AI and the energy markets: The great reshuffle

Since the COVID Pandemic in 2020, and as the global supply chain reacts to geopolitical events, flexibility and adaptation have become routine for energy commodity producers, traders, and users. Rapid advancements in AI technology and the escalation of the conflict in the Middle East are adding to the need to continuously reshuffle priorities and adapt to changing market dynamics, even faster than before.  

The dialogue on energy is becoming more nuanced, as the goal of energy security drives an enormous programme of energy addition. The debate on where data centres are located points to the complexity created by the interplay of natural resources, logistics, and insatiable demand for power to fuel AI growth.    

So, what are the new realities that energy commodity producers, traders, and users will need to adjust to? Will AI be a source of, or solution to, complexity? Will it ultimately make or break the energy markets?

Demand dynamics and data centres

The data centres behind the AI boom are hungry beasts. In the US alone, the IEA identifies 53.73 GW of installed data centre capacity today, which equates to nearly 9% of average power demand. Globally, installed capacity climbed to an estimated 114.3 GW in 2025.

For context, IRENA estimates that 114 GW of wind capacity was added globally in 2024. In other words, the numbers involved in data centre power demand are now very much material to the energy economy, capable of eating ever greater chunks out of newly installed capacity.

That’s a substantial increase in power demand. And decisions on how that demand is met – driven by political and market realities alike – will have major repercussions.

For example, energy addition requires capital investment directed toward grid development. The intention is to meet additional demand with ever higher shares of clean power generation. In some cases, that will be nuclear – the Trump administration has already pledged a $1 billion loan for the restart of Three Mile Island tied to Microsoft’s data centre plans. But in many more cases, this will entail renewable projects, particularly large-scale solar and wind assets.

The problem is that supply chains for commodities critical to the renewable roll-out are already stretched. The IEA forecasts that, under a stated policies scenario, lithium supply –  a crucial metal for batteries – will fall short of demand by 40% in 2035 and long-term supply gaps will emerge for nickel and cobalt. That’s not to even mention rare earths or the sheer volumes of steel required for mass renewable roll-outs.

We need to accept the reality of this age of energy addition, where power will be sourced from any and every resource capable of supplying accelerating demand. Signals are strong for a resurgence in natural gas to meet additional power for data centre demand, and to balance non-dispatchable renewables on the grid. So frantic is the dash for gas power that turbine manufacturers are reporting delays stretching years into the future.

And then of course – whether the electrons themselves are renewably or conventionally generated – there is the not so small matter of transmission and distribution infrastructure. Some predictions expect global grid infrastructure to more than double in size through to 2050 to connect proliferating sources of both power supply and demand. Among other things, that translates to great demand for copper, the supply of which is expected to lag demand by 30% by 2035.

No escaping geopolitics

To complicate matters further, as ever with the energy and commodity markets, we cannot ignore geopolitics. Concentration risks for key commodities are already significant, and demand crunches will only make those risks more acute. 

Currently, the majority of the world’s cobalt supply comes from the Democratic Republic of Congo (70%), followed by Indonesia (5.4%) and Russia (4.8%). Any disruption affecting the Central African country — which has faced years of violent conflict — could severely impair the supply of this critical mineral and leave buyers painfully exposed. 

And the risk of shortage is not exclusive to cobalt. More than 64% of the world’s graphite comes from China, nearly half its neodymium and dysprosium, and it is a major player in lithium and copper, among other materials. The dependence on China for the materials that underpin the energy sector is concerning. So much so that the EU passed the Critical Raw Materials Act in 2023, which aimed to both encourage domestic sourcing and discourage overreliance on a single third country. 

However, mineral deposits are not only found where it is geopolitically expedient, and new extraction and processing capacity requires large investment and long timelines. The fact is that players in the energy sector remain uncomfortably exposed to geopolitics at a time when China and the West continue tense relations, Russia remains heavily sanctioned, the US is experimenting with a new programme of trade barriers, and the Middle East is suffering a renewed outbreak of conflicts.

And let’s be clear: many of these tensions existed before AI turbocharged data centre and power demand.

Materially different markets

It’s also worth considering what the trading environment looks like for these commodities. Over the decades, the oil and gas markets have become deeply liquid, largely transparent and technologically sophisticated. With the likes of WTI and Brent serving as reference points, oil trading is overwhelmingly on-exchange. Natural gas remains more regional, but the advent of mass LNG shipping has helped globalisation.

This matters because liquidity and price transparency are the bedrock of any well-functioning market. These conditions allow counterparties to gauge risk accurately, make usable forecasts, and enter transactions with confidence. This is doubly important in commodities markets prone to boom-and-bust cycles, where traders need confidence that they can enter and exit positions at speed. 

Today, the global market is adapting to the recent escalation of the conflict in the Middle East as a supply-side shock to oil and LNG, closely watching the interplay of the mix of fossil fuels and renewables consumed in differing proportions in different regions of the world, all influenced by the natural resources acquired, stored, or accessed within a national policy agenda. 

By contrast, for other critical minerals, this is not the case. Many are not widely traded on-exchange. Instead, transactions are largely conducted over-the-counter (OTC). This limits price discovery in the wider market, makes risk management more difficult, and reduces overall transparency such that the potential for market manipulation rises.  

We are also seeing rising complexity in illiquid, long-term financial contracts used to source, store and stockpile more liquid commodities, stretching risk analysis into longer, less predictable timeframes outside the weather-influenced envelope, requiring better data, analytics, and risk management to stay appropriately hedged.  

In short: good data is more vital than ever. 

AI as a salve?

AI and the consequent rise in data centre and power demand, then, is not the root cause of many of the difficulties in navigating global energy markets: but it does add fuel to the fire. The speed and scale of the AI buildout are inflating demand, making risks more acute. For market participants who may have felt they were beginning to find their feet in an already fast-shifting energy-landscape, it will be vexing to have the deck reshuffled again so soon.

But the reason the AI proliferation is so fast and so substantial is that the technology can do amazing things. Might AI be a salve as well as an irritant?

Ultimately, short of waving a magic wand to bring all the needed new capacity online tomorrow and dissolve geopolitical tensions, success boils down to risk management.

And risk management in turn boils down to making good decisions, based on good data. This is where AI can excel if used carefully and adeptly. If market participants ensure to feed it high-quality data, AI can be used to standardise, compare, analyse, and parse that data to offer insights and recommendations to experienced traders and risk officers. These use cases are still emerging, but they are surely on their way. 

—

For now, companies can focus on fine-tuning their data procurement to maximise decisioning advantage and position themselves so that, when the time comes, AI can make their strategy rather than break it.