All rights reserved exclusively by Germán & Co.
AUTHOR’S NOTE
THE DATES MATTER
Not because dates turn writers into prophets.
They do not.
Hindsight is one of the cheapest forms of intelligence. Prophecy is worse. Anyone who claims an immaculate record in geopolitics is either extraordinarily fortunate or has been editing his own past.
I prefer the untidier truth.
Some analyses prove right.
Some prove partly right.
Some fail.
Some identify the mechanism correctly but misjudge the timing.
Some warnings never become events because circumstances change.
And sometimes the most valuable warning is precisely the one that becomes unnecessary because somebody listened.
There is no satisfaction in watching a hypothesis become a casualty report.
That is why the dates matter.
My earlier articles remain where they were published. They should remain there, with their doubts, imperfections and uncertainties intact.
They should not be retrospectively polished into prophecy.
A warning published before an event has value precisely because it still contains uncertainty.
That uncertainty is evidence of its date.
In May 2023, when Energy Central invited me to its Network of Experts and asked what most concerned me about the future of the energy industry, I answered: “I reiterate I am concerned about the trends and the hasty decisions.” I also argued that new technologies were welcome but that electricity still required firm, reliable power and secure fuel supply. (Energy Central)
Three years later, that elementary observation has become remarkably uncomfortable.
Not because renewable energy failed.
It did not.
Not because nuclear energy became perfect.
It did not.
Not because gas suddenly became virtuous.
It did not.
But because the arrival of artificial intelligence at industrial scale has forced electricity systems to confront something political fashion periodically forgets:
physics does not negotiate.
We live in an extraordinary age of information.
Almost infinite information.
But information is not knowledge.
Indeed, abundance can become its enemy.
A society can know what happened five minutes ago and understand less and less about why it happened.
We scroll rapidly enough to know every explosion.
We rarely travel backwards far enough to understand the fuse.
My working method has therefore become almost embarrassingly simple.
Before moving forwards, I try to move backwards.
From the headline to the history.
From the government to the society.
From politics to anthropology.
From price to resource.
From resource to geography.
From geography to sovereignty.
From sovereignty to fear.
From the speech to the interests surrounding the speaker.
From the battle to the inherited memory that preceded it.
A missile launched today may carry inside it the memory of a coup seventy years ago.
A gas contract may contain the history of an empire.
A prison can sometimes explain a revolution better than a hundred speeches.
Markets have memories measured in quarters.
Civilisations sometimes remember in centuries.
That difference matters.
It also requires accepting contradictory truths.
The Shah modernised Iran.
The Shah also presided over an authoritarian state whose security apparatus imprisoned and tortured political opponents.
Both statements can be true.
The Islamic Revolution emerged partly from resentment against authoritarianism and foreign interference.
The Islamic Republic subsequently constructed an authoritarian system of its own.
Both can be true.
The United States has provided security and prosperity to allies across enormous parts of the world.
The United States has also intervened abroad with consequences that were disastrous.
Both can be true.
Israel possesses formidable intelligence and military capabilities.
Formidable intelligence does not guarantee strategic wisdom.
Both can be true.
Iran is an authoritarian state.
Iran can also practise sophisticated statecraft.
Both can be true.
Wind and solar electricity can be extraordinarily cheap at the point of generation.
An electricity system containing large proportions of variable generation can still require enormous investment in networks, storage, balancing and firm capacity.
Both can be true.
Oil companies can make exceptional profits during scarcity.
Electric utilities can simultaneously make substantial profits by investing more capital.
And the consuming society can become poorer while both do so.
All three can be true.
Reality does not owe us ideological consistency.
This matters because explanation is frequently mistaken for absolution.
To explain why Iran behaves as it does is not to defend the Iranian regime.
To explain the strategic consequences of the 1953 overthrow of Mohammad Mossadegh is not to claim that every subsequent Iranian action was predetermined by Washington.
The historical record is nevertheless clear that British and American officials discussed and organised action against Mossadegh's government; the US State Department's published historical record contains the planning documents for Operation TPAJAX. (Historia del Estado)
Likewise, recognising repression under Mohammad Reza Shah does not absolve the Islamic Republic of its own repression. Evin Prison became notorious under the Shah's SAVAK and continued to imprison political opponents after 1979. (Human Rights Watch)
The gaoler changed.
The prison did not disappear.
That is not equivalence.
It is continuity.
And continuity is often where history hides.
The same method must be applied to energy.
Energy is never merely energy.
A barrel contains geology.
Geology creates geography.
Geography creates strategic value.
Strategic value creates vulnerability.
Vulnerability creates military interest.
Military interest creates alliances.
Alliances create obligations.
Obligations create wars.
And wars return to the barrel.
A kilowatt-hour is equally deceptive.
Hidden inside it may be a gas molecule, a uranium pellet, a transformer, a transmission line, an interest rate, a carbon price, a drought, a battery, a cable beneath the sea, a sanction, a regulatory decision, a war or an election.
The switch conceals the world.
And ultimately somebody pays.
Independence therefore matters.
Not because a writer can be neutral.
Nobody is.
We arrive with biographies, friendships, prejudices, commercial interests, loyalties, disappointments, fears and hopes.
Having a point of view is human.
Having a predetermined verdict is propaganda.
The obligation is not to pretend to possess no preferences.
It is to refuse to allow those preferences to dictate the evidence.
Facts are under no obligation to support my opinions.
Nor my friendships.
Nor my business interests.
Nor the expectations of those who agree with me.
Some things I have written may have cost me opportunities.
That is not martyrdom.
It is merely the price that occasionally accompanies independence.
Analysis should sometimes make its author uncomfortable.
Otherwise it risks becoming public relations.
Independence ceases to be independence the moment the conclusion needs a sponsor.
