Compute is the new oil: ‘The Entropy Trap’ author Mickey M Maini on the AI buildout
Maini explains why energy, not chips, will decide who wins the AI race
14 September, 2026
TT
16
Most market commentary on the artificial intelligence boom trades in forecasts. Mickey M Maini prefers to take measurements. The founder of Solstice Laboratory, a Dubai-based research lab that applies physics to financial and geopolitical systems, and author of The Entropy Trap: What Physics Knows That Markets Don’t, Maini reads markets the way physicists read systems under stress, separating how loaded a system is from how fast it is moving. On that framework, he argues, the debt-funded AI infrastructure buildout is showing real structural stress but has not yet tipped.
In this interview with Gulf Business, he explains why the compute will outlast some of the balance sheets financing it, why machine-speed markets have quietly removed the time economies assume they have to self-correct, and why energy, not chips, will decide who wins the AI race. For the Gulf, his message is pointed: compute is the next scarce strategic resource, and a region that spent 50 years mastering the economics of oil has answered this question before, provided it treats its position as infrastructure to maintain, not luck to enjoy.
Does the current AI infrastructure buildout, largely funded through debt, show the kind of structural stress signals your framework is designed to pick up, similar to past pre-transition periods?
Yes, but not the signal most people expect. Stress is not the same as transition. Our gauges separate the two: how loaded a system is, and how fast it is moving. Right now the AI buildout reads loaded, not yet moving.
Here is what changed this year. The buildout used to be paid for out of cash flow. Now it is paid for with borrowed money. Capital spending by the big platforms is heading toward seven hundred billion dollars this year, and by one Wall Street count it now consumes close to all of their operating cash flow, against roughly forty per cent as the average of the past decade. AI-linked debt issuance has more than doubled in a year. In physics terms, the system lost its natural brake. Cash flow is self-limiting. Debt is reflexive.
We have tested this pattern against history. British railways in the 1840s. American fibre in the 1990s. The infrastructure was real, the financing was not, and the two outcomes separated. The paper burned. The rails and the fibre carried the next fifty years of growth.
So the honest reading: the compute will survive. Some of the balance sheets will not. What we watch is one simple line: the moment new borrowing starts servicing old borrowing instead of building new capacity. It has not been crossed. It is close enough to measure.
Does AI-driven, algorithmic trading moving at machine speed break the basic assumption most economic models rely on, that markets have time to self-correct?
No, and that is the uncomfortable part. Self-correction is not a law of markets. It is an assumption, and it rests on two quiet conditions: time and diversity. Markets correct when participants have time to disagree, and when they actually disagree. Machine speed removes the first. Crowding into the same trades removes the second.
Physics has a name for this. A system stays stable when it can relax faster than it is shocked. When the shock arrives faster than the response, the system does not bend. It snaps to a new state. We have already watched corrections that once took a year compress into a quarter, then a week, then an afternoon. The machines did not change the physics. They removed the time everyone assumed they had.
Our answer is not to be faster than the machines. Nobody outside a server rack wins that race. Our answer is to be earlier. We measure the conditions under which speed becomes dangerous: how crowded positioning is, how closely assets move together, how thin the buffers are. Right now those gauges show correlations unusually high and market calm unusually deep, while the physical world runs hot. That gap is exactly where fast markets get hurt.
The practical rule is old, and it still holds. You cannot react your way out of a fast market. You can only be positioned before it. Speed rewards the prepared and punishes the reflexive.
Does AI compute becoming a sovereign-level resource rather than a commercial one change how a region like the Gulf, where sovereign capital and state strategy already play an outsized role, should be thinking about its position?
It changes everything, and the Gulf should recognise the moment, because it has lived it before. For fifty years this region has run the most successful strategic resource economy in the world. The playbook is written. The owner of a scarce input sets terms. Refine at home rather than export raw. Convert the windfall into permanent institutions. The sovereign funds of this region are that playbook made durable.
Compute is the next scarce input. Gold anchored trade. Oil powered industry. Intelligence runs on compute, and compute runs on energy, land, capital and trusted jurisdiction. Those are the four abundances of the Gulf. That is why the region holds a genuine seat at this table rather than a spectator’s ticket.
