The AI boom is becoming an energy boom

This article is part of In conversation about sustainable finance & emission reduction systems, a new series by Diego Balverde. You're reading volume 18 of the Breaking news series. Here is volume 17

Prologue

Artificial intelligence is entering a second economic phase, and it may ultimately matter more to the real economy than the first. Over the past several years markets concentrated on models, semiconductors, GPUs, hyperscalers and companies capable of constructing computing capacity, but that first race is now revealing a much larger second one.

Every new machine requires electricity, every data centre requires connectivity, every connection requires transformers and substations, every grid requires copper, every facility requires cooling and water, every interruption requires storage or backup generation, and every additional GW ultimately requires physical infrastructure that cannot be downloaded, trained or created by an algorithm.

Evidence of that shift is becoming increasingly visible this week as artificial-intelligence-related electricity demand reaches manufacturers of power infrastructure while copper remains close to record levels and markets begin to understand that the bottleneck of the digital revolution may increasingly exist outside the technology sector.

The next phase of artificial intelligence will therefore be an energy, infrastructure and capital story, and one of its most important consequences may be to transform industries previously regarded as mature into some of the largest growth markets of the coming decade.

Electricity becomes the new location factor

During the traditional era of industrial globalisation, companies primarily searched for labour, taxes, transportation, proximity to customers and regulatory stability. The new economy adds another variable that is beginning to alter all the others: available electrical power.

Not theoretical electricity, not annual generation announced in a government plan and not capacity expected a decade from now, but MW capable of being connected within the economic timetable of the project. A data centre can purchase servers quickly, a corporation can obtain capital and a government can provide land, but none of these variables can substitute for a saturated substation, a transformer requiring years for delivery or a transmission project that has not yet received approval. The digital economy is therefore discovering one of the fundamental laws of the physical economy: the world's most sophisticated asset remains dependent upon basic infrastructure.

This changes investment geography. Regions combining abundant generation, available grids, natural gas, nuclear, renewables, storage, water, fibre and industrial land can acquire an extraordinary advantage over traditionally attractive but electrically congested cities. Energy availability may increasingly influence the value of industrial land, technology parks, ports and logistics corridors in the same way that highways, airports and rail connections shaped previous economic eras. The scarce asset may cease to be simply land and become land with available MW.

The grid is no longer invisible infrastructure

For decades, electricity networks were infrastructure that consumers barely noticed. Generators produced power, utilities transported it and businesses and households assumed that connectivity would exist. Electrification is destroying that comfortable assumption.

Electric vehicles, heat pumps, industry, BESS, distributed renewables and now data centres are simultaneously increasing generation and demand while networks must manage much more complex flows. Building generation without sufficient transmission can produce curtailment; constructing demand without connectivity can strand investment; building both without flexibility can increase total system costs.

The opportunity therefore extends far beyond constructing thousands of additional kilometres of transmission. An entire productivity layer can be developed on existing networks through dynamic line rating, reconductoring, power flow control, digital substations, strategically located storage, demand response, flexible connections, distributed generation and artificial-intelligence systems capable of anticipating congestion and reorganising loads.

In certain circumstances, the best investment will not be immediately constructing another enormous piece of infrastructure but releasing capacity from assets already operating. This creates a fundamental conceptual difference. A newly constructed MW and a released MW can provide the same economic outcome for a customer requiring connection, while their costs, construction times and risk profiles can be radically different. Financial markets do not yet adequately value that distinction.

The transformer becomes a strategic asset

The artificial intelligence revolution is creating another paradox. Some of the most technologically advanced corporations on Earth can become constrained by electrical equipment based on principles more than a century old. Transformers, switchgear, cables, turbines, motors, cooling systems and high-voltage equipment are returning to the centre of the economy precisely because digital growth requires extraordinary physical expansion.

This helps explain why companies connected to electrical infrastructure are beginning to benefit from the data-centre investment cycle. The signal should not be interpreted simply as a short-term equity-market opportunity. It reveals something structural: technology CAPEX is escaping the boundaries of the technology sector. Every dollar allocated to GPUs creates indirect demand for buildings, electricity, generation, networks, cooling and materials. The true economic footprint of artificial intelligence will consequently be much larger than the revenues generated by companies selling models or semiconductors.

