The Compute Rentier

The oil was never the point.

The point was cheap energy that nobody else could reach. That advantage did not die with the oil age. It changed what it converts into.

The axiom

The Gulf's oil advantage was never the oil. It was cheap, abundant, stranded energy. A barrel in the ground is worth nothing until you can move it and sell it. The Gulf's edge was never the resource. It was the cost of turning the resource into something the world wanted, at a price no one else could match.

Read the century that way and the next move is obvious. Find the new thing the world wants that only cheap energy can produce.

That thing is intelligence.

Barrels into tokens

AI has one binding constraint, and it is not talent and it is not data. It is energy. A large model is a machine for turning electricity into inference. Training burns power. Serving burns power. Every token you read from a model was paid for in kilowatt-hours. The chip gets the headlines. The chip is useless without something to feed it, and what feeds it is a power bill.

So the scarce input in AI is not compute. It is the energy that compute consumes. Compute is downstream of watts.

Put that next to the Gulf. Industrial electricity in Saudi Arabia runs in the range of five to eight cents per kilowatt-hour, against a US average closer to nine to fifteen. Roughly sixty percent of Saudi power comes from natural gas the country already holds. The Aramco chief executive has said it plainly: cheap energy is what will make the kingdom a leader in AI data centres. He is not selling a vision. He is reading a cost curve.

The petrostate becomes the compute state. Same logic, new output. Own the cheap input everyone else needs, and price the ones who do not have it.

The rentier, redrawn

The oil rentier did one thing. He owned a cheap input the whole world required, and he collected on the gap between his cost and everyone else's. He did not need to be clever. He needed to be cheap and to be there.

Call the new version The Compute Rentier. Same position, moved one link down the chain. He does not sell the barrel. He does not even sell the electricity. He sells what only cheap electricity can make at scale: intelligence, metered, by the token. The rent is the spread between the cost of a kilowatt-hour in Riyadh and the cost of the same kilowatt-hour in Virginia or Frankfurt, multiplied across every model run on his soil.

This is why HUMAIN is not a vanity project. Launched by the Public Investment Fund in May 2025 and chaired by the Crown Prince, with Aramco taking a minority stake and PIF holding the majority, it is the state wiring its oldest advantage into its newest market. The chip agreements follow the same logic. Eighteen thousand Nvidia Blackwell chips in the first tranche, plans reaching into the hundreds of thousands, five hundred megawatts of capacity, a ten billion dollar tie with AMD. The consensus reads those numbers as capital buying its way in. Sovereign cheques and chip deals. A rich state renting relevance.

That reading is the mistake.

The inversion

Capital is not the durable advantage. Capital is abundant. Every sovereign fund on earth can write the cheque, and many are. If buying chips were the game, the Gulf would have no edge over Singapore or Japan or a US pension fund. A cheque is not a moat. Anyone can be outbid.

The durable advantage is the arbitrage. Cheap, abundant energy that cannot be moved is the one thing the buyers of chips cannot replicate by spending more. You can ship a chip anywhere. You cannot ship a gas field. The energy is stranded, which is precisely why it is cheap, which is precisely why it is an edge. Stranded energy has no other buyer. Compute is the buyer that finally showed up at the wellhead.

The Gulf is not capital buying its way into AI. The Gulf is energy that AI is buying its way into.

So invert the frame. The Gulf is not capital buying its way into AI. The Gulf is energy that AI is buying its way into. The demand travelled to the supply, the way it always does when the supply is cheap enough and fixed in place. The cheque is the visible part. The cost curve is the part that lasts.

Concede the strongest counter in one line. Compute may commoditise, chip access rather than energy may prove the real bottleneck, and inference may not care where on earth it runs.

Take each. If compute commoditises, the margin collapses back onto the one input that does not, which is energy, and the low-cost energy producer wins the commodity the way the low-cost oil producer won the last one. Chip access is a bottleneck today and a policy variable tomorrow: export approvals loosen, foundries multiply, and the constraint that money can clear is the constraint that does not last, while the constraint that geology sets does. And inference is exactly the workload that will chase the cheapest watt, because inference is price-sensitive, runs at enormous volume, and tolerates distance in a way that latency-bound work does not. The counter describes the near term. The energy position describes the structure.

The mistake underneath the consensus

The consensus prices AI as a race for chips and models. Those are the visible, scarce, expensive things right now. They will not stay scarce. Chips get built. Models get copied and open-sourced within quarters. The scarce thing is never the flashy thing at the front. The scarce thing is the input the flashy thing consumes and cannot do without.

Value migrates as an input goes abundant. When intelligence gets cheap, its price falls toward the cost of the power it takes to produce. And whoever holds the cheapest power holds the floor of the whole market. Not the smartest position. The lowest one. In a commodity, the lowest cost is the only position that survives a price war, and a price war is exactly where abundant intelligence ends.

The Gulf spent fifty years learning one lesson better than anyone: own the cheap input the world cannot do without, and let the world come to you. It is not learning a new game. It is playing the only game it has ever played, with tokens where the barrels used to be.

The rest of the world is buying chips and calling it strategy. The Gulf is selling watts and calling it AI. One of those runs out. The other is a gas field.