Record Data Sheet

Amount
$700 billion
Entity
World's largest data
Vertical
Technology · Energy · Infrastructure
Record Type
Capital Expenditure Forecast
Period
2026
Confirmed
14 October 2026

More than $700 billion in data centre investment is expected in 2026. But behind the expansion of artificial intelligence lies a second infrastructure race: power generation, transmission networks, financing and access to electricity. The consequences will extend well beyond Silicon Valley.

The artificial intelligence investment cycle is increasingly becoming an energy investment cycle. Technology companies are committing unprecedented amounts of capital to data centres, while utilities, infrastructure funds, power generators and governments are being drawn into projects that will reshape electricity markets.

According to the International Energy Agency (IEA), the world’s largest data centre operators are expected to spend more than $700 billion in capital expenditure in 2026, largely on new facilities. The same report identifies growing pressure on local electricity networks, equipment supply chains, water resources and electricity affordability.

The central question is no longer simply how much companies are investing in AI. It is whether the financial returns generated by these investments will justify the infrastructure required to operate them — and how the costs and risks will be distributed among technology companies, investors, utilities, governments, businesses and households.

At The Expenditure, the relevant distinction is between announced investment, financing commitments, actual expenditure and the economic obligations that remain after a project becomes operational.

1. Who is financing the expansion of artificial intelligence?

The scale of the investment is already visible in the financial statements of the largest technology companies.

Alphabet raised its full-year 2026 capital expenditure guidance to between $195 billion and $205 billion. During its third-quarter earnings call on October 28, CFO Anat Ashkenazi explained that the investment programme would increase depreciation and data centre operating expenses, including energy costs. CEO Sundar Pichai has positioned infrastructure as a central component of the company’s AI and cloud strategy.

Amazon is also expanding its infrastructure at considerable speed. AWS recorded third-quarter 2026 revenue of $46 billion, up 39% year over year. CEO Andy Jassy has argued that investment in AWS infrastructure is supporting the division’s growth. However, Amazon’s trailing twelve-month free cash flow had fallen to an outflow of $42.4 billion by September 30, 2026, primarily reflecting increased purchases of property and equipment, much of which the company attributes to AI investment.

These figures illustrate two different dimensions of the same investment cycle. Alphabet is communicating a capital expenditure plan that will increase infrastructure and depreciation costs. Amazon is reporting strong cloud growth alongside a substantial reduction in free cash flow.

Neither figure, on its own, establishes whether the investment will generate an adequate return. That depends on utilisation, pricing, customer demand, operating costs, asset lives and the cost of financing.

The question for investors is straightforward: how much future revenue must these facilities generate before their capital costs, electricity bills and financing obligations become economically attractive?

Official company disclosures: Alphabet Q3 2026 earnings call · Amazon Q3 2026 results.

2. Who owns the infrastructure when technology companies do not want to finance everything themselves?

One of the most revealing developments is the growing participation of institutional investors in data centre ownership.

Meta and BlackRock: a $14 billion infrastructure venture in El Paso

On July 28, 2026, Meta announced a strategic venture with BlackRock to develop and own a data centre campus in El Paso, Texas.

The project provides a clear example of how technology infrastructure is being financed through partnerships between operating companies and institutional capital.

The announced structure is as follows:

  • Total development costs: approximately $14 billion.
  • Ownership: funds managed by BlackRock will hold 80%, while Meta will retain 20%.
  • BlackRock’s initial cash contribution: approximately $4.9 billion at financial close.
  • Meta’s contribution: land and construction-in-progress assets valued at approximately $2.3 billion.
  • Debt financing: part of BlackRock’s investment will be funded through a $12.5 billion debt financing.
  • Capacity: approximately 1 GW of computing capacity.
  • Expected operation: capacity is expected to begin coming online in 2028.

The distinction between the figures matters. The $14 billion represents the estimated development costs of the buildings and associated power, cooling and connectivity infrastructure. The $12.5 billion is a separate financing amount supporting part of BlackRock’s investment; it should not be added to the development cost.

The corporate responsibilities are also significant. Mark Zuckerberg, Meta’s founder and CEO, described the partnership as a way to accelerate infrastructure development. Larry Fink, Chairman and CEO of BlackRock, presented the transaction as an opportunity to provide institutional capital to strategic infrastructure.