This essay was written in a period when the United States had once again placed abundant fossil energy at the centre of its political project. In his January 2025 inaugural address, President Donald Trump promised lower prices, increased drilling, greater energy exports and the rebuilding of strategic reserves; his administration explicitly connected expanded domestic energy production with prosperity and national security. (The White House)
The ambition was understandable.
Abundance can be power.
But power has a peculiar weakness.
It can persuade those who possess it that limits no longer apply.
Cheap oil.
Profitable domestic producers.
Europe dependent on American LNG.
Russia weakened.
China contained.
Iran subdued.
Israel protected.
The Gulf secured.
Hormuz open.
Inflation down.
Interest rates down.
The deficit down.
The dollar strong.
No endless wars.
Each objective can be defended individually.
The difficulty begins when they are demanded simultaneously.
At some point strategy ceases to be the selection of priorities and becomes the accumulation of desires.
That is where this book begins.
Not with a villain.
With a limit.
Or rather, with the refusal to recognise one.
There is an old military danger in entering a swamp.
A swamp does not need to defeat the army that enters it.
It forces the army to spend enormous quantities of energy merely to remain where it is.
The greatest danger to American power may therefore not be an enemy powerful enough to defeat it.
It may be the accumulation of ambitions expensive enough to exhaust it.
That possibility has become especially important in 2026. As of mid-August, the conflict involving the United States, Israel and Iran was still disrupting shipping through the Strait of Hormuz; competing American and Iranian claims over the waterway and renewed attacks on commercial shipping continued to influence oil markets. (Reuters)
At the same time, the United States entered the final months of fiscal year 2026 with a cumulative federal deficit of approximately $1.8 trillion through July, while the Congressional Budget Office expects rising interest costs to remain one of the major structural drivers of future deficits. (Oficina de Presupuesto del Congreso)
War has a price.
Energy has a price.
Debt has a price.
Time has a price.
And sometimes the cost of the war itself begins to manufacture arguments for continuing the war.
Which leaves one question running underneath everything that follows:
Which runs out first?
The war?
The oil?
The reserves?
The ammunition?
The money?
The patience of the allies?
Or confidence itself?
The dates are there.
The articles are there.
The facts are there.
The reader may judge them.
The argument begins on the next page.
Germán Toro Ghio Germán & Co. Karlstad, Sweden August 2026
PROLOGUE
GREED BROKE THE SACK
Wars no longer know how to stop.
They begin with something that can be explained.
Security.
A border.
A terrorist attack.
A nuclear programme.
An ally.
A shipping lane.
A threat.
Then the explanation expands.
The border becomes the region.
The region becomes the alliance.
The alliance becomes credibility.
Credibility becomes deterrence.
Deterrence becomes energy.
Energy becomes the strait.
The strait becomes the world economy.
And eventually the war no longer requires an explanation.
It requires fuel.
That is when greed breaks the sack.
Donald Trump returned to the White House promising an age of abundance.
Drill.
Produce.
Export.
Lower the cost.
Fill the reserves.
Use America's extraordinary oil and gas resources to restore industrial power.
The promise was almost magnificently simple.
Energy abundance would produce economic strength.
Economic strength would produce strategic strength.
Strategic strength would produce peace.
The official language called it peace through strength. (The White House)
But molecules have geography.
That was the flaw.
Oil does not merely come out of the ground.
It moves.
Through pipes.
Across borders.
Into refineries.
Onto ships.
Past coastlines.
Through insurance markets.
Through canals.
And sometimes through passages so narrow that the world's greatest military power can suddenly find itself arguing over a few dozen kilometres of water.
The barrel leads to the route.
The route leads to the strait.
The strait leads to the navy.
The navy leads to the enemy.
And the enemy leads back to the barrel.
An empire of molecules contains its own map of possible wars.
Britain understood this more than a century ago.
The romantic memory of empire gave us men in uniforms, desert horses, maps coloured red and, eventually, Lawrence of Arabia.
But Lawrence rode across the desert.
London calculated the subsoil.
The twentieth century would not be powered principally by the horse or the coal bunker.
It would be powered by oil.
The strategic value of Persia, Mesopotamia and the Gulf was therefore never merely territorial.
The empires drew lines across the sand.
The deeper struggle was several kilometres beneath it.
They fought for the subsoil.
And once the subsoil mattered, sovereignty above it became negotiable.
Iran carries this memory unusually deeply.
In 1953, Mossadegh's government was removed after a conflict over oil nationalisation, domestic political instability, Cold War fears and British-American covert action. The declassified American record leaves no serious doubt that British and US officials planned and supported the operation. (CIA)
The Shah returned with Western backing.
Iran modernised.
Cities grew.
Women entered universities and professions in greater numbers.
Infrastructure expanded.
Oil money transformed society.
But modernisation and repression are not opposites.
The state also built SAVAK.
It built fear.
It imprisoned opponents.
Evin became one of the symbols of that machinery. (Human Rights Watch)
Then came 1979.
The monarchy fell.
The portraits changed.
The flag changed.
The prayer changed.
The enemy changed.
But the machinery of coercion did not evaporate.
Evin remained Evin.
The gaoler changed.
The prison did not.
That continuity should destroy one of the laziest ways of understanding Iran.
The United States did not create the ayatollahs.
Iranian history is not an American puppet show.
The revolution contained religion, class, nationalism, resentment, intellectual currents, economic disruption, clerical organisation and profound domestic conflict.
But foreign intervention became part of that memory.
And states remember humiliation differently from markets.
A market remembers last quarter.
A civilisation can remember 1953.
That does not justify what it does with the memory.
But it may explain why the memory matters.
Explanation is not absolution.
Then came another misunderstanding.
Military superiority was confused with strategic completion.
A radar can identify an object.
A satellite can locate a facility.
An intelligence service can find a commander.
A missile can destroy the building.
What none of those instruments can guarantee is an ending.
This distinction matters because modern warfare has become seductively precise.
A target appears on a screen.
A line moves across a map.
A weapon arrives within metres.
The building disappears.
Victory appears almost mathematical.
