Power Is No Longer A Given

Energy used to be the boring part of running a large organisation. You paid the bill. You managed the supplier relationship. You budgeted for modest annual increases and moved on. The interesting decisions lived elsewhere.

That assumption is now a liability.

The organisations discovering this aren't small businesses caught on the wrong tariff. They are large infrastructure operators, property groups with substantial portfolios, and utilities managing the grid itself, now facing a strategic exposure that their financial models weren't built to handle.

Here is what changed.

Artificial intelligence has become one of the largest new sources of electricity demand in history, and it is accelerating. The International Energy Agency reported that global data centre electricity consumption grew 17 percent in 2025, with AI-specific facilities surging at nearly three times that rate. The IEA's base-case projection puts total data centre electricity demand roughly doubling by 2030. Within that, the most advanced AI facilities are already growing far faster. The IEA now calls these cutting-edge facilities AI factories, and they have tripled in capacity in the past 18 months alone. Big Tech capital expenditure exceeded 400 billion US dollars in 2025 and is expected to jump a further 75 percent in 2026.

That is the global supply pressure. The more consequential shift for Australian infrastructure leaders is what it means for grid behaviour.

The electricity grid was designed around assumptions that no longer hold. Power flowed in one direction, from large, centralised generators to consumers. Demand was predictable. Supply could be planned years in advance. The system rewarded scale and stability.

AI doesn't work that way.

A typical hyperscale data centre now draws around 100 megawatts, enough to power 100,000 homes. The largest AI clusters demand two to three times that. These loads don't arrive gradually. They arrive in blocks, often with short notice, requiring the grid to absorb step-changes in demand that transmission infrastructure was never rated for. In markets where this is already playing out, the result is price and reliability behaviour that surprises even experienced energy buyers.

Australia is not immune from this pressure. Wholesale prices on the National Electricity Market have been falling, driven by record renewable generation crossing 51 percent of NEM supply in the final quarter of 2025. The average, though, conceals the real risk. On a single July day last year, South Australia recorded 68 pricing intervals above 3,000 dollars per megawatt hour, seven of them above 13,000 dollars per megawatt hour. The average looked manageable. The day did not.

Volatility, not the average cost, is the exposure that matters. And volatility is structural. Coal is retiring faster than new firm capacity is built. Renewables are creating intermittency that storage is only beginning to absorb. Global AI demand is adding structural upward pressure that will flow through to any market connected to international capital and technology investment.

For large property operators, this creates a balance sheet problem that few are actively managing. Energy costs across a commercial portfolio are typically passed through to tenants or treated as a fixed-cost line. Neither approach prices in the real exposure correctly. A portfolio running 20 or 30 large commercial buildings has meaningful price risk sitting unhedged on both sides of the lease. The question most property executives haven't been asked yet is a simple one. What does a sustained period of price spikes cost us, what does it cost our tenants, and at what point does it start affecting asset values and lease retention?

For utilities and network operators, the challenge is different and, in some ways, harder. The infrastructure investment decisions being made now, covering transmission upgrades, interconnection capacity and grid balancing tools, are built on demand forecasts that are already being rewritten. In Australia, 20 percent of renewable energy developers are already waiting two to three years for grid connection approval. That backlog is a supply constraint, and supply constraints tighten price outcomes for everyone downstream. Utilities not actively stress-testing their planning assumptions against an accelerated AI demand scenario are planning for a world that is already out of date.

The organisations getting ahead of this have started with a question most haven't asked yet. Not what energy costs on average, but what a bad month costs specifically. Across every asset, every tariff structure, every tenant arrangement. That number, once calculated, tends to change the conversation quickly.

From there, two moves matter. The first is procurement structure. Most large property operators and infrastructure businesses are either on contracts that don't reflect current market dynamics or have no active hedging position at all. The organisations managing this well are treating energy the same way they'd treat any significant input cost with genuine price volatility. Forward contracts where they make sense, real-time consumption visibility across assets, and a defined tolerance for spot exposure rather than an accidental one.

The second is planning assumptions. The standard approach is to model energy costs off average price trajectories. That works when the grid is stable. It doesn't work when the tail risk is a single day that costs more than a quarter's budget. Scenario planning that includes a sustained volatility period, not just a high-average-price year, gives leadership a realistic view of where the business is exposed and what levers exist to respond.

Neither of these requires a major capital commitment to start. They require energy to be on the agenda at the right level, with the right information in the room.

The energy bill is not a fixed cost anymore. Grid stability, predictable supply, manageable price progression. All three of those assumptions are being dismantled in real time. Large organisations that haven't revisited them are carrying a risk they probably haven't quantified.

The time to work through that is before a bad day on the grid forces the conversation.

𝛀

If you found this Independent A.I. Think Tank release useful, consider subscribing to D!sruptive newsletter. Built for business leaders who need to cut through the noise and understand what AI means for their organisation.

If you need the help discussed in this article, get in touch to see how we can work with your business 🇺🇸 +1 512 518 4196, 🇦🇪+971 585 629 975 or hello@byDisrupt.com