WHY ARE OILMEN BUYING THE AI FUTURE?

 

The old energy economy buys the new one.

For years, the story seemed almost settled. The future belonged to software. Then cloud computing. Then artificial intelligence. The companies that mattered would be the ones building models, chips, platforms and applications. Oil belonged to another age—a vast industrial inheritance from the twentieth century, increasingly challenged by electrification, renewables and the transition away from fossil fuels.

Then the machines got hungry. Not metaphorically. Literally.

Artificial intelligence needs enormous quantities of electricity, and as the machines become more powerful, the AI revolution starts looking less like a software story and more like the biggest industrial construction project in decades. Data centres need land. They need substations, transformers, cables, cooling systems, water, construction capacity and, increasingly, dedicated generation. The digital economy has discovered that it still has to live inside the physical one. And the old energy economy has noticed.

On August 31, SLB, the world's largest oilfield-services company, announced that it had agreed to acquire Kelvion, a global provider of thermal-management and heat-exchange technologies, for approximately $3.4 billion in cash plus about $700 million of assumed debt. Kelvion's largest and fastest-growing end market in 2026 is expected to be data centres, with data-centre revenue of roughly $1.2 billion to $1.3 billion. SLB says the combined data-centre business could generate $4.5 billion to $5 billion in revenue by 2028.

An oilfield-services company is spending billions to deepen its position in the machinery needed to build and cool data centres. That is not simply diversification. It is a signal. The company that spent generations helping the world extract energy from the earth is positioning itself to help the machines that consume extraordinary amounts of energy operate.

The future did not replace the past. It signed a contract with it.

The conventional image of artificial intelligence is wonderfully clean. A model exists somewhere in a cloud. A user types a question. An answer appears. The interface conceals almost everything that makes the answer possible: buildings, substations, cables, generators, turbines, cooling loops, transformers, pipelines and industrial equipment. AI looks weightless because somebody else is carrying the weight. And that weight is becoming enormous.

The International Energy Agency estimates that global data-centre electricity consumption rose 17% in 2025, while electricity consumption from AI-focused data centres increased by 50%. Its latest base case sees total data-centre electricity consumption roughly doubling from 485 TWh in 2025 to about 950 TWh in 2030. But the really important number is not simply how much electricity AI needs. It is how quickly it needs it.

A city can spend years expanding its grid. A transmission line can take years to permit and construct. Transformers and other critical equipment can have long lead times. AI companies, meanwhile, are competing in a technological race in which being late can mean losing customers, developers and strategic position. That creates a peculiar economic problem: the world's most advanced digital companies suddenly need the world's oldest industrial capabilities. And this is where the oilmen enter the story.

Consider Chevron. In June, Chevron announced a 20-year power agreement with Microsoft to develop a 2.67-gigawatt co-located natural-gas power facility in West Texas serving a Microsoft-operated data centre. The project, known as Project Kilby, is expected to deliver first power in 2028, subject to the necessary conditions and Chevron's final investment decision. The project is designed to provide dedicated, dispatchable electricity directly to Microsoft while reducing the project's dependence on the regional grid.

Look at the two names. Microsoft is one of the defining companies of the artificial-intelligence economy. Chevron is one of the defining companies of the petroleum economy. And they are building electricity infrastructure together. This is not a sentimental reunion between the old economy and the new. It is a transaction. Microsoft needs power. Chevron knows how to build energy projects. The market is discovering that those two facts may be worth billions.

The distinction matters because the AI energy problem is not simply about producing more electricity somewhere. It is about producing enough electricity in the right place, at the right time, with sufficient reliability and on a timetable that matches the construction of the data centre itself. That is a completely different problem. A gigawatt is not a line on an investor presentation. It represents generation equipment, fuel supply, grid connections or dedicated infrastructure, transformers, cooling, land, permitting, engineering, construction and years of capital expenditure.

And time is becoming an asset.

The IEA says electricity grids are already under strain and estimates that around 20% of planned data-centre projects could face delays if grid risks are not addressed. It says building new transmission lines can take four to eight years in advanced economies, while waiting times for critical grid components such as transformers and cables have roughly doubled over the past three years.

