EWIT enlightenment: IREN

As with any investment, capital is at risk.

As artificial intelligence becomes part of our daily lives, so too do the pipes and power that enable it. Independent estimates now forecast multi‑trillion‑dollar investment in AI infrastructure this decade, led by hyperscale datacentres, high‑voltage grid connections and advanced cooling. Mega-projects such as Prometheus (operated by Prometheus Hyperscale), Stargate (operated by OpenAI, Oracle, and Microsoft) and Colossus (operated by Elon Musk’s xAI) are pushing the cumulative tab to roughly $3 trillion by the late 2020s.

For context, at the height of the internet build-out in the early 2000s, US telecom carriers peaked at roughly $120bn of annual capital outlay. Today’s AI Capex (capital expenditure, money spent by a business to acquire, upgrade, or maintain long-term assets to generate future benefits or returns) by the hyper-scalers (such as Amazon’s AWS, Google Cloud Platform and Microsoft Azure) alone is running at the low hundreds of billions per year, with it forecast to go higher. Back in the early 2000s, the internet’s physical infrastructure was limited by the challenges of laying fibre. Today, it looks likely that the bottleneck will be the availability of electricity.

AI has a vast appetite for power, which is already straining the grid. The International Energy Agency (IEA) projects that global electricity demand from datacentres will double by 2030, roughly to Japan’s total current consumption.

As AI demand surges, resulting shortages are local as much as global. Accordingly, the largest cloud providers are expanding multi-year capital plans and pre-booking grid connections, power contracts and critical equipment. This soaks up scarce near-term capacity and leaves power markets tighter for everyone else. Substation lead times in key US markets now stretch years, and, as a result, developers are prioritising sites with existing high‑voltage infrastructure and room for liquid‑cooled halls. That’s why IREN’s renewable datacentre proposition is so powerful. It addresses the worst bottleneck: power.

From bitcoin to AI infrastructure

Founded in 2018, IREN operates vertically integrated datacentres, designed in-house and run on low‑cost predominantly renewable power. The company started with bitcoin mining, but its strategic direction has always pointed towards the energy‑intensive computing with the greatest earnings potential – which is now AI. As such, IREN paused further mining expansion in March, reallocating capital into liquid‑cooled AI halls.

Today's operating model spans three pillars: its cash-generative bitcoin mining business, its AI Cloud offering that lets clients rent GPU compute, and a pipeline of high‑density datacentre campuses designed for next‑generation systems. Importantly, IREN has secured approximately 2.9 gigawatts (GW) of grid‑connected power across North America, providing a multi‑year runway to convert scarce megawatts (MW) into rentable compute. Today it operates in the following sites:

• Sweetwater, Texas (2.0GW campus): Flagship, multi‑phase site designed for liquid‑cooled AI training and inference
• Childress, Texas (~750MW site): Home to Horizon‑1, a 50MW liquid‑cooled facility targeting delivery in late 2025
• British Columbia, Canada (~160MW across sites): Live operations combining bitcoin and AI Cloud on shared infrastructure, with the option to expand

Why IREN’s edge is hard to copy

First, IREN starts with power, the scarcest input. Thanks to the background of its founders (co-founder and CEO Daniel Roberts previously worked in infrastructure asset management, and invested in wind farms and pipelines), IREN has demonstrated a knack for identifying undervalued real estate with strong connections to the grid that are already present or can be easily established. 

From there, IREN has already proved it can build and run high-density datacentres quickly and cheaply. Its bitcoin mining arm is clear evidence of this, where tight engineering and low-cost power translate into substantial profits, outperforming major peers like CleanSpark and Marathon. IREN uses the same playbook for its liquid-cooled AI halls, where the company is applying its proven cost discipline to denser, more valuable compute.

The company also signals flexibility in commercial models allowing it to match contract structure to customer appetite as conditions evolve.

Potential upside

Financially, IREN has reported back‑to‑back profitable quarters, with profit margins near 60 per cent, while maintaining a conservative balance sheet. This is impressive, given that the company is only in the early stages of commercialising its AI Cloud. 

From here, the plan is to add customers at its liquid-cooled Horizon‑1 facility at Childress, while advancing conversations with larger tenants at Sweetwater. Management’s own scenarios illustrate how monetising the current c.2.9GW of secured power could translate into multi‑billion‑dollar annual revenue late in the decade. We can foresee a path to this site generating around $9–10 billion in annual revenue by 2030 at full utilisation. We’ll monitor the swing factors – securing big tenants at scale, maintaining timely access to advanced GPUs, building liquid‑cooling capacity at pace – but we’re optimistic.

Conclusion

At its core, the AI era is a power and plumbing story. Significant infrastructure is required for the technology to deliver on its tantalising potential. Thus, the impending spending wave is real and dwarfs that of previous cycles. 

Pleasingly, the portfolio has increasing exposure to this, with several holdings participating in or helping to drive it across the value chain, from energy infrastructure to cutting-edge chip design.

However, power and grid capacity are already a potential bottleneck to growth, placing a premium on developers who control real power, permitted land and high‑density designs.

We’re excited by how IREN addresses this. The business has the opportunity to turn scarce power into durable compute economics. If the company converts its sites into contracted AI datacentre capacity, and scales delivery in the coming years, we see considerable fundamental growth opportunities. 

Important information

This article does not constitute, and is not subject to the protections afforded to, independent research. Baillie Gifford and its staff may have dealt in the investments concerned. The views expressed are not statements of fact and should not be considered as advice or a recommendation to buy, sell or hold a particular investment.

Baillie Gifford & Co and Baillie Gifford & Co Limited are authorised and regulated by the Financial Conduct Authority (FCA).

Some of the views expressed are not necessarily those of Baillie Gifford. Investment markets and conditions can change rapidly, therefore the views expressed should not be taken as statements of fact nor should reliance be placed on them when making investment decisions.

Baillie Gifford & Co and Baillie Gifford & Co Limited are authorised and regulated by the Financial Conduct Authority (FCA). The investment trusts managed by Baillie Gifford & Co Limited are listed on the London Stock Exchange and are not authorised or regulated by the FCA

A Key Information Document is available at bailliegifford.com.