But strategy begins after the explosion.
What happens tomorrow?
Who replaces the commander?
What does the population remember?
What does the enemy learn?
How much ammunition was consumed?
How quickly can it be replaced?
How many additional targets appeared because the first target was destroyed?
Sophistication does not abolish arithmetic.
Iran has understood another form of arithmetic.
Time.
Washington can count installations.
Israel can count commanders.
Tehran can count time.
That does not mean Iran is invulnerable.
Far from it.
Its military infrastructure can be destroyed.
Its economy can be strangled.
Its allies can be weakened.
Its commanders can be killed.
Its population can become exhausted.
But states do not always define victory by the preservation of every piece.
In chess, the loss of pieces can be tolerated if the king survives.
Survival itself can become a form of victory.
Military supremacy is not the same as military endurance.
And then there is Hormuz.
Hormuz is not an oilfield.
It is more powerful than that.
It is the valve.
A field produces the molecule.
A valve determines whether the molecule moves.
For decades the strategic assumption was simple enough: the United States and its allies possessed overwhelming maritime power and would protect navigation through the Gulf.
But a strait is not sovereign property merely because a navy can dominate it.
Iran occupies one shore.
Oman occupies the other.
Arab producers depend upon it.
China depends upon flows through it.
Japan depends upon flows through it.
India depends upon the broader Gulf system.
Europe depends upon the consequences of whatever happens there.
A great power may protect a door.
That does not necessarily make the door its house.
By August 2026, shipping through Hormuz had fallen sharply from pre-war levels amid attacks, blockades and competing claims about whether the route was open or closed. Reuters reported Brent around $91 a barrel on 19 August, with uncertainty over exports through the strait still supporting prices. (Reuters)
The extraordinary thing is not that oil has already reached two hundred dollars.
It has not.
The extraordinary thing is that the world has spent months discovering just how many mechanisms are required to prevent it.
Strategic reserves.
Alternative routes.
Spare production.
Demand destruction.
Diplomacy.
Naval power.
Insurance.
Rerouting.
Inventories.
Political restraint.
None of them manufactures infinity.
This is where greed becomes something more interesting than avarice.
Greed is not merely wanting money.
For states, greed can be the refusal to select.
Cheap oil and expensive oil.
Cheap oil for consumers.
Expensive oil for producers.
Weak rivals.
Dependent allies.
Open trade.
Controlled trade.
Lower inflation.
Higher strategic leverage.
Less war.
More coercive power.
The ability to close the tap.
And the insistence that nobody else should close it.
At some point the sack breaks because too many incompatible things have been placed inside it.
Perhaps that is why 1913 is more useful than 1939.
The frightening historical analogy is not always the dictator consciously preparing catastrophe.
Sometimes it is a system populated by governments making decisions that appear rational from inside their own capitals.
One mobilisation.
One guarantee.
One alliance.
One ultimatum.
One calculation.
One reaction to somebody else's reaction.
Nobody needs to desire the final disaster.
They merely need to desire something else strongly enough.
The absurd can be created by rational actors.
That will matter later.
Trump wanted, in political mythology at least, to become something like Lawrence of Arabia: the Westerner capable of understanding the desert, commanding alliances and rearranging the map.
The tragedy is that Lawrence eventually discovered that the desert adventure was not simply about freedom.
Behind him stood governments.
Behind the governments stood treaties.
Behind the treaties stood resources.
Lawrence rode across the desert.
London calculated the subsoil.
And today Washington is again discovering that control over the molecule can become a trap.
The man who wanted to become the world's fossil-fuel tsar may discover that the more aggressively fossil energy is weaponised, the stronger the incentive becomes for everyone else to escape from it.
China electrifies.
Europe builds renewables.
States reconsider nuclear.
Batteries improve.
Transport electrifies.
Industry searches for substitution.
A valve possesses enormous power only while everyone continues using the same pipe.
This is the deeper irony.
The fossil empire may contain the mechanism of its own erosion.
The more frightening the molecule becomes, the more valuable the electron becomes.
But the electron has its own problem.
A molecule can be stored.
Moved.
Put in a tank.
Loaded on a ship.
Burned when required.
Electricity is more demanding.
It must be balanced.
Delivered.
Stabilised.
Transmitted.
Stored indirectly.
Maintained second by second.
The molecule monetises scarcity.
The electron manages complexity.
And complexity costs money.
Which leads us away from the battlefield and towards somewhere apparently insignificant.
A kitchen.
A bill.
A switch.
Because every empire eventually meets the household.
And that is where the absurd begins.
CHAPTER I
ABSURD
The Cloud Has Shareholders. The Grid Has Ratepayers.
It begins in a kitchen.
Not in Silicon Valley.
Not in Abu Dhabi.
Not inside a nuclear reactor.
Not at FERC.
Not at an oil terminal.
Not in Washington.
A kitchen.
Someone opens an electricity bill.
It is higher.
Again.
Perhaps not spectacularly higher.
That is precisely how these things often happen.
A few dollars.
A new adjustment.
Transmission.
Capacity.
Reliability.
Infrastructure.
Fuel.
The refrigerator is the same.
The cooker is the same.
The lamp is the same.
The person opening the bill has not purchased ten thousand GPUs.
Has not commissioned a hyperscale data centre.
Has not requested two additional gigawatts.
Has not ordered a transmission line.
Has not asked for another gas turbine.
Has not decided that the United States must win the global race for artificial intelligence.
He merely switched on the light.
And yet the world may already have entered the bill.
THE SHIRT
In May 2023 I was asked what worried me about the energy industry's future.
My answer was not solar.
Not nuclear.
Not oil.
Not gas.
It was trends.
And hasty decisions.
At the time, that answer probably sounded conservative in the original meaning of the word: cautious.
It was not intended to defend any technology.
All technologies were welcome.
The warning was about confusing enthusiasm with architecture.
For a period, green energy became more than an engineering transformation.
It became a social credential.