This changes the economics of energy companies. For decades, the oil and gas industry was largely understood through the commodity it sold: oil and gas. The price went up. The price went down. But an energy company operating in the AI buildout can potentially sell something else: certainty.

A data-centre developer does not merely want electricity. It wants to know that the electricity will be there when the servers arrive. That distinction creates an extraordinary opportunity for companies with experience in large-scale physical projects. Oil and gas companies understand how to secure and develop industrial sites. They understand project finance, heavy equipment, complex engineering, fuel logistics, construction schedules, regulatory processes and long-lived infrastructure. They understand what happens when a project fails because one physical component is late. Technology companies are learning that lesson now.

The AI industry can design a new model in months. It cannot manufacture a high-voltage transformer in a weekend. It can announce a data-centre campus. It cannot make a transmission queue disappear. It can buy thousands of GPUs. It cannot persuade the laws of physics to deliver another gigawatt simply because the next training run is urgent.

Power is not a button.

That is why the energy story surrounding AI is becoming much more interesting than the familiar argument over whether oil and gas will eventually disappear. The more immediate question is whether the companies that understand energy infrastructure can become indispensable to the companies building intelligence infrastructure.

The evidence is already appearing. The IEA reports that orders for new natural-gas-fired power plants reached 130 GW in 2025, a 25-year high, with US data-centre demand a major driver. It also says strong demand in the United States and Middle East is limiting turbine availability for near-term deployment elsewhere.

But this does not mean the AI economy is simply becoming a gas economy. That would be too crude—and wrong. The IEA expects renewables to provide about half of the global growth in electricity demand from data centres, while natural gas, nuclear and other sources also expand. In its base case, natural gas supplies an additional 175 TWh of generation for data-centre demand through 2035.

The deeper story is not which fuel wins. It is who can reliably convert energy into usable power at the speed AI requires.

That is a much bigger market, and it explains why companies from the traditional energy ecosystem are moving toward the data-centre economy. PwC estimates that US AI-linked natural-gas demand could reach between 7.58 and 11.47 billion cubic feet per day by 2035, depending on its scenario assumptions. More importantly, its analysis argues that existing pipelines, storage assets and rights-of-way can have strategic value because they can reduce permitting and development delays.

That phrase—reduce development delays—may ultimately matter more than the gas itself, because the AI economy is beginning to create a new hierarchy of value. Cheap land is useful. Land with power is better. Land with power and transmission access is better still. Land with power, transmission access, cooling, fibre, fuel infrastructure, permits and a credible customer is something else entirely. It is no longer simply land.

It is time-to-compute.

And time-to-compute may become one of the most valuable forms of infrastructure in the AI economy. This is where the old energy economy becomes unexpectedly relevant. Oil and gas companies spent decades solving the problem of moving enormous quantities of physical energy through hostile environments. They built pipelines, processing facilities, power systems, industrial campuses and engineering ecosystems. Now AI is creating an enormous new customer for those capabilities.

The old energy economy does not have to build the smartest AI. It does not have to train the largest model. It does not have to beat Nvidia, Microsoft, Google or OpenAI at software.

It only has to become indispensable to the machines those companies are building.

That is a very different competitive strategy. SLB's Kelvion acquisition makes the point particularly clearly. SLB says its existing Data Center Solutions business is already growing rapidly, with revenue expected to have increased at a compound annual rate exceeding 90% between 2024 and 2026, while cumulative delivered capacity is expected to surpass 2 GW by the end of 2026. Its strategy combines modular manufacturing, off-site construction, engineering and digital capabilities with Kelvion's thermal-management technology.

The significance is easy to miss. SLB is not trying to become another software company. It is trying to become an industrial technology company for the AI infrastructure era. That may be the more natural transition for an oilfield-services company anyway. The industry has spent decades learning how to make complex machines work under extreme conditions. Data centres are simply creating a new extreme condition: heat, density, power, speed and reliability. The machines may be different. The industrial problem is strangely familiar.

