An AI Strategy Needs an Energy Strategy

Artificial intelligence is becoming a strategic priority for many companies. The focus is on productivity, new business models, automation and speed.

What is still missing from many presentations is the physical foundation behind it: AI needs electricity — reliably, affordably and in growing volumes.

A recent Financial Times report on New York Climate Week describes a notable shift in the debate. Discussions are no longer focused solely on emissions targets. Energy prices, security of supply and the additional electricity demand created by AI data centres are moving to the foreground.¹

In my view, this is not a side issue for utilities or data-centre operators. It is a strategic topic for every executive team considering digitalisation, electrification, growth or a new site.

Energy Is Once Again a Strategic Production Factor

For many years, companies often treated energy as an operating-cost item: procure it, consume it, pay for it.

That view is no longer sufficient.

Volatile prices, constrained grid capacity, longer connection lead times, electrification and the growing demand of digital infrastructure are changing the decision logic. Companies investing in new processes, AI applications, electric heat, charging infrastructure or additional production capacity need to consider energy supply from the outset.

The key question is no longer simply:

What does energy cost us today?

It is:

What energy, grid and flexibility structure will our business model require tomorrow?

This is where energy becomes a C-level issue. It affects capital allocation, scalability, site quality, resilience and competitiveness.

The AI Boom Is Significant — but Its Impact Is Highly Localised

The International Energy Agency estimates that data centres consumed around 415 TWh of electricity worldwide in 2024. In its base case, this figure could more than double to roughly 945 TWh by 2030.²

Globally, this would still represent a limited share of total electricity consumption. Locally, however, the effect can be substantial. Data centres concentrate high loads in a small number of locations. These are often the same grid nodes where industry, new housing, heat pumps and electric mobility also require additional capacity.

The IEA therefore warns that grid bottlenecks could delay some planned data-centre projects. At the same time, new transmission lines, substations and major grid components often require much longer lead times than the digital infrastructure itself.²

That is the real constraint: the digital world scales faster than the physical energy infrastructure behind it.

Most Mid-Sized Companies Will Not Build a Data Centre. They Are Still Affected.

This does not mean that every mid-sized company suddenly needs its own server campus. Many AI applications operate in the cloud.

The consequences still reach the company site:

Electricity prices reflect generation, grids, peaks and availability.
Grid connection capacity can become a bottleneck for expansion and electrification.
Digital production, electric heat and e-mobility increase demand at the same time.
Customers, banks and owners expect robust statements on cost, resilience and emissions.
Investments compete for capital and must make economic sense as part of the overall system.

Looking at these issues separately may produce several technically valid solutions — but not yet a robust business case.

Flexibility Is Becoming an Economic Asset

The FT report highlights a point I consider particularly important: data centres could adapt their electricity use more closely to power-system conditions. Computing workloads can be reduced or shifted during peak periods and increased when power is readily available.¹

This principle is not limited to data centres.

In industry, commercial buildings and real estate, flexible loads, storage, charging infrastructure, generation and energy management can also be controlled as one system. The economic value does not come from a single asset. It comes from the interaction:

on-site electricity is used when it is available,
loads are shifted over time,
demand peaks are limited,
grid connection capacity is used more deliberately,
mobility and heat are integrated into the site strategy,
new processes can be prepared without treating every investment in isolation.

At our own HSG Innovation HUB, we follow precisely this system logic: photovoltaics, a 270 kWh battery, heat, AC/DC charging infrastructure and energy management are assessed together. The battery does not make the grid unnecessary. It does, however, help shift energy over time, buffer charging power and significantly limit additional peaks.

For me, this is the decisive difference between buying technology and developing an entrepreneurial energy strategy.

AI Is Not Only a Consumer — It Is Also a Tool

The debate would be too simplistic if AI were portrayed solely as an additional electricity consumer.

The IEA also identifies considerable potential: better forecasting of renewable generation, faster fault detection, optimised production processes, more flexible buildings and more efficient grid utilisation. In buildings, existing AI-enabled applications could significantly improve the way heating, cooling and electricity use are managed.²

Yet the same principle applies here: technology alone does not create economic value.

Data quality, processes, responsibilities, investment and operations must fit together. Otherwise, we create more dashboards but not better decisions.

AI can manage energy intelligently. It cannot replace the entrepreneurial decision about which system we should build in the first place.

Five Questions for Executive Management and the CFO

Decision Area	C-Level Question
Demand	How will digitalisation, AI, heat, mobility and production change our power and energy requirements over the next five to ten years?
Grid	What connection capacity is actually available — and what do waiting time, expansion or insufficient capacity cost us?
Flexibility	Which loads can be shifted, stored or controlled intelligently without impairing operations?
Business Case	Which combination of CAPEX, OPEX, on-site generation, tariffs, storage, controls and potential revenues is economically viable?
Execution	Who connects the target state, data, technology, financing and specialist partners into an executable project?

These questions should not be raised only after a grid connection has been refused or an electricity bill has escalated. They belong at the beginning of the strategy process.

My Conclusion

The international debate is shifting from a climate-only discussion towards cost, security and competitiveness. I consider that a necessary development.

Not because climate protection has become less important, but because sustainable solutions will only endure if they are economically sound, robust and executable.

For companies, the conclusion is clear:

An AI strategy without an energy strategy is incomplete.

Companies planning new digital processes, electrification or growth should assess electricity, heat, process heat, mobility, generation, storage and the grid together. Not as a collection of technical work packages, but as one investment decision for the entire site.

This is where my entrepreneurial SPARRING begins. We first clarify the target state, dependencies and decision logic. REVOLUTION E can then use actual site data to develop a finance-oriented business case — before capital is committed and equipment is ordered.

From Strategic Issue to Initial Orientation

The free All-Electric QuickCheck is the fast starting point. Using industry benchmarks and model assumptions, it indicates the potential order of magnitude for operating costs, CO₂ and investment. It does not replace an individual site assessment and makes no promise of a specific result or financing outcome.

Start the All-Electric QuickCheck →

If you are currently considering AI, electrification, site expansion or new production capacity, we should not discuss software alone. We should discuss the energy system behind it.

References

This article is an independent entrepreneurial perspective by Frank Hummel. The Financial Times report is the trigger and one of the sources, but neither its text nor its structure has been reproduced.