Governments wore it.
Banks wore it.
Corporations wore it.
Consultants wore it.
Cities wore it.
Investors photographed themselves wearing it.
A company without a net-zero promise seemed almost improperly dressed.
The green transition became, in sections of Western corporate and political culture, something like the fashionable shirt of the season.
Everyone wanted to be seen wearing it.
The problem was never the shirt.
The shirt was excellent.
The problem was pretending it was the entire wardrobe.
Solar became dramatically cheaper.
Wind became a major industrial technology.
Batteries became extraordinarily useful.
None of that is fictional.
The IEA still expects renewables to grow strongly through 2030, with solar supplying much of the increase in global electricity generation; renewables and nuclear together are projected to provide around half of global electricity generation by the end of the decade. (IEA)
The mistake was different.
We confused a generation technology with a power system.
A cheap solar megawatt-hour is not a transmission line.
A wind farm is not frequency control.
A renewable PPA is not necessarily electricity available in the same place at every hour.
A battery is not infinite-duration storage.
Installed capacity is not identical to firm capacity.
An annual energy balance is not a functioning grid at 03:17 in the morning.
The physics never changed.
Only the clothing did.
EVOLUTION, NOT A WARDROBE
Electricity is not a beautiful shirt from H&M, worn brilliantly to one party and replaced before the next.
A power system has no seasonal collection.
It evolves.
A hydroelectric plant built by one generation may stabilise a grid transformed by another.
A forty-year-old transmission line may carry electricity from a solar farm completed six months ago.
A nuclear reactor designed before the internet may supply a data centre containing computers its designers could never have imagined.
A modern battery may provide frequency services to an electrical architecture older than its software engineers' grandparents.
The new does not delete the old.
It inherits its obligations.
Energy transitions are not wardrobe changes. They are evolutionary processes.
And evolution has an awkward characteristic.
It takes time.
Political time is measured in elections.
Electric-system time is measured in decades.
A government can announce a target in minutes.
A parliament can close a technology in an afternoon.
A television presenter can declare an energy revolution before dinner.
A transmission corridor can require a decade.
A nuclear industry requires engineers, regulators, manufacturers, financing and institutional memory accumulated over generations.
Large transformers do not appear because a minister has held a press conference.
The paradox is almost beautiful:
Electricity moves at nearly the speed of light. Electricity systems evolve at the speed of civilisation.
SPAIN
On 28 April 2025, Spain and Portugal went dark.
Ideology arrived before the engineers.
One side declared that renewable energy had caused the catastrophe.
Another behaved as though the generation mix were irrelevant.
Neither position survived serious investigation.
ENTSO-E's final report, published in March 2026, identified a combination of interacting factors including oscillations, gaps in voltage and reactive-power control, different voltage-regulation practices, rapid reductions in output and generator disconnections, producing fast voltage increases and cascading failures. (ENTSOE)
The lesson was not:
renewables do not work.
The lesson was far more important.
Megawatts are not a power system.
Spain did not merely need electricity.
It needed control of electricity.
Voltage.
Reactive power.
Frequency.
Synchronism.
Protections.
Reserves.
Stabilising capabilities.
All the invisible things that do not fit elegantly on a political poster.
For years politicians had counted installed megawatts.
On one April afternoon, physics asked a different question:
Can you hold the system together?
That is the question a grid must answer every second.
THE FUTURE ARRIVED HUNGRY
Then came AI.
The cloud has always been badly named.
There is very little cloudy about it.
The cloud is concrete.
Steel.
Copper.
Silicon.
Transformers.
Substations.
Transmission lines.
Cooling systems.
Backup power.
Batteries.
Gas.
Nuclear.
Water.
And land.
Artificial intelligence may be digital.
Its metabolism is industrial.
US data centres consumed about 4.4 per cent of total American electricity in 2023, and the Department of Energy's cited estimates place their potential share at roughly 6.7 to 12 per cent by 2028. (The Department of Energy's Energy.gov)
The IEA's updated analysis sees global data-centre electricity consumption roughly doubling from about 485 TWh in 2025 to around 950 TWh in 2030, while electricity consumption from AI-focused centres grows considerably faster. (IEA)
In the United States, data centres are expected to account for a very large share of electricity-demand growth through the end of the decade. (IEA)
Then the fashionable shirt met the server rack.
Artificial intelligence does not ask:
How many renewable gigawatts did you install last year?
It asks:
Will the electricity be there?
Tonight.
Tomorrow.
During the heatwave.
When the wind falls.
When cloud cover changes.
When another line trips.
When another hyperscaler asks for another gigawatt.
AI does not merely consume electricity.
AI consumes firmness.
That distinction is worth billions.
Energy and availability are not the same product.
Wind produces energy.
Solar produces energy.
A data centre requires energy plus reliability.
To reconcile those things you need combinations.
Wind.
Solar.
Batteries.
Hydro.
Nuclear.
Gas.
Geothermal.
Demand response.
Transmission.
Interconnection.
Overbuilding.
Control.
There is nothing scandalous about this.
The scandal was ever pretending one component would eliminate the need for the others.
THE BATTERY
Then came the battery.
It is a magnificent technology.
It can respond in milliseconds.
Move solar production into the evening.
Stabilise frequency.
Reduce peaks.
Absorb surplus generation.
Provide reserves.
Support weak networks.
And battery technology will continue improving.
But the battery does not abolish time.
A four-hour battery contains four hours at rated discharge.
Four hours can be priceless.
Four hours is not three days.
Nor does a battery abolish ageing.
Cells degrade.
Temperature matters.
Cycling matters.
Depth of discharge matters.
Systems can be augmented and modules replaced.
But storage does not become immortal because the electrons entering it were green.
This matters economically.
A hydroelectric dam may serve several generations.
A nuclear reactor may operate for many decades.
A major transmission asset may remain useful for forty or sixty years.
Battery systems can require augmentation or replacement during the lives of those assets.