And there is another reason this matters. Once energy companies begin selling directly into AI infrastructure, the boundaries between utilities, oil and gas producers, data-centre developers, technology companies and infrastructure investors begin to blur. Microsoft becomes an electricity customer. Chevron becomes a power developer. SLB becomes a data-centre infrastructure supplier. Gas producers become potential suppliers to dedicated AI generation. Cooling companies become strategic infrastructure assets. And the data centre itself begins to look less like an office building filled with computers and more like a new kind of industrial plant.

That could change how the energy sector is valued. A commodity sold into a volatile market is one thing. A long-term fuel arrangement attached to a major AI customer is another. A pipeline serving contracted industrial demand is not economically identical to a pipeline exposed entirely to spot-market fluctuations. A power plant built around a long-term technology customer is not merely another generation asset. Infrastructure begins to acquire something the commodity business has historically struggled to provide: visibility.

That may be the real attraction. The oil and gas industry does not need AI to rescue every barrel. It needs AI to create new, durable demand for the infrastructure and expertise it already possesses.

And this is where the story becomes uncomfortable.

Artificial intelligence was supposed to make physical geography less important. Software could move instantly. Information could travel everywhere. The cloud was supposed to make location almost irrelevant. Instead, AI is making location important again. A data centre wants electricity, land, cooling, fibre, water or an alternative cooling architecture, transmission access, generation and regulatory permission. And it wants all of these things close enough to one another to make the economics work.

The digital revolution is therefore creating a strange reversal. The more intelligent the machines become, the more valuable certain pieces of physical geography may become. That is why the old energy economy is not necessarily standing outside the AI revolution.

It may be standing underneath it.

And perhaps the most important change is not that oil companies are becoming technology companies. It is that technology companies are becoming industrial customers. That changes the relationship. For years, Silicon Valley could behave as though the physical world were an invisible service layer beneath software. Now the physical layer is becoming strategically scarce: electricity, transformers, turbines, cooling, transmission, pipelines, land, water, construction capacity and time.

The AI economy is bidding for all of them simultaneously. And the companies that already know how to acquire, develop, connect and operate them suddenly possess something Silicon Valley cannot manufacture with code: industrial memory.

That is why the SLB deal matters. That is why Chevron and Microsoft matter. And that is why the resurgence of gas-fired generation matters—not because fossil fuels have somehow defeated the energy transition, but because AI has made reliable energy delivery strategically valuable again.

The future may still be renewable. It may eventually become increasingly nuclear. It may involve storage technologies we cannot yet predict. But whatever the final energy mix becomes, someone will have to build the physical system that delivers those electrons to the machines. And that is the part the old energy economy understands.

The oilmen may therefore be making a bet that looks strange only if we think AI is primarily a software revolution. If AI is instead becoming a gigantic industrial system, their bet starts to look rational. They are not necessarily buying artificial intelligence. They are buying pieces of the machinery that intelligence cannot live without.

That is a much bigger opportunity.

Because the real competition may not ultimately be between the company with the best model and the company with the second-best model. It may be between the companies that can scale intelligence and the companies that can supply the physical conditions required to scale it. The first group writes the algorithms. The second group keeps the lights on.

And when the world's most valuable technology becomes dependent on a scarce physical input, the supplier of that input acquires something every technology company understands:

Leverage.

The old energy economy spent the last century learning how to control energy flows. The new intelligence economy is creating an enormous new demand for them. Perhaps that is why the oilmen are coming. Not because they believe the future belongs to oil. Because they have recognized something more subtle.

The future still needs power.

And whoever learns to control the infrastructure between power and intelligence may discover that the most important business created by AI is not artificial intelligence at all. It is the industrial system that keeps artificial intelligence alive.

The AI revolution may not be replacing the old energy economy.

It may be hiring it.

---

You May Also Like

The Next Billionaires Aren't Buying Things—They're Buying the Next Scarce Resource: Permission

The Phantom Asset: How the Future Is Already Worth Billions

What Do Billionaires Know About Land That the Rest of Us Don't?

By: Manish Kumar, Founder & Editor, Explain It Clearly