The effect can extend into unexpected industries. Oil & Gas possesses expertise in turbines, generation, large projects, engineering, pipelines and drilling. Nuclear can provide firm electricity to high-utilisation loads. Renewables can deliver enormous quantities of competitive energy when appropriately combined with grids and flexibility.

BESS can respond to peaks, provide grid services and improve connection utilisation. Advanced geothermal can deploy techniques developed during decades of petroleum engineering. Ports can become privileged locations for new energy and digital industries because they concentrate transmission, industrial land, logistics, fuels and access to heavy equipment. Artificial intelligence is not replacing the old economy. It is increasing the value of selected parts of it.

Gas and nuclear return for a different reason

Electricity-demand growth is also changing the generation debate. Data centres operate around the clock and their economic value depends upon extremely high availability. This creates demand for firm electricity that can be provided through different combinations depending on each market.

Gas-to-power offers speed and dispatchable capacity in regions possessing gas infrastructure. Nuclear provides high-utilisation firm generation and can acquire new economics through long-term contracts with large consumers. Solar and wind can deliver enormous volumes of competitive electricity, particularly when combined with storage, transmission and flexible load management. Hydropower and pumped storage can provide flexibility. Geothermal can offer continuous generation.

The conclusion is not that one technology has defeated another. It is precisely the opposite: electricity demand growth may be sufficiently large to increase the value of multiple technologies simultaneously.

This represents a fundamental departure from much of the previous decade's energy debate. If global electricity demand expands rapidly through electrification, industrialisation and computing capacity, the problem is no longer how to divide relatively static demand among competing sources. It becomes how to construct sufficient generation, transmission, storage and flexibility without destroying competitiveness through excessive system costs. Energy therefore becomes industrial policy again.

Copper explains why software cannot solve everything

Electricity infrastructure also faces a material constraint. Copper remains near historically elevated levels while artificial intelligence, grids, electric vehicles and manufacturing compete for the same metal. The International Energy Agency's Global Critical Minerals Outlook 2026 warns that global critical-mineral investment fell by approximately 9% during 2025 even as supply concerns increased, geographic concentration in processing continued to grow and new trade restrictions transformed theoretical vulnerabilities into immediate economic risks.

The most important issue is not simply copper's price. It is the relationship between relatively small quantities of selected minerals and enormous amounts of economic production dependent upon them. The IEA estimates that full implementation of certain rare-earth restrictions could put approximately $6.5 trillion per year of downstream production outside China at risk across automotive, high-tech, defence and energy industries. This demonstrates one of the defining characteristics of modern systems: a comparatively small component can control an enormous economic flow.

Mining therefore can no longer be analysed separately from artificial intelligence, energy or defence. Copper connects all three. Rare earths connect motors, defence and electronics. Graphite connects batteries and mobility. Uranium connects energy security and nuclear generation. The geopolitics of the next decade will increasingly revolve around who extracts these materials, who processes them, who manufactures components and who controls the technologies required to do so.

Latin America: The opportunity is not to extract more, but to capture more value

Latin America possesses an exceptional opportunity precisely because it participates significantly in multiple mineral supply chains, but repeating the historical model of extraction and export would once again transfer most industrial value to other regions. The real opportunity is to use mineral resources to construct productive ecosystems around them.

A copper mine should conceptually connect to electricity, water, processing, ports, component manufacturing, international contracts and finance; lithium should connect to active materials, cells, BESS, management software and recycling; energy resources should be used to attract industries requiring competitive power.

This creates a concrete role for BalGreen. Not to compete with miners in extraction or manufacturers in equipment production, but to work on the efficiency of the system connecting those assets. A mine can lose money through energy, water, maintenance, transportation, equipment downtime and port logistics. A processing facility may be constrained by electricity. A port may create congestion that reduces exportable tonnes. BESS can reduce peaks and increase stability. Local generation can reduce exposure. DOIX can measure every variable before and after intervention.

