In 2011, a retired theoretical physicist from Würzburg published a book that should have detonated mainstream economics. Reiner Kümmel, who had spent thirty years at the intersection of thermodynamics and growth theory, made a claim so simple it sounded like a provocation: energy conversion and entropy production determine the growth of wealth. He called it the Second Law of Economics. Not a metaphor. A mathematical result.
The argument rests on a production function called LinEx — Linear-Exponential — that Kümmel developed as an alternative to the Cobb-Douglas model still taught in every economics department on Earth. Cobb-Douglas treats capital and labour as the primary engines of output. Energy appears as a rounding error — typically assigned a weight of 5% or less, in line with its share of total production costs.
Kümmel’s LinEx function, fitted to sixty years of macroeconomic data for Germany, Japan, and the United States, returns a radically different picture. Energy’s actual productive weight — its output elasticity — is roughly 50%. An order of magnitude larger than its cost share. And labour’s contribution is correspondingly much smaller than its cost share suggests.
The standard reaction from economists: energy is cheap, therefore it cannot be important. Kümmel’s response: that is precisely the error. You are confusing price with productive power. Energy is cheap because we extract it from nature and pay only the extraction cost, not the value of the resource itself. But productive power is determined by physics, not by price. And physics says: nothing moves, transforms, computes, or metabolises without energy conversion. Every joule that enters the economy either becomes useful work — exergy — or dissipates as waste heat. That ratio, and nothing else, sets the frontier of what an economy can produce.
Labour without energy is a corpse. Capital without energy is a sculpture.
— Steve Keen, paraphrasing Kümmel & Ayres
What made the LinEx function dangerous was not the claim itself — biophysical economists like Robert Ayres and Nicholas Georgescu-Roegen had argued similarly for decades. What was dangerous was the fit. When Kümmel plugged energy into the production function alongside capital and labour, the Solow Residual — that mysterious leftover that economists call “technological progress” — essentially vanished. The growth that standard theory attributes to innovation, human capital, and institutional quality turns out, empirically, to be overwhelmingly explained by our ability to channel increasing energy flows into the economic process and convert them into useful work with increasing efficiency.
Technological progress, in this framing, is not an exogenous miracle. It is the downstream consequence of learning to convert energy more effectively. The steam engine. The transistor. The turbofan. The GPU.
Bitcoin: Thermodynamic Money
If Kümmel is right — if wealth is fundamentally a function of energy conversion — then Bitcoin is not an anomaly. It is the purest monetary expression of that principle ever designed.
Proof of Work is thermodynamic commitment made legible. Every block requires the irreversible expenditure of electricity — real energy, converted into heat, dissipated into the environment, with a SHA-256 hash as the only ordered residue. That hash is proof that entropy was produced. It cannot be faked, reversed, or duplicated. The second law of thermodynamics guarantees it.
The Bitcoin network currently draws approximately 25 gigawatts of continuous power — roughly 128 terawatt-hours per year. The hashrate touched 1 zetahash per second in January 2026 for the first time in history. After the April 2024 halving reduced the block reward to 3.125 BTC, the average production cost per bitcoin for public miners rose to around $37,856 — a direct function of electricity price and hardware efficiency.
This is what Kümmel’s framework makes visible: Bitcoin’s value does not come from narrative, adoption curves, or regulatory clarity alone. It comes from the physical fact that producing each coin requires irreversible energy expenditure. Every bitcoin is, thermodynamically, a crystallised unit of entropy production. It is the closest thing the monetary system has ever produced to a physically grounded store of value — not because energy is “backing” it in the gold-standard sense, but because the second law ensures that the work done to create it cannot be undone.
Critics who say “Bitcoin wastes energy” are making the same error that neoclassical economists make when they weight energy at 5%. They are evaluating thermodynamic work by its cost share, not by its productive power. The energy consumed by the Bitcoin network is not wasted. It is converted into the highest-assurance settlement layer on Earth — one that has never been hacked, never been reversed, and never been shut down. That is useful work, in the precise thermodynamic sense.