The battery does not make system cost disappear.
It rearranges when and how the cost appears.
The useful question is therefore not:
Are batteries good?
They plainly are.
The useful questions are:
How many?
What duration?
What duty cycle?
At what cost?
With what degradation?
Replaced or augmented when?
Connected through which network?
And what happens after they are empty?
Politics dislikes questions like these.
Engineering lives inside them.
THE ATOM WE THREW AWAY
Europe had experienced another fashion.
Nuclear rejection.
The history must be treated fairly.
The anti-nuclear movement was not owned by one ideology.
Greens were important.
Parts of the European left were important.
Environmental movements were important.
Public fear after Chernobyl mattered enormously.
Fukushima transformed political calculations again.
And conservative governments made closure decisions too.
Fear was not invented.
Nuclear accidents can have terrible consequences.
Waste presents extraordinarily long-term responsibilities.
Some large Western reactor projects have suffered extraordinary delays and cost overruns.
Those truths belong in the argument.
But another truth existed beside them.
Nuclear provided something Europe would later discover it still valued:
large quantities of firm, domestic, low-carbon electricity.
The strategic error was not questioning nuclear technology.
The strategic error, in countries that pursued rapid phase-outs, was removing firm low-carbon capacity before equivalent firm low-carbon replacements existed at adequate scale.
Europe feared the atom.
And parts of Europe increased their vulnerability to the molecule.
Gas.
Germany closed its final reactors in 2023.
Yet nuclear power did not vanish from Europe. In 2024 it still produced 23.3 per cent of all EU electricity. (European Commission)
Now the wheel is rotating again.
The IEA describes renewed support for nuclear power across more than forty countries, driven by energy security, stronger electricity demand and growing demand for stable low-emissions generation — including from data centres. (IEA)
Sweden provides almost comic historical symmetry. After decades in which nuclear phase-out shaped national debate, the Swedish government decided in June 2026 to take a 60 per cent share in Videberg Kraft, a company planning new reactors at Ringhals. (Regeringskansliet)
Yesterday:
nuclear is the past.
Today:
the digital future wants nuclear.
Europe spent decades trying to retire nuclear power. Then the economy of the future arrived asking for precisely what nuclear power was good at providing.
That does not mean every reactor should have remained open forever.
It means electricity has no fashion season.
A reactor cannot be closed in spring and ordered back from the catalogue in autumn.
Industrial memory does not reboot.
Engineers retire.
Suppliers disappear.
Manufacturing capacity moves.
Regulatory expertise erodes.
A parliamentary vote can close a plant.
It cannot manufacture thirty years of accumulated knowledge the following Monday.
BACK TO THE MOLECULE
And so the circle begins closing.
The future arrived demanding electricity.
Not just electricity.
Reliable electricity.
The IEA expects renewables to provide a large share of additional global data-centre demand, while nuclear plays an increasing role; but gas and coal also contribute significantly to meeting the increase, with regional mixes varying substantially. (IEA)
In Europe, the IEA expects renewables and nuclear together to provide most of the incremental electricity required by data centres by 2030. (IEA)
That contradictory truth must remain in the book.
Otherwise the argument becomes ideological.
The conclusion is not:
green failed and gas won.
It is:
The system wants everything.
The future needs wind.
Solar.
Storage.
Firm low-carbon power.
Networks.
Control.
Gas where required.
Hydro where available.
Geothermal where available.
Demand flexibility.
Capital.
Water.
And time.
Lots of time.
Politics likes winners.
Engineering likes portfolios.
Politics demands:
Which technology is the answer?
Physics replies:
It depends.
WATER
The cloud has another hidden dependency.
Water.
Computer chips generate heat.
Heat must leave.
The IEA estimates that cooling can account for anywhere from about 7 per cent of electricity consumption in efficient hyperscale facilities to more than 30 per cent in less efficient enterprise centres, depending upon design. (IEA)
Cooling technologies vary enormously.
Some consume substantial quantities of water.
Others use closed-loop liquid cooling, dry systems or other designs that can reduce direct water consumption dramatically. The US Department of Energy has emphasised both efficiency improvements and closed-loop approaches as ways of reducing water pressure. (The Department of Energy's Energy.gov)
Again, the issue is not technological impossibility.
Engineering can solve extraordinary problems.
The interesting question is what has to be built around the solution.
MEGALOMANIA IN THE DESERT
Perhaps nowhere is this more visible than in Abu Dhabi.
The United Arab Emirates and US partners are developing a 5-gigawatt UAE-US AI campus, with Stargate UAE planned as a 1-gigawatt compute cluster inside it and an initial 200 MW expected to enter service in 2026. The developers say the full campus will combine nuclear, solar and natural-gas power. (G42)
Five gigawatts.
In the desert.
This is not an argument that the Emirates are foolish.
Quite the opposite.
The project demonstrates what an extraordinary concentration of capital, energy, political ambition and engineering can accomplish.
But the setting reveals something profound about our age.
The UAE is one of the world's most water-constrained states and relies heavily on desalination for drinking water; the UAE federal atlas says desalination supplies more than 90 per cent of drinking water. (Atlas de la Unión de los EAU)
So we arrive at an almost perfect physical circle.
The servers require electricity.
The servers produce heat.
The heat requires cooling.
Cooling may require water.
Much of the usable water must be produced.
Producing water requires infrastructure and energy.
The electricity requires generation.
The generation requires fuel, uranium, sunlight, networks, storage and capital.
And the enormously material product produced at the end of the chain is still called:
the cloud.
Nothing could be less like a cloud.
We have begun building machines that can require water in places where water itself must first be manufactured.
Perhaps megalomania does not begin when engineering attempts something difficult.
Human civilisation has always advanced by attempting difficult things.
Megalomania begins when engineering asks Can it be built? and politics forgets to ask:
What must be built around it to make it possible?
The desert once had oil beneath it.
Now we are placing intelligence above it.
