The grid has started saying no to artificial intelligence

This article is part of In conversation about sustainable finance & emission reduction systems, a new series by Diego Balverde. You’re reading volume 27 of the Breaking news series. Here is volume 26

Artificial intelligence has just collided with a frontier that no algorithm can cross on its own: the physical capacity of the electricity grid, and what is happening in Texas this week should be read far beyond the United States because one of the world’s most aggressive territories for data-centre development is requiring new interconnection projects to be audited before they can continue advancing while proposed demand has expanded to several times the historical maximum load of the system and an extraordinary share comes from data centres themselves, turning what appears to be a technical interconnection problem into a global warning about the speed at which the digital economy is colliding with physical infrastructure.

The message is much larger than Texas: artificial intelligence can attract hundreds of billions of dollars, technology companies can buy chips, funds can finance campuses, developers can acquire land and governments can offer incentives, but if generation, transmission, transformers, substations, storage, cooling, permits and interconnection capacity do not exist, the project has no electricity and without electricity artificial intelligence does not exist. The digital economy has reached the point where the decisive scarcity is no longer only advanced processors but usable electrons delivered at the correct location and time.

This week is simultaneously exposing the opposite side of the problem in California, where rising volumes of solar and wind generation are being curtailed because grids and demand cannot absorb all available production during certain hours, revealing an extraordinary paradox: one part of the American electricity system is attempting to slow new loads because it fears it cannot serve them while another part is wasting renewable electricity because it cannot move, store or use that power when it appears.

The energy problem of the twenty-first century is therefore not simply producing more. It is connecting better. Shifting better. Storing better. Coordinating generation and demand. Building transmission before capital becomes trapped waiting for an interconnection. Making data centres, industry, ports, electric vehicles, BESS, renewables, nuclear power, flexible gas and hydropower operate within one architecture rather than as independent markets competing for insufficient infrastructure.

The Texas pause also turns electricity interconnection into a visible financial variable because behind every request sit acquired land, contracts, deposits, engineering studies, suppliers, construction commitments, potential debt and capital waiting to begin generating returns. When an electricity approval is delayed, the project does not simply stop technically. Capital becomes immobilised. Land remains unproductive. Interest continues accumulating. Contracts may need renegotiation. Equipment prices can change. Developers lose visibility over schedules. Banks increase discounts applied to future cash flows. Investors demand greater returns to compensate for regulatory and execution risk. The grid bottleneck becomes a financial spread.

This represents a fundamental transformation because interconnection was treated for years as an engineering issue sitting far away from investment committees. It now belongs at the centre of credit analysis. A data-centre project without firm access to electricity is not yet a complete digital asset. It is an option on a future interconnection. A BESS without access to the correct market is not yet profitable infrastructure. It is equipment waiting for a revenue architecture.

A renewable park behind a congested network is not simply clean generation. It is an asset potentially exposed to curtailment and lost revenue. A factory seeking to electrify processes without understanding the actual capacity of its substation is pushing an operating risk into the future. Capital will increasingly differentiate brutally between projects controlling their energy and projects that have merely reserved a position in a queue.

This also changes bond architecture because the enormous investment needed to reinforce grids, construct substations, add storage and create firm generation will require utility issuance, infrastructure bonds, private debt, project bonds, corporate hybrids, debt funds, asset finance and new structures in which major consumers participate directly in the cost of infrastructure they require. Artificial intelligence can therefore become one of the largest indirect sources of infrastructure-bond supply of the next decade. The securities may not say artificial intelligence. They may say transmission, utility, power, storage, data-centre infrastructure, grid modernisation, generation or digital infrastructure, but the economic engine increasingly will be the need to supply electricity to a digital economy whose consumption is growing much faster than networks.

Goldman Sachs, BlackRock, CPP Investments, Brookfield, Macquarie, KKR, Apollo and other large global capital providers have different profiles, but all represent the scale of financial capacity that can participate in an energy rebuild requiring long-duration debt, patient equity and sophisticated allocation of generation, demand, regulatory and technology risk. Galdana Ventures and other innovation-focused investors can find opportunities in energy-management software, grid intelligence, load optimisation, distributed energy systems and technologies capable of increasing the productivity of existing infrastructure.

BalGreen’s opportunity lies precisely in not competing with them as capital providers but in constructing the product that such capital can finance. BalGreen needs to arrive before the bond. Before the fund. Before the bank. It must identify the operating loss, demonstrate why it exists, design the solution, quantify the capacity released and create sufficiently strong evidence for financial markets to convert that improvement into cash flow.