AI: Thermodynamic Intelligence
Now run the same analysis on the other great energy consumer of the 2020s.
Global data centre electricity consumption reached approximately 415 TWh in 2024 and is projected to double to 945 TWh by 2030. The IEA projects that in the United States, data centres will account for nearly half of all electricity demand growth through the end of the decade. The industry’s new internal metric — tokens per watt per dollar — reveals the underlying truth: AI is not a software phenomenon. It is an energy-conversion phenomenon. Every inference is a thermodynamic act. Every training run is an entropy-producing process that converts electricity into organised information — statistical structure extracted from data, encoded in billions of floating-point weights.
The parallel to Bitcoin is not superficial. Both systems convert raw electricity into a form of order that did not previously exist. Bitcoin converts energy into monetary assurance. AI converts energy into cognitive capability. Both are entropy machines. Both create value by dissipating energy.
And both are exposing the same structural bottleneck: the world does not have enough electricity for what it is trying to build. Combined, AI data centres and Bitcoin mining now consume over 540 TWh per year — more than the annual electricity consumption of France. By 2030, data centres alone are projected to consume as much electricity as Japan. Hyperscalers are forecasting a quadrupling of IT capacity in five years. US grid infrastructure — 70% of which is approaching end of life — was not built for this.
What It Means
The market is evaluating the largest capital reallocation toward energy-to-value conversion in human history using a production function — Cobb-Douglas — that assigns energy a weight of 5%. Kümmel’s empirical result says the correct weight is ten times that. If he is even half right, then power delivery infrastructure, grid modernisation, and on-site energy generation assets are mispriced by a category error, not a valuation error. This is not a question of whether the market has the right multiple. It is a question of whether the market is using the right model.
Bitcoin is the first monetary system whose production cost is denominated in the same unit that drives all wealth creation: energy. Every block is proof of irreversible entropy production. Proof of Work is not an inefficiency to be engineered away. It is a thermodynamic anchor that makes the ledger physically unforgeable. If you understand Kümmel, you understand why this matters.
AI is the largest peacetime reallocation of energy toward a single productive purpose in history. The industrial revolution converted chemical energy into mechanical work. The AI revolution converts electrical energy into cognitive work. The pattern is identical. The scale is unprecedented. The country or bloc that controls the most efficient energy-to-intelligence conversion pipeline will dominate the next economic epoch — exactly as Kümmel’s model predicts.
The second law of economics is not a metaphor. It is a constraint. And constraints, eventually, price themselves in.
Flight Log — Dispatch from Altitude
Every A320 has a parameter called fuel flow. It is displayed on the engine page in kilograms per hour, and it tells you exactly how much chemical energy — kerosene, Jet A-1 — is being converted into thrust at any given moment. Most passengers never think about it. Most economists never think about its macroeconomic equivalent. But every pilot knows: if fuel flow goes to zero, nothing else on the aircraft matters. The flight management computer, the autopilot, the hydraulics, the pressurisation — all of it is downstream of energy conversion. No fuel flow, no flight. The aircraft does not negotiate with thermodynamics.
Kümmel understood this at the level of an entire economy. He looked at GDP the way a pilot looks at the engine page — not as a number produced by capital and labour, but as a number produced by energy flowing through the system. The Cobb-Douglas model is the economic equivalent of an instrument that shows airspeed and altitude but hides the fuel gauges. It tells you where you are, but not what is keeping you there. Kümmel put the fuel gauges back on the panel. And when he did, the Solow Residual — that unexplained “technological progress” that economists had been treating as a free gift — turned out to be fuel flow all along.
The Second Law does not care about your model. It does not care about your policy. It does not care about your cost-share theorem. Entropy increases. Energy converts. And wealth follows — for those who understand what drives it.