The shift to make is mental. Compute stops being procurement and becomes statecraft. A large compute agreement today deserves the questions an oil concession deserved in the 1930s. Who owns the asset? Who operates it? Who sets the price of its output? Where the knowledge ends up. Regions that asked those questions built national champions. Regions that did not spent decades buying back what sat under their feet.
And one measurement, because measurement is our trade. The region should track its share of global compute capacity with the same seriousness it once tracked its share of oil capacity. What you measure, you manage. What you manage, you keep.
Does the energy demand from AI data centres, rather than chip supply, represent the bigger constraint on who actually wins the AI race, in your view?
Energy, without much doubt. Chips are a factory problem, and factories scale. You can build a chip plant in three years. Electrons are a physics and permissions problem. Power plants, transmission lines and grid connections move on decade clocks, and the queue for a grid connection in the West now runs to years. Money is not the constraint either. Capital is chasing this buildout at historic scale. The constraint is the socket.
The numbers say it plainly. Data centre capacity worldwide is set to roughly double by 2030, while the utilities that must feed it are investing at a fraction of the required pace. When one input is abundant, and another is fixed, value migrates to the fixed one. That is not opinion. That is how every shortage in history has been priced.
So the AI race will not be won by whoever designs the cleverest model. It will be won by whoever delivers dispatchable power at scale, fastest, with the fewest permits in the way. Read the map through that lens, and it redraws itself toward energy-rich jurisdictions that can build. The UAE understood this early. Nuclear at the Barakah plant, gas, solar at scale, and land next to all three.
The industrial age asked who has the oil. The intelligence age asks who has the electrons. It is the same question wearing new clothes, and this region has answered it before.
Does a neutral hub like the UAE gain a lasting edge as AI becomes a geopolitical fault line alongside debt and fragmentation, or is that advantage more temporary than people assume?
It can be lasting. It is not automatic. Neutrality is a position you maintain, not a prize you keep. In a fragmenting world, flows do not stop. They reroute. The node that stays open, connected and trusted collects the toll on that rerouting. That is the physics of hubs, and it has held across a thousand years of trade history.
But the position has a maintenance schedule. Four things sustain it. Energy, which this country has. Connectivity, which it keeps building. When the region’s main artery came under pressure this year, the answer was new pipeline capacity toward Fujairah and new corridors, which is exactly what an adaptive hub looks like. Rule of law and capital mobility are why families and firms keep arriving. And now compute, the fourth utility.
History grades both outcomes. Venice held the middleman position for three centuries because it kept investing in the things that made it central. Amsterdam lost the crown to London in a generation once it stopped. The edge is temporary for hubs that treat it as luck, and lasting for hubs that treat it as infrastructure. On the evidence of the past few years, the UAE is behaving like the second kind.
The gauge we watch is stickiness. Of the capital and talent that arrived in the stress years, how much stays once calm returns? Passing traffic is revenue. Traffic that settles is destiny.
You call yourself a measurer, not a forecaster. You say the AI buildout is loaded but not yet moving. What are the gauges that will tell us it is tipping, and what should investors here do when they move?
Three dials, and all of them can be read from public information.
The first is funding quality. Watch whether new borrowing in the AI buildout is creating capacity or servicing old borrowing. Every infrastructure boom in history crossed that line quietly before it broke loudly. It has not been crossed yet. When it is, the weakest financing goes first while the strongest assets keep running.
The second is power delivered against power promised. Announcements are free. Megawatts are not. When the gap between connected capacity and announced capacity widens, timelines slip. Slipped timelines are what turn debt from fuel into weight.
The third is the price of trust. Watch what gold does against the money supply while all of this is being financed. It is the oldest gauge of confidence in paper promises, and it currently sits at levels that reward attention.
What to do about them is architecture, not heroics. Hold a foundation that does not depend on the financing weather: real assets, and the physical layer of the buildout itself- power, grid, the inputs intelligence cannot scale without. Keep paper claims short. And keep liquidity, because the discipline that feels like dead weight in calm times becomes ammunition on the day the dials move.
When the ground shakes, the sorting begins, and the prepared side of the room does the buying. Our work is making sure you can see the dials before that day.
