Artificial intelligence is transforming economies at breathtaking speed. From generative models to autonomous machines, the world is entering what many describe as the “intelligence age.” Yet beneath the excitement surrounding algorithms and computing power lies a far more fundamental constraint: electricity. Accessing a fast and reliable supply of electricity and maximising its efficient use in data centres is increasingly becoming AI’s biggest challenge, catalysing a wave of innovation across grid infrastructure, energy storage and efficient power electronics.

Minimising the ‘time to power’

Securing a grid connection can take years, while AI developers operate on far shorter timelines. As a result, data centre operators are increasingly turning to onsite power generation – often combining natural gas turbines, battery storage and renewable inputs – to ensure fast, reliable electricity supply.

This trend has far-reaching implications. Onsite generation requires not only turbines and engines but also transformers, switchgear, power electronics and sophisticated control systems to manage fluctuating loads. Onsite energy storage plays a vital role, enabling the reduction of electricity consumption during peak demand hours (peak shaving), grid balancing and resilience during outages.

We believe the electrification challenge is not just about producing more power; it is about managing power flows intelligently, efficiently and securely.

Growing the grid

While onsite solutions are gaining traction, the broader electricity grid remains central to the energy transition. After years of underinvestment, grid infrastructure is entering a new growth phase as renewable penetration rises and demand becomes more volatile.

Accessing a fast and reliable supply of electricity and maximising its efficient use in data centres is increasingly becoming AI’s biggest challenge

Modern grids must handle bidirectional flows, integrate intermittent generation and respond dynamically to changes in consumption. This is driving demand for advanced transformers, high-voltage cables, smart meters and digital grid-enhancing technologies. These upgrades are essential not only for AI, but for the wider electrification of transport, buildings and industry.

Grid energy storage is another critical enabler. Battery systems smooth renewable intermittency, provide frequency regulation and increasingly generate revenue through capacity and ancillary services markets. With system costs falling and performance improving, storage is becoming a cornerstone of modern power systems.

Power electronics to drive efficiencies

At the heart of the transformation towards greater efficiencies lies power electronics – the semiconductors and systems that convert, control and optimise electricity from source to end use. Whether in grid management, data centres, industrial motors, electric vehicles (EVs) or solar inverters, small efficiency gains can translate into substantial energy savings at scale.

New materials such as silicon carbide and gallium nitride are enabling higher switching speeds, lower losses and more compact designs. In AI data centres, advanced power conversion can reduce energy waste between the grid connection and the processor, directly improving operating economics.

We believe efficiency is the most underappreciated lever in the energy transition. It allows us to maximise the use of each electron produced.

From data centres to physical AI

Hyperscale facilities consume vast amounts of power, often equivalent to that of mid-sized cities, and their requirements are rising sharply as compute power increases. In the US alone, data centres are expected to account for a double-digit share of total electricity consumption by the end of the decade.

However, we see this as only the beginning. Beyond digital AI lies what we refer to as ‘physical AI’: humanoid robots, autonomous vehicles, drones and intelligent machines that operate in the real world. These systems do not just process data. They move, lift, sense and act – activities that are inherently energy-intensive. As AI moves from the cloud into factories, cities and households, the challenges ensuring the systems and devices are constantly and efficiently powered is becoming paramount.

The electrification of everything – from data and mobility to industry and buildings – is a structural trend supported by technology, economics and policy

Crucially, the long-term economics of physical AI will depend on energy efficiency. A humanoid robot that can operate longer per charge, wastes less power and requires fewer battery swaps will deliver a far lower total cost of ownership. This creates powerful incentives for innovation across efficient power semiconductors, motors and actuators.

A long-term investment opportunity

For investors, these dynamics point to sustained, multi-decade opportunities rather than short-term cycles. The electrification of everything – from data and mobility to industry and buildings – is a structural trend supported by technology, economics and policy.

The Polar Capital Smart Energy Strategy focuses on companies providing the necessary ingredients for this transition: power semiconductors, grid equipment, energy storage, electrification technologies and efficiency solutions. Rather than betting on any single technology or outcome, we seek diversified exposure across the energy value chain.

Our portfolio has considerable exposure to the AI theme, strategically weighted towards technology enablers that improve data-processing efficiencies and essential electrical infrastructure suppliers. We anticipate the coming years will be decisive in determining the leaders of the AI race, expecting significant capital expenditure allocations to persist across the industry. Concurrently, we have initiated positions within the emerging humanoid robotics supply chain – a new segment projected for strong structural long-term growth – targeting companies delivering energy-efficient solutions within this burgeoning field.