions. The Loudoun figure is a set of deals reported by an interested party, not a price index. And George’s general law, that progress pushes land values up, did not hold for a long stretch after he wrote. Knoll, Schularick and Steger found that across 14 advanced economies real house prices stayed roughly flat from the late nineteenth century to the middle of the twentieth, with residential land prices broadly constant despite strong growth in population and income. Rising land prices then explain about 80% of the boom after the Second World War. The law describes some periods well and others not at all.
Book V: speculation and the bust
George's example, 1879
George blames recurring depressions on land speculation: owners hold land for the price it will bring later, rents run ahead of what production can pay, and work stops. He writes of San Francisco’s unemployed beside unused land, and he dismisses the popular blame on railroad overbuilding as far too small a cause.
“to assign to this wasting of capital such a widespread industrial dead-lock seems to me like attributing an unusually low tide to the drawing of a few extra bucketfuls of water.”
The compute era
The obvious risk today is the one George waved away: overbuilding. Alphabet and Amazon each spent more on property and equipment than their operating cash flow in the second quarter of 2026 (Alphabet, Amazon). If demand falls short, today’s scarcity premium could turn into a write-down, as the numbers page says.
Where it's weak
Book V doesn’t map onto compute, and I won’t force it. George’s mechanism is good land held idle. The evidence points the other way: with 0.2% vacancy in Northern Virginia and most new space preleased, powered sites are being used, not withheld. If a bust comes, it will look more like the railroad mania George dismissed than the land speculation he blamed. That would count against Book V, not for it.
Book VI: remedies that miss the cause
George's example, 1879
George reviews the remedies of his day: thrift in government, education and hard work, trade unions, cooperation, government regulation, and breaking up large estates. Each, he argues, helps a little or helps some people, but none touches the cause, because whatever wages gain, rising rent takes back. Then he states his own remedy in six words.
“We must make land common property.”
The compute era
The compute era has its own list of remedies that may miss the ground. Open weights are the best example: a real gift, but as the essay puts it, a commons you can’t run is a museum. Meanwhile the scarce input that sets the price is often the grid connection, and PJM’s market monitor puts the added cost of data center load to other customers at $23.1 billion over three capacity auctions.
Where it's weak
George’s remedy only reaches land. Chips, buildings, cooling plants and software are things people made, and George would leave them untaxed. So even if you accept every word of Book VI, it reaches only the ground, the power allocations and the water rights under compute, not the margins at the chip and cloud layers. I suspect that is the smaller share of the money.
Book VII: the miners' rule
George's example, 1879
For an American case of land held in common, George points to the California gold fields. Faced with a new kind of land, the miners set aside the usual land system. A claim lasted only as long as it was worked, and no one could take more than he could use. The government and the courts went along while placer mining mattered.
“it was by common consent declared that this gold-bearing land should remain common property, of which no one might take more than he could reasonably use, or hold for a longer time than he continued to use it.”
The compute era
Compute rests on public grants of the same sort: zoning approvals, water rights and places in the grid queue. In May 2026 Box Elder County, Utah, approved a 40,000 acre data center project, and in October a water rights application tied to it proposed moving water from irrigation to the campus. Both are described, with sources, in Compute Is the Soil. The miners’ question fits: should a public grant last only as long as it is used, and return something to the public that gave it?
Where it's weak
The miners’ rule worked for claims one man could work with a pan and a sluice. I suspect a campus of that size needs long and secure tenure to be financed at all, and a claim that lapses with disuse is a hard thing to lend against. Book VII’s moral argument is also the most contested part of the book; Pope Leo XIII’s Rerum Novarum defended private property in land as a natural right in 1891, as session 5 of the guide notes.
Book VIII: tax the land, spare the machines
George's example, 1879
George turns the principle into a tax. Owners keep their titles and the public takes the rent through a tax on land values, with all other taxes abolished. He is careful to separate land from other monopolies: he would not tax patents, and he notes that the value of a railroad or telegraph line mixes monopoly with real labor and capital.
“But all other monopolies are trivial in extent as compared with the monopoly of land.”
The compute era
Loudoun County does nearly the reverse. Its fiscal 2027 budget expects $879.1 million from its tax on data center computer equipment and $417.7 million from real property tax on data centers, and computer equipment pays $4.15 per $100 of value against $0.805 for real property. George would tax the location and spare the machines; Loudoun taxes the machines at about five times the rate.
Where it's weak
George’s line that other monopolies are trivial beside land is the sentence in this Book that has aged worst. Nvidia’s margin and the cloud margins are, in his terms, closer to his railroad and patent cases than to land, and a land value tax doesn’t reach them. Where his tax has been tried in pieces it has also been hard to keep. Pittsburgh gave up its split rate after a 2001 reassessment, and Altoona dropped its land-only tax in 2016; session 6 of the guide has the sources.
Book IX: the spring released
George's example, 1879
George promises that ending taxes on industry would release production like a weight lifted from a spring. His examples are the taxes his readers knew: on the house a man built, the ship, the factory and the railroad.
“if a railroad be opened, down comes the tax-collector upon it, as though it were a public nuisance”
The compute era
Data centers are the factories and railroads of this list, and they are taxed heavily as he describes: Loudoun’s tax on the computers inside them brings in about twice what its real property tax on data centers does. Yet Northern Virginia still had 0.2% vacancy, and more than 80% of the 7,481 megawatts under construction across North America was preleased.
Where it's weak
This is the weakest Book to test, and the evidence points against its size. A heavy tax on the machines has not visibly stopped building where the location is good enough. Where land taxes have been raised relative to building taxes, the measured effects are real but modest: more and denser housing in Pennsylvania cities, in Banzhaf and Lavery’s study, and little effect on new businesses, in the studies the Chicago Fed reviewed. Neither looks like a spring released.
Book X: the barons and the decline
George's example, 1879
George ends with a theory of history. Progress comes from people associating on equal terms; when inequality grows, civilizations decay. He saw the signs in his own America, in private citizens who controlled railroads, land and livelihoods, and who chose governors and senators.
“We have simple citizens who control thousands of miles of railroad, millions of acres of land, the means of livelihood of great numbers of men;”
The compute era
Concentration is the clearest parallel. Epoch AI estimates that five hyperscalers held 71% of the world’s AI compute at the end of 2025, up from 63% in early 2024, and Synergy put Amazon, Microsoft and Google at 28%, 20% and 15% of the cloud market. The compute barons hold the chokepoints of this economy the way George’s barons held the railroads.
Where it's weak
George’s barons named governors. I have no evidence of that kind about compute, and I won’t imply it. His larger forecast, that inequality would turn civilization back toward barbarism, has not been borne out by the long fall in extreme poverty since 1990. The concentration is worth watching. The prophecy is not evidence.
These notes are written for this site by GANDALFO. George's words are quoted from his original text and linked to their paragraphs; page numbers are those of the PDF edition. Outside figures link to their sources and are as of October 2026.