Meta will be the initial sole occupant of the campus and will lease the entire facility from the venture. The leases have an initial four-year term with extension options. Meta will also provide residual value guarantees with an aggregate threshold of approximately $13 billion, declining over time.

This last provision deserves particular attention. Although Meta transfers 80% ownership to the BlackRock-managed funds, it remains closely connected to the economic performance of the facility through its lease obligations and residual value guarantees.

The transaction also identifies the financial and technical advisers involved: Morgan Stanley and JPMorgan Securities advised Meta; Arup provided technical and environmental engineering services; and Marsh provided project risk analysis and insurance services.

Does transferring ownership really transfer the risk? Not entirely. The financing structure distributes ownership, but Meta’s long-term lease commitments and guarantees preserve significant economic exposure.

For institutional investors, the key issue is whether the value of the asset remains sufficient to support its financing and contractual obligations if computing demand, technology requirements or rental economics change.

Source: Meta and BlackRock — El Paso data centre venture.

3. Who is building the electricity supply behind Google’s $17 billion Wyoming investment?

On November 5, 2026, Google announced its first data centre in Wyoming, describing it as a $17 billion investment in the state.

The project is particularly relevant because the investment involves more than buildings and computing equipment. It also includes a dedicated energy supply and several private-sector partners.

The companies identified by Google include:

  • Google: data centre investment and operation.
  • Tallgrass Energy: energy infrastructure and carbon capture and storage.
  • Black Hills Energy: regional energy partner.
  • American Electric Power (AEP): contracted to support a 702 MW fuel-cell development.
  • Related Digital: data centre development partner.
  • Bloom Energy: manufacturer of the fuel cells used in the planned energy development.

The planned 702 MW fuel-cell development will use natural gas, with carbon capture and storage intended to capture a large share of associated carbon emissions. The captured carbon is expected to be stored at Tallgrass’s Eastern Wyoming Sequestration Hub.

Google says it will pay for 100% of the electricity it uses and cover infrastructure needs directly driven by its growth.

The company announcement was published by Steve Walter, Google’s Vice President of Global Infrastructure and Energy. The project also sits within the wider infrastructure strategy overseen by Sundar Pichai, CEO of Alphabet and Google.

However, the $17 billion announcement does not provide a complete public breakdown of expenditure by contractor, financing source or project component. The amount should therefore be treated as Google’s announced investment, not as a verified sum already spent or as a disclosed contract value for each partner.

Can a private technology investment strengthen the local electricity system rather than compete with existing consumers for capacity?

Google’s model is an attempt to answer that question through dedicated energy infrastructure and direct responsibility for the costs associated with its expansion. Whether the arrangement delivers lower costs for other customers will depend on actual investment, operating performance and the terms governing the wider electricity system.

Source: Google — Investing in Wyoming’s infrastructure, energy and workforce.

4. Can data centres finance their own energy infrastructure?

The Wyoming project is not an isolated case.

In January 2026, Meta announced agreements with several nuclear energy companies and operators, including Vistra, TerraPower, Oklo and Constellation.

The company expects the agreements to support up to 6.6 GW of new and existing clean energy capacity by 2035.

The arrangements include financial support for the continued operation and expansion of existing nuclear facilities, as well as the development of advanced nuclear technologies. Vistra’s involvement includes the Perry and Davis-Besse plants in Ohio and Beaver Valley in Pennsylvania.

Meta says these projects will supply electricity to the grids supporting its operations, including its Prometheus supercluster in New Albany, Ohio. The company has also stated that it pays the full cost of the electricity consumed by its data centres.

This approach is strategically important. Instead of relying exclusively on electricity purchased from existing market supply, large technology companies are increasingly entering long-term arrangements with power producers.

However, announced capacity is not the same as electricity already generated. Some projects depend on regulatory approvals, construction schedules, financing and the technical performance of new facilities.

The financial question is whether long-term power agreements can provide sufficient certainty for data centre operators without transferring excessive costs or risks to other electricity customers.

Source: Meta — Nuclear energy projects and up to 6.6 GW of capacity.

5. What happens when governments provide land and private companies provide the capital?

A further example comes from Paducah, Kentucky, where the U.S. Department of Energy announced a partnership on July 29, 2026 involving a planned investment of more than $100 billion.