DOIX has a much more important function here than producing dashboards because it can measure demand curves, peaks, available capacity, transformer utilisation, congestion, interruptions, curtailment, BESS behaviour, cooling consumption, asset availability and savings resulting from each intervention. Using this information, BalGreen can determine whether the real problem requires a battery, microgrid, substation reinforcement, flexible-demand management, cooling efficiency, distributed generation, solar integration, firm-power contracting or simply elimination of unnecessary consumption before requesting more capacity from the network.

This sequence matters because the most expensive way to solve an energy problem is to build infrastructure to supply inefficiency. If a data centre can reduce or shift part of its load during the most critical hours, the size of some required investments can fall. If BESS removes peaks, an interconnection can be used more productively. If cooling systems reduce their consumption, megawatts become available without building new generation. If selected workloads can operate at different times, computing begins to behave as flexible demand. If data centres in different regions redistribute selected workloads according to electricity availability, artificial intelligence can use its own software to reduce pressure on the infrastructure it requires.

This is one of the most powerful innovations of the new system because it changes the paradigm from completely rigid digital demand toward partially manageable demand. The data centre stops being only a grid problem and can become a resource when an economic relationship exists that rewards flexibility. A capacity market can pay for availability. A utility can provide tariffs linked to peak reductions. BESS can buffer curtailment. Contracts can differentiate critical loads from shiftable processes. Software can decide within seconds which operations must remain online and which can wait. Electricity and computing therefore begin integrating into a single operating system.

This also creates an enormous storage market because Texas does not necessarily need to choose between accepting or rejecting all new demand if part of that demand arrives accompanied by new generation, BESS, flexibility, microgrids and contractual obligations to reduce consumption when the system requires it. The same logic applies globally. New major consumers should arrive with energy, not only demand.

A data centre requesting five hundred megawatts could finance part of the additional generation, storage and network reinforcement required to integrate it. An industry electrifying a massive process could participate in the new substation it requires. A port introducing shore power, electric cranes and truck charging should combine those loads with BESS, generation, intelligent management and efficiency improvement. The energy transition will accelerate when new consumers stop approaching networks simply as requests and begin arriving as integrated packages of demand plus capacity.

This is particularly relevant for BalGreen Ports because ports combine major loads, land, grids, industrial customers and logistics within one location. A port can become an energy platform combining BESS, solar, shore power, heavy charging, industrial refrigeration, data centres, microgrids and grid services. DOIX can measure every flow, identify peaks and verify how much value coordination releases. BalGreen can design the operating package. Improvement can reduce costs, release capacity and generate new revenue. Capital can finance the resulting assets. The port moves from electricity consumer to energy node. The same architecture applies to industrial parks, airports, cities, hospitals and logistics corridors.

The Texas shock should also change the employment debate because digital expansion constrained by infrastructure means the relevant labour shortage will not exist only among software engineers. Economies will lack electricians, high-voltage specialists, BESS technicians, system operators, civil engineers, transformer experts, commissioning technicians, maintenance teams, SCADA operators, EMS specialists, industrial cooling professionals, energy-cybersecurity experts, transmission-construction workers and infrastructure-finance analysts. A region seeking to attract ten billion dollars of digital investment while lacking sufficient electrical workforce will discover the human bottleneck can be as serious as the grid bottleneck.

BalGreen Academy should therefore form part of the solution from the beginning rather than after assets are built. Every project should include the skills map needed to operate it for twenty years. Training becomes risk mitigation. An investor has greater confidence when a local workforce can maintain the asset. A utility gains resilience when it does not depend entirely on foreign specialists. A port reduces downtime when systems can be repaired internally. An industry protects production when trained technicians are available. Employment creation can therefore enter the financial model because it reduces operating risk and improves continuity.

The grid saying no to artificial intelligence is also sending governments a message: competing through cheap land, low taxes and rapid permits is insufficient. Energy infrastructure must become part of industrial policy. A region should know how many megawatts it can offer immediately, how many it can add within three years, which lines must be constructed, which transformers need to be secured, what share can be supplied by renewables, how much storage is required, which firm capacity remains and what the total cost of every major new interconnection will be. This information should exist before investments are attracted, not afterward. Otherwise governments convert announcements into queues and queues into unproductive capital.