The objective is to transform efficiency into productive capacity. And then transform productive capacity into financial yield.

The opportunities

The first major market is grids. Transformers, cables, substations, reconductoring, storage and management software will be required regardless of which generation technology dominates each region. The second is firm and flexible generation because data centres, industries and cities cannot operate only when variable production is available.

The third is BESS and long-duration storage, not merely to store renewable electricity but to release connections, manage peaks, provide ancillary services and defer selected grid investments. The fourth is gas and LNG where they can provide capacity rapidly while additional infrastructure is constructed. The fifth is existing nuclear, new large reactors and eventually SMRs in markets capable of solving cost, construction and regulatory challenges.

The sixth is copper, uranium, graphite, lithium, rare earths and processing. The seventh is cooling and water because computing infrastructure transforms both into industrial inputs. The eighth is ports because expansion in grids, offshore wind, LNG, fuels, minerals, transformers and heavy equipment requires maritime infrastructure.

There is finally a ninth market that may become larger still: financing the productivity of all the others. Institutional capital does not merely need new projects. It needs measurable cash flows, contracts, understandable risk and evidence of performance. An investment releasing 20 MW from an existing facility, reducing an industry's energy consumption by 15% or increasing port throughput without constructing another port can produce a return as real as an entirely new asset.

BalGreen and DOIX: From bottleneck to yield

The method can become a repeatable architecture. DOIX identifies and quantifies the bottleneck. BalGreen determines which operating combination can solve it. Implementation reduces the loss or releases capacity. DOIX verifies the difference. Savings or incremental revenue create cash flow. That flow can potentially be structured through project finance, debt, bonds, private credit or other instruments according to the asset and jurisdiction.

Innovation does not necessarily require inventing another technology. It can mean making an improvement financeable that previously remained hidden inside operating expenditure.

If a network can connect an additional 100 MW through storage, digitalisation and intelligent management before constructing another transmission line, those 100 MW have value. If a factory produces more while using the same electrical connection, that productivity has value. If a port reduces turnaround time, that capacity has value. If a mine reduces water and energy consumption per tonne, that difference has value.

DOIX measures the value. BalGreen designs how to recover it. Financial markets can purchase the yield.

Analysis

Artificial intelligence is teaching an economic lesson extending far beyond technology. The more digital the economy becomes, the greater its dependence on certain physical assets can become. Models require chips, chips require factories, factories require electricity and water, electricity requires networks, networks require copper, copper requires mines and processing, mines require energy and logistics, and all of that infrastructure requires capital.

There is no digital economy separated from the physical economy. There is one system.

Understanding that connection allows us to anticipate where value may migrate next. During the first AI phase, scarcity existed primarily in advanced chips and computing capacity. The second phase is shifting part of that scarcity towards electricity and data centres. The next may shift it towards transformers, turbines, copper, water, connected land, generation capacity and professionals capable of constructing and operating those assets. Whenever one bottleneck begins to be resolved, capital immediately searches for the next.

The opportunity is to arrive first.

Conclusion

The great mistake would be to interpret the artificial-intelligence boom as an exclusively technological story. It may ultimately become one of the largest energy and industrial infrastructure cycles of this century, not because AI possesses an energy agenda but because computing has unavoidable physical requirements.

The winner will not necessarily be the country announcing the greatest number of data centres. It will be the country capable of delivering electricity, networks, water, minerals, infrastructure, permits and capital before its competitors. Nor will the winning company necessarily be the one producing the greatest amount of energy. It may be the one capable of making existing energy usable.

The world economy is moving from asking how much we can generate towards asking how much we can deliver. Then it will ask how much economic output we can obtain from every MW delivered.

That is where the next frontier begins. Not merely producing more. Connecting more. Storing better. Losing less. Using more. Measuring it. And financing the difference.

illuminem Voices is a democratic space presenting the opinions of leading Sustainability Thought Leaders, their views do not necessarily represent those of illuminem.

The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today.