And manufacturing parts of the physical ecosystem in between.
Oil financed the desert city.
The desert city finances AI.
AI demands electricity and cooling.
Electricity and cooling send us back to energy.
The circle is complete.
THE PRIVATE LOAD
Now we reach the distinction upon which this chapter stands.
A data centre uses the electricity system.
It is not the electricity system.
A data centre may use the public water system.
It is not the public water system.
Connecting a private industrial facility to public infrastructure does not transform the private facility into public infrastructure.
This seems obvious.
Yet billions can depend upon forgetting it.
Suppose a city has a fifty-year-old substation serving two hundred thousand people.
It must be replaced.
That is public infrastructure.
Suppose a transmission corridor must be hardened against storms or wildfire.
Public resilience.
Suppose a municipal water main serving thousands of households has reached the end of its life.
Public infrastructure.
But now suppose a corporation arrives requiring two gigawatts of continuous power.
The existing system cannot supply it.
New transmission is required.
A substation.
Perhaps generation.
Perhaps gas infrastructure.
Perhaps storage.
Perhaps dedicated water systems.
Billions of dollars.
Then there is one wonderfully simple test:
Would we have built this if the data centre had not arrived?
If the answer is no, then the cost has a cause.
WHO CAUSED THE COST?
Regulators call it cost causation.
A dry phrase.
An explosive principle.
Who caused the cost?
Who benefits?
Who carries the risk?
Who owns the commercial operation?
Who receives the profit?
And finally:
who pays?
There is nothing improper about a utility earning money.
Utilities require enormous capital.
Transmission does not finance itself.
Neither do substations, generation, resilience or storage.
A regulated utility invests.
The regulator determines whether the investment is prudent.
The company earns an authorised return.
Customers finance the system over time.
That mechanism has helped build modern civilisation.
The problem appears when the boundary between the common system and the exceptional private load becomes blurred.
FERC has now recognised the issue explicitly.
On 18 June 2026, the Commission ordered the six regional grid operators under its jurisdiction to justify or reform their tariffs governing data centres and other large loads, framing the initiative simultaneously as a way to accelerate connections and to protect existing customers. (Federal Energy Regulatory Commission)
The accompanying FERC material expressly says the Commission is acting to guard against inappropriate cost shifting among transmission customers, while leaving state regulators responsible for preventing equivalent shifting among retail customers. (Federal Energy Regulatory Commission)
Those two words matter enormously.
Cost shifting.
Because if the danger did not exist, nobody would need rules to prevent it.
THE CLOUD HAS SHAREHOLDERS
Imagine the structure.
The data centre belongs to the corporation.
The servers belong to the corporation.
The models belong to the corporation.
The intellectual property belongs to the corporation.
The customers belong to the corporation.
The revenue belongs to the corporation.
The increase in market value belongs to the shareholders.
Then ask:
Who pays for the transmission line built principally because it arrived?
Who pays for the additional generation?
Who pays for the substation?
Who pays if the forecast proves wrong?
This is not an attack on AI.
Artificial intelligence may produce enormous benefits.
Medicine.
Engineering.
Research.
Productivity.
Education.
Scientific discovery.
It may become one of the most consequential technologies humanity has created.
Precisely because the upside could be enormous, there is no obvious economic principle requiring an ordinary electricity customer to subsidise the infrastructure necessary for somebody else to capture it.
The beneficiary should pay for the incremental cost it causes.
If a hyperscaler requires extraordinary reliability, charge it for extraordinary reliability.
If its arrival requires a dedicated substation, allocate that cost appropriately.
If it causes a transmission expansion principally for its own load, it should carry the corresponding responsibility.
If it brings genuine benefits to the common system, credit those benefits.
If both are true, divide the cost honestly.
This is not anti-capitalism.
It is capitalism with the invoices attached.
The household should pay for the grid it needs. It should not be forced to finance the factory that makes somebody else a billionaire.
VIRGINIA
Virginia has already confronted the question.
Its data-centre concentration is enormous, and the state's Joint Legislative Audit and Review Commission concluded that existing data centres were generally paying their current cost of service.
That fact is important.
It prevents a good argument from becoming dishonest.
The issue is not that data centres simply refuse to pay their electricity bills.
The deeper issue is prospective infrastructure.
JLARC concluded that strong data-centre demand could drive new generation and transmission requirements and estimated that, by 2040, generation- and transmission-related costs for a typical Dominion residential customer could be $14 to $37 per month higher in real terms, depending upon how the system develops. (JLARC)
Thirty-seven dollars.
For a hyperscaler, it is invisible.
For a pensioner, it may be medicine.
Food.
Heating.
A telephone bill.
A meal.
This is how enormous transfers can become politically quiet.
Divide the sum among millions of people.
Spread it over decades.
Give each component a technical name.
No individual line looks revolutionary.
The total can be.
THE RISK THAT CAN LEAVE
There is another asymmetry.
The household is captive.
The hyperscaler is mobile.
A family cannot move its kitchen to another state because a tariff becomes inconvenient.
A data-centre developer can cancel.
Delay.
Reduce.
Move.
Redesign.
Become more efficient.
Discover that projected demand was exaggerated.
But a transmission line built in anticipation of that load does not pack a suitcase.
A substation cannot migrate.
A turbine does not telephone the bank and explain that the customer changed its mind.
A bond remains a bond.
The data centre can leave.
The ratepayer remains.
That is stranded-asset risk.
And it reveals one of the oldest laws of political economy:
the least mobile participant is often the easiest participant upon whom to leave the residual cost.
Regulation exists partly to prevent precisely that.
THE OIL COMPANY GETS THE HEADLINE
Now compare this with oil.
When war drives oil sharply higher, everybody sees the mechanism.
Yesterday:
X dollars.
Today:
Y dollars.
Multiply by millions of barrels.
The producer captures scarcity.
The quarterly results appear.
Politicians speak about windfall profits.
Investors celebrate.