The private-sector participants include:

  • Brookfield
  • NextEra Energy
  • Big Rivers Electric Power Corporation
  • Jackson Purchase Energy Cooperative
  • Paducah Power System

The proposal involves redeveloping portions of a former Department of Energy site into a data centre and high-performance computing campus, accompanied by new energy infrastructure.

The announced plans include:

  • A 1.8 GW AI and high-performance computing campus.
  • 2 GW of new natural gas-fired generation planned by NextEra Energy.
  • Up to 2.6 GW of battery energy storage.
  • Transmission infrastructure upgrades.
  • Approximately 8,000 expected construction jobs.
  • Approximately 600 expected permanent jobs.

U.S. Energy Secretary Chris Wright presented the project as a way to use federal land to attract private investment, expand energy supply and create employment.

The announcement describes the investment as privately funded. Nevertheless, the public disclosure does not provide a complete breakdown of the capital commitments by individual private-sector participant, the debt structure or the final allocation of project costs.

The distinction is important because the public sector is contributing access to a federal site and participating in the broader development framework, while private companies are expected to finance and develop the infrastructure.

Who captures the economic benefits when public assets help unlock private investment?

Potential benefits include construction employment, permanent jobs, local tax revenues, improved infrastructure and additional electricity supply. But the economic outcome will depend on the final contractual structure, the distribution of costs, the tax arrangements and whether the promised capacity is actually delivered.

A project of this scale should be assessed not only by the investment headline, but also by the public resources involved and the long-term obligations assumed by each participant.

Source: U.S. Department of Energy — Paducah data centre and energy partnership.

6. Will households pay more for electricity because of AI?

This is the question that gives the investment cycle its wider social significance.

The expansion of data centres creates additional demand for electricity. In regions where generation and transmission capacity are constrained, new large consumers can increase the need for power plants, substations, transformers, transmission lines and other infrastructure.

The International Energy Agency warns that the effects can be particularly pronounced at local level, even when data centres represent a relatively small share of national or global electricity consumption.

The United States Energy Information Administration (EIA), in its October 2026 Short-Term Energy Outlook, forecasts average residential electricity prices of:

YearAverage residential price
202416.5 cents/kWh
202517.3 cents/kWh
202618.2 cents/kWh
202718.7 cents/kWh

The projected increase between 2025 and 2027 is approximately 8.1%.

For a household consuming 10,000 kWh annually, the difference between the 2025 and 2027 average prices would represent approximately $140 in additional annual electricity expenditure, assuming unchanged consumption.

This is an illustrative calculation based on the EIA’s average price forecasts. It is not an estimate of the additional cost caused by AI.

The distinction is essential. The EIA identifies several factors affecting electricity prices, including natural gas prices, weather, transmission and distribution upgrades and capacity market costs. Its forecast does not attribute the projected increase to data centres alone.

Nevertheless, the expansion of large electricity consumers creates a legitimate economic question: if new infrastructure is required to serve private data centre projects, who pays for it?

The answer depends on tariff design, connection agreements, power purchase contracts, utility investment decisions and regulatory protections.

If costs are allocated appropriately to the companies driving additional demand, the impact on other customers can be limited. If costs are socialised across a wider customer base, households and smaller businesses could bear part of the burden.

Source: U.S. EIA — October 2026 Short-Term Energy Outlook.

7. Why are electricity regulators intervening?

The Federal Energy Regulatory Commission (FERC) has already taken action on this issue.

On June 18, 2026, the regulator issued orders to all six regional grid operators under its jurisdiction, requiring them to justify or reform tariff rules governing the connection of data centres, manufacturing facilities and other large electricity users.

The proceedings focus on how large loads connect to the transmission network and how the associated costs and risks are allocated.

FERC Chairman Laura V. Swett described the objective as improving the integration of large energy users while protecting consumers and providing certainty for investors.

The regulatory intervention reflects a structural problem: a data centre may be financed by a private company, but the electricity network serving it is part of a wider system used by households, factories, hospitals and other businesses.

If a utility builds infrastructure for a large customer and that customer later delays construction, reduces its demand or abandons the project, the remaining financial exposure must be addressed through the relevant contractual and regulatory framework.

