The AI ​​crisis no one is talking about

In today’s Finshots we talk about the energy crisis that goes unnoticed by the masses.

But before we start, if you are someone who likes to keep track of what is happening in the world of business and finance, then press subscribe if you haven’t already. If you’re already a subscriber or you’re reading this on the app, you can just go ahead and read the story.

The Story

When people talk about artificial intelligence, the conversation usually drifts to chips, algorithms, and trillion-parameter models. The assumption is that if you have enough GPUs (Graphics Processing Units) and enough data, the rest will somehow fall into place. But over the past year, another limitation has started to emerge, and it has nothing to do with GPUs or code.

You see, modern AI data centers are nothing like traditional server farms. Hyperscale campuses can demand hundreds of megawatts of continuous power, equivalent to a mid-sized city. Let’s see why with one simple example:

A watt-hour (Wh) is a unit of energy that measures electrical consumption, representing the energy used when one watt of power is consumed for one hour. It is calculated as Watts × hours and tells us the total amount of energy needed to perform something.

A typical Google search consumes approx 0.3Wh of electricity. However, one ChatGPT request consumes approx 2.9Wh of electricity. Considering 9 billion searches per day, this would require approx 10 TWh of additional electricity in 1 year. And all of this is just routed to a handful of servers that these companies run.

The problem is that power grids were never designed for such a claim to appear quickly in a single place. Transmission lines, substations and grid upgrades take years to plan and build. AI infrastructure, on the other hand, advances in a matter of months. Meanwhile, each major AI advance requires significantly more energy than the last.

AI’s energy demand will go from just 8 TWh in 2024 to a staggering rise 652 TWh by 2030. That’s more than an 80x increase in electricity consumption just through AI in 6 years. We can clearly see increased demand in this estimate by the IEA:

Compared to 2022, this demand for AI data centers will grow tremendously in 2026.

And that gap is now being bridged in a way we probably never expected.

One of OpenAI’s data center partners, Crusoe Energy, ordered 29 gas turbines of Boom Supersonic to power its facilities. Yes, you read that right. The company that will bring back commercial supersonic flights is also about to help drive your late-night ChatGPT requests.

Sidebar: Supersonic flights are aircraft designed to carry passengers at speeds faster than the speed of sound. They have been banned in the US and most other regions around the world due to noise pollution from sonic booms. However, the ban was lifted in June 2025 by a US presidential executive order.

And they are not the only company. Elon Musk’s xAI already uses large methane-powered turbines to keep its AI clusters running (and is also fine for that). Retired aircraft engines such as the PE6000once bolted to long-haul aircraft and military transport, are transformed into mobile power plants.

It also makes sense from the point of view of reducing waste. Rather than completely scrapping jets at the end of their flight lives, these engines find a second life in the energy sector, and that shift could unlock economic value beyond data center operations. Park them near a data center, hook them up to a fuel feed, and you have power off the grid in a matter of days.

What this trend really shows is a contradiction at the heart of the AI ​​boom. Some of the most advanced computer systems ever built now rely on aerospace hardware and the burning of fossil fuels just to stay online. These turbines are efficient for what they are, but they were never intended to support the digital economy. And yet demand has become so intense that most available industrial turbines are already sold out 2029.

But it also creates an unexpected upside. Since decommissioned aircraft no longer need to be fully scrapped, this creates a new opportunity for the Maintenance, Repair and Overhaul (MRO) sector. Countries with good MRO capabilities can refurbish, adapt and maintain these turbines for stationary use.

And India, which is actively pushing both data center investment and domestic MRO capacity through tax and infrastructure incentives sit right at this intersection. This expands the addressable market for engine specialists by adding energy power conversion to their repertoire. Firms experienced in high-precision engine work are uniquely positioned to capture this shift because no one else has matching technology and certification skills for heavy turbine work.

However, the environmental costs are also hard to ignore.

Let’s take methane turbines for example. They emit harmful nitrogen oxides into the air, which are known to cause cancer and respiratory diseases such as asthma.

These recycled jet engines also emit significant amounts of carbon dioxide. After all, they are intended as temporary solutions. But the risk is that it becomes temporarily semi-permanent. As more data centers come online and network connections slow down, downtime may turn into a structural feature of AI infrastructure. This will push local pollution higher and complicate decarbonisation goals, especially in regions that already struggle with air quality.

More importantly, it points to a deeper problem. The grid itself is not ready for sustained, concentrated loads of this magnitude. Instead of waiting for utilities to catch up, companies are taking energy generation into their own hands. That shift is blurring the line between technology companies and power producers, and it’s changing who bears responsibility for emissions and reliability.

In fact, some of the biggest players are already looking beyond jet engines. Amazon and Google have both shown interest in powering future data centers small modular nuclear reactors (SMRs). And their logic is simple. Nuclear energy provides continuous, low-carbon power without the intermittent problems of wind and solar power.

SMRs, in particular, promise faster construction, lower upfront costs and better containment compared to traditional reactors, and are not a short-term solution, but a bet on where this problem will eventually go. Seen this way, jet engines are not the solution. They are the symptom.

They tell us that if AI scales, the real bottleneck may not just be chips or data. It could be electricity. And until grids adjust to that reality, companies will keep reaching for whatever works, even if it seems outdated.

Of course, for countries like India, there is a chance to build a new revenue stream around the overhaul and operation of turbine-based power through a strengthened MRO ecosystem. However, there is a policy challenge to ensure that emergency solutions do not undermine long-term climate goals.

Fixing this permanently will require a combination of grid upgrades, faster transmission expansions and truly scalable clean power. SMRs may play a role. So there would be a better plan to align data center locations with renewable capacity.

Until then, regulators may be forced to walk a tightrope, allowing flexibility for critical digital infrastructure while placing limits on carbon-intensive backup power.

So yes, jet engines can keep AI servers running for now, but it also makes one thing clear:

The future of artificial intelligence will be shaped as much by energy policy as by algorithms.

If this story helped you better understand the AI ​​energy crisis, feel free to share it with your friends, family or even strangers WhatsApp, LinkedIn and x.

Did you know? Almost half of Indians are unaware of term insurance and its benefits. Are you among them?

If so, don’t wait until it’s too late.

Term insurance is one of the most affordable and smartest steps you can take for your family’s financial health. This ensures that they do not face a financial burden if something happens to you.

Ditto’s IRDAI certified advisors can guide you to the right plan. Book a FREE consultation and find what coverage suits your needs.

We promise: No spam, just honest advice!