Despite recent investor nerves around the risks of the AI buildout, power remains the key bottleneck – and investment opportunity – that will determine its speed and success, says BlackRock Australia Fundamental Equities Strategist Alex Boschen.
Key takeaways
-
01
Power is emerging as a critical bottleneck for AI’s continued growth
-
02
Long-term contracts are supporting durable earnings growth for utilities and energy infrastructure providers
-
03
Strong fundamentals and recent share price weakness may be creating an attractive opportunity, which investors can selectively take advantage of through active strategies
Across the US, abandoned industrial areas are once again becoming hotbeds of activity, as hyper scalers and neoclouds rush to build the compute facilities backing the AI boom. These facilities are often operational in a matter of months, demonstrating how quickly compute capacity can be deployed, but in many cases, access to the electricity grid comes much later.
This dynamic has led to novel solutions that help bridge this gap but also highlights a less appreciated challenge facing the AI buildout: while attention often centres on physical constraints such as semiconductors and memory, power is emerging as an equally important chokepoint.
Power underpins the entire AI ecosystem, and the growing scarcity of reliable, quickly deployable electricity is creating a multi-year investment cycle across generation, transmission and grid infrastructure. Despite this, stock prices have actually fallen since February 2026 - partly due to increased public skepticism around AI infrastructure, and partly due to an underappreciation of earnings strength in the sector.
Fundamentally though, we believe the buildout is likely to keep driving these companies to new heights.
Unprecedented earnings growth
Since the launch of ChatGPT* in 2022, demand for compute and corresponding investment has risen exponentially. This, along with an increasing population and the ubiquity of electrification, has fuelled an uptick in electricity demand around the world.
US Electricity Demand
Sources: Bloomberg, retrieved on 19 January 2025.
We see this most clearly in AI-exposed utilities, such as Entergy Corporation* in the US. Current consensus estimates forecast earnings for Entergy to grow at almost 13% this year, accelerating to 16% in 2027, and remaining over 10% to the end of the decade1. This growth rate, well above the sector’s historical growth rate of low-to-mid single digits, is being driven by the buildout of power plant and transmission infrastructure to service new hyperscale data centres.
The sustainability of these earnings is underappreciated in the market. In the hunt for power, hyperscalers are making long-term commitments that allow utilities to invest confidently in new generation, transmission and grid infrastructure. These investments support durable and accelerating earnings growth for utilities. Contract terms typically feature 10-20 year minimums and contain protections to the utility if power is underutilised.
For example, to provide energy to a massive AI compute facility in Louisiana, Meta* entered into a contract with Entergy that was signed under the company’s “Fair Share Plus” framework. The contract features a 15-year minimum term and explicit minimum bill requirements that protect the utility and its existing ratepayers from financial risks of underutilisation or early termination2.
Terms like these are becoming more common, and they are underwriting significant and durable growth in a sector where we see a number of companies able to sustain double digit earnings growth over the coming decade, making them very compelling investments.
Bottlenecks are powering new opportunities
Ongoing developments of compute facilities continue to highlight a broad mismatch: computing equipment can be deployed far more quickly than major generation, transmission, and grid connections. The gap is creating demand for behind-the-meter (BTM) power solutions that can bring computing capacity online years earlier than traditional grid connection.
BTM power solutions, which often involve the use of natural gas turbines on site, are gaining interest as hyperscalers combat pressures for faster deployment and need workarounds to meet surging demand.
What began as a temporary workaround is developing into a new contracted infrastructure market. Companies such as Solaris Energy Infrastructure* and Williams* are combining generation, natural-gas infrastructure and operating expertise to provide dedicated power at scale. For customers, the value lies in speed and certainty; for investors, it lies in long-duration, contracted cash flows. Some industry forecasts project up to 30% of future AI data center projects will include some degree of BTM power.3
Like regulated utilities, these businesses are increasingly supported by long-term contracted revenues - making them significantly more ‘infrastructure-like’, and attractive from an investment point of view, than traditional power generation.
Compelling valuations
Utilities have pulled back from their February highs as bond yields rose and investors became more cautious about the scale and timing of AI data center related investments. Yet earnings expectations remain strong, creating an attractive disconnect between near-term market sentiment and the sector’s improving longer-term growth outlook.
However, earnings projections have not fallen - in fact, the 3-5 year earnings outlook for some of these companies is the strongest it has been for decades. In a sector where the returns of companies are historically driven by fundamentals (see below), this current dynamic may provide a compelling entry point for what we see as an undervalued sector.
Source: FactSet as of 30 September 2026. Chart shows S&P 500 Utilities sector price (blue line) versus next 12 months projected earnings per share (red line).
Concerns around AI’s impact
The AI buildout raises important questions about how the costs of upgrades and expansion to existing infrastructure will be shared, and how local concerns around land use, water, cost and quality of life will be addressed. These issues have continued to attract headlines as concerns that began in the US spread globally across Europe, Australia and Asia.
While such concerns could affect the speed and economics of the rollout, at the same time, the growth of compute demand shows little sign of slowing. In the US, state-level policy debates are attracting the most attention; however, we see concerns at the local level as the more durable challenge.
The impact of this, we think, is that investment in compute capacity will likely favour expansion around existing clusters where community acceptance is already established. Therefore, we do not see the growth in AI infrastructure investment ultimately slowing down, despite the important choices it raises for policymakers.
The bottom line: AI’s growth depends on power
AI’s growth depends on power, and the infrastructure required to supply it cannot be built overnight. This is creating a multi-year investment cycle across generation, transmission, grid infrastructure and behind-the-meter power.
Regulated utilities are converting new demand into earnings growth supported by long-term customer commitments, while BTM power providers are monetising the premium hyperscalers place on speed-to-power. We believe the market continues to underestimate the duration and scale of this opportunity, making selected utilities and energy infrastructure companies compelling investments through the end of the decade and beyond.
While earnings growth is accelerating across the AI ecosystem, utilities and energy infrastructure companies stand out for the duration and visibility of that growth, supported by regulated returns and long-term contracted cash flows. Investors can tap into the power behind the AI buildout through BlackRock’s Global Listed Infrastructure Fund, an actively managed strategy that has generated alpha every year since inception3 and provided investors with best-in-class risk adjusted returns.4