New industrial policy will increasingly be megawatt policy. Territories will compete on connection time in the same way they previously competed on labour costs or taxation. A company may prefer paying more for electricity if it receives an interconnection three years earlier. A utility capable of delivering capacity with certainty can become more valuable than one offering theoretically inexpensive power while forcing projects to wait. Electrical time becomes money.

This principle will directly affect real-estate valuations, municipal bonds, utility debt and industrial location. Land close to nodes with available capacity can multiply in value. Existing infrastructure can become a competitive advantage. Older generation sites with valuable interconnections can acquire second lives. Closed industrial sites can become digital campuses because transmission and substations already exist. The real asset may not be the land or building but the grid connection. That logic will transform entire infrastructure markets.

It also forces a discussion of who pays. If a large technology company needs billions in additional network investment, a significant share of that cost should be contractually connected to the project so society does not finance dedicated infrastructure for a private customer that may change plans years later. Deposits, guarantees, take-or-pay agreements, capacity contracts, long-term commitments and obligations to add new supply can become normal conditions for major loads.

Developers with genuine projects may accept greater discipline in exchange for a more predictable path to interconnection. Speculative projects will disappear from queues. This can eliminate enormous quantities of phantom demand and enable grid operators to plan more accurately. For credit markets this can be positive because uncertainty declines, although initial costs rise and strong projects become separated from projects dependent on cheap capital and excessive expectations.

A new opportunity will appear for investors capable of identifying infrastructure supported by genuinely contracted demand. The risk of an artificial-intelligence bubble is not located only in technology valuations. It can also appear in infrastructure built for loads that never arrive. Every power plant, transmission line, battery and substation should therefore be analysed under alternative scenarios. What happens if the data centre reduces scale? Can the asset serve other customers? Can BESS participate in different markets? Can the line supply industry or housing? Can generation sell to the wider grid? The greater the possible reuse, the lower the stranded-asset risk. Multi-user infrastructure will be more defensible than infrastructure dependent on one customer.

This philosophy should guide BalGreen when designing assets because the optimal solution is not necessarily the one maximising capacity for the current client but the one creating residual value for the entire system. DOIX can verify that value and produce the information banks, insurers, funds and bond markets require. The real financial innovation is converting released capacity, avoided interruptions, recovered energy and reduced peaks into variables capable of supporting debt.

A system demonstrating recurring annual savings can finance against part of those savings. A BESS with contracted services can support project finance. A portfolio of grid improvements can be packaged for institutional investors. Thousands of smaller projects can be aggregated under one methodology until they reach scale. BlackRock or CPP Investments do not need to analyse one thousand batteries individually if a platform offers governance, data, standardisation and a diversified portfolio.

Goldman Sachs and other investment banks can find more value structuring assets with verifiable cash flows than financing a narrative. Galdana Ventures can seek software converting operating complexity into automated intelligence. BalGreen can occupy the centre of the system if it succeeds in connecting operations, DOIX, execution and capital without becoming merely another hardware seller.

The grid has started saying no to artificial intelligence, but that is not the end of digital growth. It is the beginning of a more disciplined phase in which AI must learn to coexist with physics, finance the infrastructure it needs and demonstrate the value it brings to the territories receiving it. Texas is showing that even one of the world’s strongest energy economies can reach a point where accumulated requests exceed planning capacity and force the system to stop and distinguish genuine demand from speculative demand. California simultaneously demonstrates that producing more electricity without being able to move or store it also destroys value.

Between those two extremes lies the opportunity: connect new loads with new generation, recover energy currently being wasted, build storage, reinforce grids, use existing assets more effectively, digitalise operations and finance only improvements capable of demonstrating performance.

The next major energy company will not necessarily be the one possessing the most generation. It may be the one capable of coordinating generation, network, storage and demand more effectively. The next major winner from artificial intelligence may not only be the company building the best model. It may be the company securing the megawatts required to run it. The next major bond market may not emerge from climate policy but from the urgent need to finance the grid supporting the digital economy.

Which governments actually know how much new demand they can connect before compromising reliability, which utilities will distinguish real digital projects from speculative requests, which technology companies will accept financing generation, storage and transmission in exchange for firm access, which banks will turn electricity capacity into a central credit variable, which global funds will finance bottlenecks rather than only chasing the visible companies of artificial intelligence, which regions will transform technical training into competitive advantage, which ports can become energy and computing hubs and how much value can BalGreen create if DOIX measures where the constraint exists, BalGreen designs the package releasing capacity, execution reduces losses and congestion, DOIX verifies the outcome and institutional capital converts those recovered megawatts into measurable, verifiable and credible financial yield?

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.