Consumers complain at the pump.
Television draws a chart.
War.
Scarcity.
Price.
Profit.
Visible.
Electricity is quieter.
A data centre appears.
The forecast rises.
Transmission expands.
Generation is contracted.
A substation is built.
Storage is added.
Gas demand changes.
Capital is borrowed.
The utility earns on approved investment.
Tariffs are adjusted.
The cost lives inside an asset for thirty years.
Nobody sees the single moment of transfer.
There may not be one.
It happens through:
depreciation,
financing,
capacity,
transmission,
allowed returns,
future rate cases,
and infrastructure.
The petrol station announces the oil shock in enormous numbers.
The electricity system hides its shock in architecture.
Perhaps that is why we speak so loudly about extraordinary profits earned by oil companies during war and so little about how extraordinary digital demand redistributes the cost of public infrastructure.
AI AND THE MOLECULE
AI alone will not send oil to two hundred dollars.
That would be too simple.
Data centres do not predominantly burn crude oil.
Their direct pressure falls first upon electricity systems and, in important markets, natural gas.
But AI can consume something else.
Spare capacity.
Spare transmission.
Spare gas generation.
Spare transformers.
Spare engineering.
Spare construction capacity.
Spare capital.
Spare time.
Those reserves of flexibility matter.
If an energy system enters a geopolitical shock with fewer cushions, the shock travels further.
AI does not need to be the detonator.
It can become part of the accelerant.
A strait becomes dangerous.
LNG cargoes reroute.
Gas prices move.
Electricity demand does not disappear.
Data centres continue computing.
Utilities seek dispatchable generation.
Markets reprice risk.
The links may be indirect.
Energy systems are networks of substitution.
Pressure introduced in one place can emerge somewhere else.
THE GREAT IRONY
Now assemble the pieces.
Europe feared nuclear power.
Some countries closed reactors.
Europe built enormous amounts of renewable capacity.
Gas remained important to balancing, heating and industry.
Russia invaded Ukraine.
Energy security returned.
Spain and Portugal suffered a blackout reminding everyone that installed megawatts are not equivalent to a stable grid.
AI arrived.
Data-centre demand accelerated.
Governments declared artificial intelligence strategic.
Technology companies looked again towards nuclear.
Utilities looked towards gas.
Storage became indispensable.
Transmission became urgent.
Water entered the discussion.
And consumers were told that vast new investments were required.
Nothing in that chain requires conspiracy.
That is what makes it fascinating.
The environmentalist wants decarbonisation.
The utility wants reliability.
The technology company wants computing capacity.
The government wants strategic dominance.
The investor wants return.
The gas producer wants demand.
The consumer wants an affordable bill.
Every desire is rational.
The sum can become absurd.
The absurd is not always created because everybody behaves irrationally. Sometimes it is created because everybody behaves rationally inside a system whose final result nobody owns.
THE TREND
And so we return to 2023.
Trends.
Hasty decisions.
The problem with a trend is not that it is necessarily wrong.
The problem begins when questioning it becomes socially unacceptable.
Energy technologies should never become religions.
Nor demons.
Solar is not virtuous.
Gas is not wicked.
Uranium has no ideology.
A battery does not vote.
An electron has never attended Davos.
A turbine has never read an ESG report.
At 03:17 in the morning, the system has only one question:
Can you deliver the power?
Correct voltage.
Correct frequency.
Now.
You can lie to voters.
You can lie to shareholders.
You can lie to financial markets for a while.
It is considerably harder to lie to frequency.
THE LAST LINE
And now we return to the kitchen.
The bill is still there.
The person reading it does not know what reactive power is.
He has never attended a FERC hearing.
He has never read a utility rate case.
He does not know how many gigawatts a hyperscaler has requested.
He does not know what happened in Spain on 28 April 2025.
He does not know which reactor was closed two decades earlier.
He does not know when the battery will need augmentation.
He does not know how the data centre cools itself.
He does not know whether the water hundreds of kilometres away was pumped from a river, recycled or desalinated.
He does not own the gas field.
He does not own the transmission company.
He does not own the data centre.
He does not own the algorithm.
He does not own the intellectual property.
He owns a refrigerator.
A cooker.
A lamp.
Perhaps an electric heater.
And an obligation to pay before the due date.
Above him, everyone possesses some mechanism for capturing value.
The producer captures scarcity.
The utility earns a return on approved capital.
The lender earns interest.
The data centre monetises computing.
The AI company monetises intelligence.
The shareholder captures appreciation.
The government calls the whole enterprise strategic.
And then the cost starts moving.
Through fuel.
Through generation.
Through storage.
Through transmission.
Through substations.
Through water.
Through financing.
Through regulation.
Through the tariff.
Down.
Down.
Down.
Until it reaches the participant who cannot send the invoice anywhere else.
The household.
The last line.
That is ABSURD.
The cloud has shareholders. The grid has ratepayers.
And the water system has households.
The profits may be private.
The infrastructure must not therefore automatically become socialised.
A public grid should modernise for the future.
It should not become venture capital for private fortunes.
The household should pay for the electricity system it needs.
It should not be forced to finance the factory that makes somebody else a billionaire.
The man looks once more at the number on the bill.
Then at the lamp above the kitchen table.
He switches it off.
The data centre does not.
EPILOGUE
$200
Nobody knows whether oil will reach two hundred dollars a barrel.
I certainly do not.
That is not the question.
The question is:
Why does the number no longer sound absurd?
As this book is being completed in August 2026, Brent is nowhere near two hundred dollars; it was trading around the low nineties on 19 August. But shipping through the Strait of Hormuz remains severely disrupted and markets continue pricing the possibility that the conflict may again restrict Gulf exports. (Reuters)
Two hundred dollars is therefore not a forecast.
It is a stress test.
A question placed at the end of the system.
A barrel does not reach two hundred dollars because somebody in Houston wakes up greedy one morning.
