At the moment, the Oil & Energy sector weight in the S&P500 is ~3.30%, while if we put together Hyperscalers & Cloud Providers, Semiconductors & AI Chips, Hardware & Networking Suppliers, and Software & Security names, all linked, in one way or another, to the “AI” trade, the weight of the whole group is over 50% (hyperscalers alone are ~20%). Which one is more critical to the economy, AI or Oil & Energy?
Pretty much everyone at this moment will answer “AI” because of all the future biblical economic growth AI companies promised to deliver, provided they managed to raise enough money to build all the infrastructure they claim they need TODAY, of course. If investors did not give them enough money, failure won’t be their fault but the market’s, which did not believe enough in a technology that is going to bring humanity into the next stage of its evolution – so they say.
More than 1 trillion USD later and almost four years since the launch of ChatGPT, all AI companies still manage to deliver are promises, lots of promises, but don’t ask them to open their books and share with the world how much money AI can really make. Today is the 11th of September, and Anthropic is still aiming to go public next month, after delaying the IPO several times already, but wants to do that without disclosing how much revenue its products are making. Not profits, of course, let’s not even dare to get there, but at least some basic figures like revenues, assets, and liabilities, not convenient ARR or all sorts of nonsense so far shared with the usual cheerleading mainstream media that is very good for PR campaigns but less good when you are asking investors to purchase shares in your business at a $ 1.75 trillion valuation. It’s pointless to spend any words on OpenAI, the frontrunner of AI for years and now a company that does not even know what it wants to be as an adult. If you zoom out and you look at the whole ecosystem supporting all AI Labs, all you see is debt, a giant pile of debt, and the same amount of cash circling around from one pocket to another, eventually landing, most of the time, in Nvidia’s pockets after being taken out of the pocket of some eager investor or lender somewhere in the globe.
We are currently in the worst energy crisis in human history, and one of the key assumptions for the whole AI ecosystem to be sustainable as a business in the long term was access to cheap energy. Have you noticed how this topic completely disappeared from the public narrative? How can all these companies become profitable with the crude oil price at $ 100, with all the repercussions in the energy ecosystem, when they were not even profitable with the crude oil price at $50? I believe this is a very fair question, but somehow there is nobody out there who dares to ask it and, to be fair, the reason is rather obvious, isn’t it?
While over a $ trillion has been spent on AI, the vast majority of this amount has all gone into US companies. The last crude oil refinery in the US, Marathon Petroleum’s facility in Garyville, Louisiana, was built in 1977. The last oil refinery built in Europe was the Mongstad refinery in Norway in 1975. Yes, the US embraced the whole shale oil “revolution” with billions of dollars poured into the sector in the early 2010s, but how did it end? Most of the companies either went bust or have been absorbed by oil majors because their revenues weren’t enough to run the business profitably and repay their debt and investments with the oil price below $90. Natural gas, solar, nuclear, and all other alternative sources of energy remained, as a whole, wishful thinking; after all, why invest in strengthening local resiliency while there were so many countries out there rich in oil and gas and eager to export as much as they could, especially those in the Arabian Peninsula, willing to foot the bill of infrastructure investments and export the products at rather low prices?
Not surprisingly, after 2008, the whole oil and energy sector was relegated to a corner of financial markets, with all the capital needed to keep the global economy resilient fleeing into the coffers of software and internet companies that, yes, repaid their investors handsomely, but did they really create real economic growth? Partially, because at the end of the day, the clients of these companies that either sell software subscriptions or advertising are, at the end of the day, those companies making widgets, distributing food and goods, producing energy, shipping anything around the globe, and so on. What is the common denominator of all these businesses operating in the real world and not on the internet? Energy, and without enough of it for everyone, the whole global economic machine starts slowing down to the point it can potentially grind to a halt. As I said before, fuel is like the blood carrying oxygen in a body, and you just need the blood pressure to drop for a body to collapse. The blood pressure in the global economy body started dropping on the 28th of February this year and is now closer and closer to that critical level where the body faints. AI is part of the body, not an entity of its own. This is the reason why it will never be able to outgrow the revenues the system generates, by definition being the AI business at the service of the real economy, of factories and workers. The conclusion is therefore banal, and that’s precisely why it’s so uncomfortable: AI will not refill the oil tanks.
You can financial-engineer an income statement, you can inflate narratives, you can recycle the same trillion dollars through the same balance sheets and call it “investment”, but you cannot negotiate with physics. You cannot prompt your way out of an energy constraint. Every additional model, every additional inference, every additional “agent” deployed at scale is not an abstract cloud service; it is a real, measurable flow of electrons that must be generated somewhere, transported somewhere, paid for by someone. And the price of that flow is no longer anchored to the fantasy of permanently cheap energy that made the last decade’s growth story possible.
This is why the current “AI = inevitable” narrative is so fragile. It assumes a world in which capital is always available, energy is always abundant, and the real economy is an afterthought. In the real world, the chain of dependency runs the other way around: chips depend on fabs, fabs depend on power, power depends on fuel and infrastructure, infrastructure depends on long-cycle investment, and long-cycle investment depends on stable economics. When that stability breaks, the first thing that gets repriced is not the dream; it’s the inputs. And once inputs are repriced, margins don’t care about your TAM PowerPoint slides.
So yes, some AI applications will survive, even thrive: the ones that actually save time, reduce waste, optimize logistics, improve yields, cut costs in energy-intensive processes. In other words: the ones that are forced to compete like any other industrial tool, against the cost of the energy they consume. Everything else, the “biblical growth” promised on the back of unlimited compute, will face the same fate as every other story built on cheap money and cheap fuel: it works beautifully until it doesn’t.
The market can keep calling this an “AI cycle” if it wants. It’s not. It’s an energy cycle, and AI is just the latest claimant on a finite resource. And when the body is short of blood, it doesn’t matter how exciting the new organ is; the organism prioritizes survival.
Dear readers and followers, after many requests over the years for more direct access to me and premium content, I launched a Premium Subscription on Slice Finance. Slice Finance not only lets me share premium content, commentary, and trade ideas for a market-professional audience, but also provides many tools you can use to directly interact with me easily: live chats, group chats, and weekly live streams. I’ve set the subscription fee at a fair, accessible price of 39$ a month with no fixed time commitment. If you’re interested, you can subscribe to my Slice portal by clicking this link: https://bit.ly/JustDario