AI is an electricity story disguised as a software story. By 2030, the International Energy Agency expects data centres to use around 945 terawatt-hours (TWh) of electricity each year, more than double today’s level and roughly Japan’s annual electricity consumption.
That single figure captures why power has become one of the most compelling ‘picks-and-shovels’ ways to participate in the AI revolution. Models and platforms will compete. Chips and architectures will evolve. Business models will be disrupted. But whatever wins, AI still needs the same thing: reliable power, delivered continuously, at scale.
Most electricity networks were designed for a simpler era: large power stations pushing power one-way to passive consumers. Today, that world is disappearing. Demand is rising as electric vehicles, heat pumps and electrified industry shift energy use from fossil fuels onto the grid.
Supply is becoming more variable as renewables grow. Power flows are increasingly two-way, with generation and storage appearing everywhere from rooftops to industrial parks.
The world is already much further down the renewables road than many people realise. In 2025, renewables supplied roughly 32% of global electricity for the first time. The milestone is meaningful, and it reveals a hard limit. Clean power is being built faster than grids can connect and move it, leaving large volumes of renewables stuck in interconnection queues or curtailed.
In the UK, net-zero pathways imply electricity demand will be around 50% higher by 2035, and roughly double by 2050.
As networks become larger and more complex, the cost of keeping them stable rises too. These system costs are easy to underestimate, but they matter. Centrica CEO Chris O’Shea warned UK electricity prices in 2030 could exceed levels seen after Russia’s invasion of Ukraine, driven largely by those system costs.
The next phase of the transition is increasingly about the grid. Decarbonisation will only endure if it delivers reliable power without becoming too expensive for consumers.
AI accelerates electrification in three ways that matter for infrastructure. Firstly, it concentrates demand. Data centres cluster in specific places, creating local pinch points even when national totals look manageable. In the UK, Ofgem has disclosed that around 140 proposed data centre projects are seeking about 50 gigawatt (GW) of connection capacity. Not all will be built, but this demand is still enormous when compared with Britain’s current peak demand of around 45 GW.
Secondly, it compresses timelines. A data centre can be built in roughly two years, while major grid upgrades can take ten years to permit and deliver.
And thirdly, it changes the shape of demand. AI power demand can be ‘spiky’ compared with many traditional loads. A cluster of 10,000 GPUs moving from idle to training can add tens of megawatts quickly, broadly comparable to switching on tens of thousands of homes at once.
So the challenge is no longer simply to build more clean power; it is a dash to make the electricity system usable at scale.
Many investors have jumped on the most obvious beneficiaries of increased electrification: utility providers, whose shares have risen in value over the past three years. But will these companies really be able to generate exceptional long-term returns at the expense of the broader economy?
Spending will concentrate where the grid is physically constrained, making the most acute pressure points transmission build-out and flexibility.
Two companies with a genuine, defensible edge that can also benefit from demand increases that can alleviate these pressures are cable maker Prysmian and battery maker CATL.
Prysmian
Clean energy generation is expanding but grids cannot keep pace. By mid-2024, around 1,650 GW of wind, solar and hydropower in advanced development were still awaiting grid connection.
Sometimes the constraint shows up in stark local terms. In Scotland, wind farms were paid not to generate 37% of their potential output in the first half of 2025 because the cables to move the power south did not exist.
Prysmian makes high-voltage and subsea cables that connect offshore wind to the shore, reinforce overloaded transmission corridors and link electricity markets so clean power can flow to where it is needed.
AI adds urgency because data centres create dense new pockets of electricity demand that need to be connected quickly. Prysmian’s data centre order book already stands at more than €700m, evidence that this is becoming a meaningful end-market.
More broadly, the annual high-voltage cable awards market, worth roughly €3bn in 2019, is expected to reach six times that figure by the end of the decade.
CATL
Clean electricity is being wasted because power systems cannot absorb it when it arrives in a rush. In California’s main grid in 2024, 3.4 TWh of wind and solar output were curtailed – a 29% jump on the year before and enough power to support a large city for a year.
Without flexibility, renewables raise system costs and keep high-emissions peakers in place to fill the gaps. Storage is the missing piece. It absorbs surplus electricity when renewables are abundant and releases it when demand spikes. As renewable penetration rises, storage shifts from useful to essential.
CATL , a company best known for its electric vehicle (EV) batteries, has the scale and systems capability that translates directly into grid storage.
In 2024, CATL reported 475 GWh of total battery system sales and 676 GWh of manufacturing capacity – industrial throughput that can serve a fast-growing storage market on a meaningful scale.
By 2030, that market is expected to be far larger. Industry projections suggest annual energy storage shipments could approach 1,400 GWh, equivalent to roughly 23 million EV-sized battery packs repurposed for the grid.
Data centres prize uninterrupted power and face tighter grid connections, making on-site buffering and backup increasingly practical. Some projections suggest data centre storage demand alone could exceed 300 GWh by 2030, equivalent to around 5 million EV-sized packs in a single year, creating a fast-emerging demand pool that plays directly to CATL’s ability to deliver at scale.
Many of the companies building AI models and platforms are extraordinary businesses, but it is not clear-cut which will capture enduring value over the long term. It is entirely possible to be right about AI’s importance and wrong about who wins.
The electricity system stands apart. It must be upgraded regardless: for electrification, for renewables integration, and now for AI.
Cables and storage, alongside other parts of the enabling layer, can offer a more winner-agnostic way to participate in the AI build-out while also contributing to a central impact challenge: a cleaner, more resilient and affordable energy system that supports societies.
Nduka Amadi is an investment specialist on Baillie Gifford Positive Change. The views expressed above should not be taken as investment advice.