It reaches extraordinary prices when enough cushions disappear at the same time.
A route closes.
A pipeline burns.
A refinery fails.
Inventories fall.
A strategic reserve shrinks.
Spare production becomes uncertain.
Insurance prices war.
Tankers reroute.
Gas becomes expensive.
Governments panic.
Traders anticipate governments panicking.
Consumers hoard.
The market begins purchasing tomorrow's fear today.
And eventually price ceases to be the cause of the crisis.
It becomes the measurement of everything that failed before it.
The United States entered the 2026 Gulf crisis with an enormous oil industry but finite strategic stocks.
In March 2026, Washington authorised a release of 172 million barrels from the Strategic Petroleum Reserve as part of a coordinated international response to the energy shock. (The Department of Energy's Energy.gov)
By the week ending 7 August, US SPR inventories had fallen to roughly 299 million barrels, according to reported EIA data — a reminder of the distinction between producing oil and possessing emergency time. (The Wall Street Journal)
Strategic reserves do not manufacture petroleum.
They manufacture time.
Production provides strength.
Stocks provide something different:
time to decide.
And time, like oil, can run out.
China has spent years thinking about inventories and routes.
Japan understands what import dependence means.
Europe has learnt brutally what energy dependence means.
Russia understands that geological reserves become strategic power only when accompanied by wells, pipelines, ports, refineries, tankers and customers.
Venezuela demonstrates the opposite lesson.
Oil beneath the ground is potential.
Oil connected to electricity, diluent, wells, processing, storage, pipelines, ports and buyers is power.
Geology is not infrastructure.
A reserve is not a barrel.
A barrel is not a route.
A route is not guaranteed passage.
And passage is not sovereignty.
Every stage creates another vulnerability.
Now add debt.
The United States is not financing strategy from an empty balance sheet.
The federal deficit reached approximately $1.8 trillion in the first ten months of fiscal year 2026, and CBO projects persistent large deficits over the coming decade, increasingly burdened by net interest. (Oficina de Presupuesto del Congreso)
This creates a circularity.
War costs money.
War can raise energy prices.
Higher energy prices can sustain inflation.
Persistent inflation can keep interest rates higher than they otherwise would be.
Higher interest rates increase debt-service costs.
Higher interest costs enlarge deficits.
Larger deficits require more borrowing.
And more borrowing reduces the room available for the next crisis.
At some point the cost of defending power becomes one of the forces that weakens the financial foundations beneath the power.
This does not mean the United States is bankrupt.
It means arithmetic remains undefeated.
Sophistication does not abolish arithmetic.
Neither does empire.
At fifty dollars, oil is a commodity.
At one hundred, it becomes politics.
At one hundred and fifty, it begins entering almost every conversation.
Inflation.
Transport.
Aviation.
Fertilisers.
Food.
Petrochemicals.
Freight.
Interest rates.
Currencies.
Subsidies.
Budgets.
Elections.
At two hundred dollars, the barrel becomes something else.
It becomes a referendum on the architecture of the world we built.
Not because two hundred contains mystical significance.
Because sufficiently extreme prices reveal where resilience ended.
They reveal whether abundance was real or conditional.
Whether diversification was real or rhetorical.
Whether reserves were adequate.
Whether alliances could endure.
Whether governments could protect households.
Whether debt mattered.
Whether the producer, utility, investor and state could continue passing costs downwards without eventually meeting political resistance at the last line.
Somewhere, perhaps, a number appears on a screen.
$198.70.
Or $201.30.
It does not matter.
Numbers such as these are crossed long before markets print them.
They are crossed politically.
Strategically.
Militarily.
Financially.
One decision at a time.
Behind the price stands the barrel.
Behind the barrel stands the tanker.
Behind the tanker stands the strait.
Behind the strait stands Iran.
America.
Israel.
Oman.
Saudi Arabia.
The Emirates.
China.
India.
Japan.
Behind them stands the war.
Behind the war stands history.
Behind history stands empire.
Behind empire stands oil.
And beneath all of it:
the subsoil.
We return to where we began.
Lawrence crossed the desert.
London calculated beneath it.
A century later we possess satellites, algorithms, hypersonic weapons, artificial intelligence and machines capable of processing quantities of information unimaginable to those old imperial planners.
But somewhere beneath all that technology remains a molecule.
And somewhere beside it remains a human weakness.
The belief that possessing power grants permission to want everything.
Perhaps oil never reaches two hundred dollars.
We should hope it does not.
Because if it does, the important question will not be:
Who predicted it?
The important question will be:
Who created the conditions that made it possible?
Was it the war?
The sanctions?
The blockade?
The producer?
The consumer?
The failure to diversify?
The dismantling of infrastructure?
The obsession with control?
The belief that every strategic objective could be pursued simultaneously?
Or simply the oldest error in economics and politics:
the refusal to recognise a limit?
The answer may be all of them.
Which returns us to the sack.
Greed did not make the barrel expensive by itself.
Greed did something more dangerous.
It kept placing desires into the same sack after the sack was already full.
Cheap energy.
High producer profits.
Military dominance.
Technological dominance.
AI dominance.
Control of trade routes.
Low inflation.
Low interest rates.
Low deficits.
Weak enemies.
Dependent allies.
No sacrifice.
No limits.
Eventually the sack breaks.
And when it breaks, those at the top may still possess oilfields.
Utilities may still possess assets.
Technology companies may still possess algorithms.
Banks may still possess bonds.
Governments may still possess aircraft carriers.
But somewhere, in a kitchen, somebody is still opening the bill.
That person was there at the beginning of this book.
He is here at the end.
The last line.
Perhaps he has already switched off the lamp.
The server is still running.
The tanker is still moving.
The missiles are still flying.
The debt is still accumulating.
And beneath the desert, indifferent to every ideology constructed above it, the molecule waits.
The question is not whether we have enough power.
The question is how much power we can afford to want.
End
All rights reserved exclusively by Germán & Co.
