Why This Topic
This series started with one question. Suppose the law treated a data center like a plain warehouse of the same size. It could draw only a warehouse’s share of grid electricity, city water and aquifer water, and it would get no tax break written for its industry alone. Everything else, it would have to generate, find or buy at full price. Could it still make money?
Five episodes took the question apart: whether states should pay industry to come, who actually pays the bill, who owns the water and air, whether farms can be taken for power lines, and what the remedy should be. The finale puts it back together. We gave it to Walter Bagehot and Alexander Hamilton, because one of them spent his life watching investment manias and the other wrote the founding American case for helping new industries grow.
Here is the ledger Bagehot reads from. These figures are per megawatt of capacity per year for a large AI campus. The revenue and cost come from published estimates. The cost of the rule is our own arithmetic, built from the sources in the earlier episodes, and Bagehot calls it that on air.
LinePer megawatt per yearRevenue, industry estimate10 to 12 million dollarsTotal cost, independent modelabout 8.5 millionMargin1.5 to 3.5 millionCost of the warehouse ruleabout 0.45 to 1.7 million
The cost of the rule is made of four pieces: making power on site from gas instead of buying it from the grid, cooling without water, paying sales tax on the servers like any other business, and paying full property tax. On the best campuses that takes a dent out of the margin. In the middle it takes about forty percent. At the worst, it takes all of it.
The picture changes for everyone else. Older data centers that rent space to many customers earn roughly a quarter as much per megawatt, so the same costs would wipe out a normal return. Facilities inside cities have no room to build their own power plants at all. And many of the projects that make up the boom have not been built yet. When Ohio asked developers to sign binding contracts for the power they wanted, most of the forecast demand disappeared. That is the evidence the finale turns on: whether the rule would end an industry, or only the speculative part of it.
Why Walter Bagehot
Bagehot edited The Economist for seventeen years and wrote Lombard Street, the 1873 book that became the handbook for how central banks handle panics. He watched Victorian Britain throw money at railways and other world-changing ventures, many of which did change the world while ruining their investors. His most famous line on manias is that “John Bull can stand many things, but he cannot stand two per cent.” When safe returns are low, money chases anything that promises more.
We kept Bagehot’s established character from our Bagehot vs Bryan debate on cryptocurrency: a Christopher Hitchens style performer, cool and exact, generous about a good hit and most dangerous when quietest. His opening line sets the tone: “A business that cannot pay for its own electricity is not an industry. It is a mood.”
Why Alexander Hamilton
Hamilton’s 1791 Report on Manufactures argued that a young country could not wait for its industries to grow up on their own, because established nations would crush them first. He recommended government encouragement, including bounties paid to new manufacturers. It is the founding document of American industrial policy, and every modern argument for subsidizing a strategic industry descends from it.
We kept his established character from our Jefferson vs Hamilton debates on data center secrecy: clipped, impatient and hot, with a pomposity he knows about and resents. Any suggestion that he is an aristocrat sets him off. Bagehot does it deliberately, and Hamilton’s angry reply, about arriving as an orphan from Nevis, is one of the best moments in the episode.
Who Else We Considered
Adam Smith vs Alexander Hamilton. The classic free trade against industrial policy pairing. We have already run Hamilton vs Smith on tariffs, and Smith would have argued principle where we wanted someone who would argue the numbers.
Walter Bagehot vs Leland Stanford. The banker against the railroad baron would have been a fine fight about manias. Stanford already had Episode 1, and we wanted the finale to include someone who could defend public support on principle rather than self-interest.
Friedrich List vs Walter Bagehot. List wrote the German version of Hamilton’s argument and attacked British free trade directly. He would have been a new voice, and Hamilton’s open-bounty position gave us a better ending.
Why Each Man Takes the Position He Does
Bagehot’s case is that the rule does not kill the industry but sorts it. The best AI campuses can absorb the cost. Older buildings that rent space to many tenants earn far less per megawatt and cannot. Data centers inside cities cannot build a gas plant in the parking lot. His strongest evidence is Ohio: when one utility required data centers to pay for 85 percent of the power they reserved for up to twelve years, its forecast of new demand fell from about 30,000 megawatts to about 5,700. “The rule does not destroy the industry,” he says. “It destroys the mania.”
Hamilton’s case is the infant-industry argument. New industries face early costs they will never face again, nations that wait are overtaken, and the help is repaid in taxes. Loudoun County’s 1.2 billion dollars a year from data centers is his best exhibit.
Bagehot gets his shot at Hamilton’s record. Hamilton’s own industrial venture, a planned manufacturing town at the Great Falls of Paterson, New Jersey, struggled almost from the start. Paterson later recovered, and Bagehot notes it went on to produce William Graham Sumner, who debated in Episodes 2 and 5. Sumner really was born there.
The turning point is Hamilton’s concession. His Report favored open bounties, and in the debate he draws the line we think the whole series was heading toward. He will defend help that is voted by a legislature and visible in the public accounts. He will not defend a subsidy hidden in a widow’s electric bill, buried in a sealed contract, or slipped through a capacity market. “If the nation wants these buildings,” he says, “let Congress vote a bounty in daylight and put its name to it.”
Bagehot makes concessions of his own. He agrees the warehouse yardstick is crude. Jeremy Bentham asked in Episode 3 why a warehouse and not a steel mill or a hospital, and Otto von Bismarck pointed out in Episode 4 that forcing data centers to make their own power puts gas turbines next to the neighbors. Bagehot’s revised rule is simpler: “The honest rule is not be a warehouse. It is pay what you cost.”
He also admits what the ledger cannot show. The revenue figure is the industry’s own, and the real contracts are sealed. If true revenue is ten million rather than twelve, the costliest campuses are under water. And the chips wear out in about five years while the power lines last thirty, so somebody is holding the difference.
A Note on the Sources
The revenue and cost figures come from Cleanview, Epoch AI and industry cost analyses, and the gas power premium comes from Rabobank. The cost of the rule is our own estimate, built from those sources and laid out in the research file behind the series. Bagehot’s lines draw on Lombard Street, Hamilton’s on the Report on Manufactures. A few details, including Bagehot’s marriage into the family that founded The Economist and the precise fate of the Paterson venture, are from general knowledge and flagged in our notes for checking.
The commentary booth closes out the series. Samuel Johnson reads his Dictionary definition of a stockjobber, “a low wretch who gets money by buying and selling shares in the funds.” Niccolo Machiavelli applies for a job at Hamilton’s Treasury and learns that his application has been forwarded to Thomas Jefferson. Then they read their full scorecards for all six bouts, disagree on nearly every fight, and deliver their verdicts. Johnson: the strong campuses could pay, so they ought to. Machiavelli: they could pay, which is exactly why they never will.
The one thing both commentators agree on, and the one thing every debater in the series agreed on, is that the contracts are sealed and nobody outside can check the sums. As Johnson puts it: “Open the drawer, sir, and half this quarrel settles itself.”
What Comes Next
That is the end of the series. Thank you for watching all six.
If you want the next series as it drops, subscribe here and on YouTube.
Sources
Axis Intelligence: AI data center cost per MW, citing Epoch AI
Rabobank: Data centers are building a parallel energy system
POWER Magazine: Regulator approves AEP Ohio’s data center tariff
Ohio Capital Journal: AEP Ohio says new data center tariff is working
Every debate on PhilosophersTalk.com is made with AITalkerApp.com. Write a script, pick your voices, and it turns the conversation into an animated video. If you want to make your own, start at AITalkerApp.com.