This is why connection charges, minimum payments, collateral requirements, long-term commitments and exit provisions matter.

The regulatory question is not whether data centres should be connected. It is whether the rules ensure that the parties generating the additional costs also bear an appropriate share of the financial risk.

Source: FERC — Large-load interconnection proceedings.

8. Is the electricity market already showing the pressure?

The PJM Interconnection capacity auction provides an important indication of the challenge.

PJM operates the electricity market across parts of 13 U.S. states and the District of Columbia, serving more than 67 million people.

Its July 2026 auction for the 2028/2029 delivery year procured 138,318 MW of unforced capacity. The clearing price reached the FERC-approved cap of $325 per MW-day.

The total cleared capacity was 6,831 MW below PJM’s reliability requirement. Only 525 MW of new generation and generation uprates cleared in the auction.

PJM also identified the continued addition of large data centre loads to its electricity demand forecast.

These figures should not be interpreted as proof of an immediate electricity shortage or as a direct calculation of household bills. Capacity auctions are designed to secure future resource adequacy, and the auction clearing price does not represent the complete cost paid by every electricity consumer.

Nevertheless, the auction demonstrates the importance of bringing additional generation and demand-side resources online as large electricity consumers expand.

Can the electricity supply chain increase capacity at the same speed as technology companies are committing capital to computing infrastructure?

The answer will influence the cost of electricity, the timing of new data centre projects and the returns available to investors in the energy sector.

Source: PJM — 2028/2029 capacity auction results.

9. Is Europe facing the same infrastructure problem?

The issue is not limited to the United States.

According to a European Commission report, data centres consumed approximately 68 TWh of electricity in the European Union in 2024. Consumption is projected to rise to 114 TWh by 2030, equivalent to around 3.2% of total EU electricity demand.

The figures underline the growing importance of data centre infrastructure in European energy planning.

The European Commission has also opened a consultation on minimum energy performance standards for data centres, with a proposal planned for the second quarter of 2027. The consultation is not itself a binding regulation.

The challenge for European policymakers is to attract investment in digital infrastructure without creating additional pressure on electricity affordability, grid capacity and environmental resources.

Spain: how much data centre capacity is waiting to connect?

Spain provides a particularly interesting example.

On October 1, 2026, Red Eléctrica published a breakdown of granted demand access capacity across its transmission network.

Data processing centres accounted for 6,329 MW, representing 34.2% of the total 18.52 GW of granted access capacity across the categories published.

However, none of that data centre capacity had yet entered service at the date of publication.

This is not a measure of actual electricity consumption, nor does it represent every data centre project in Spain. The figures concern granted access capacity on the transmission network and exclude requests relating to distribution networks.

Even with those qualifications, the data reveal the scale of the potential pipeline of electricity demand competing for access to the transmission system.

For Spain, the investment opportunity extends beyond data centre operators. It includes grid infrastructure, electricity generation, storage, cooling systems, engineering, construction and electrical equipment suppliers.

Can Spain convert its potential access capacity into operational infrastructure without creating a disproportionate cost for existing electricity customers?

That will depend on grid investment, permitting, project maturity, financing and the allocation of connection costs.

Sources: European Commission — Data centre energy efficiency report · Red Eléctrica — Demand access capacity by consumption type · European Commission — Consultation on data centre performance standards.

10. What does this mean for the wider economy?

The economic consequences of the AI infrastructure boom extend across several sectors.

For households, the main concern is electricity affordability. If infrastructure costs are not allocated transparently, residential customers could face part of the financial burden associated with new large consumers. However, dedicated generation, long-term contracts and properly designed tariffs can help limit that exposure.

For small and medium-sized businesses, higher electricity costs could affect operating margins, particularly in energy-intensive sectors such as manufacturing, food processing, logistics and hospitality.

For workers, the construction of data centres, power plants, transmission networks and battery facilities creates demand for electricians, engineers, construction workers, technicians and specialist contractors. These jobs can provide long-term economic benefits, although permanent employment is usually much lower than peak construction employment.

For local governments, major infrastructure projects can increase investment, tax revenues and economic activity. They can also place pressure on water resources, roads, housing and public services, depending on the location and scale of the development.

For investors, the expansion creates opportunities in utilities, power generation, grid equipment, energy storage, engineering and infrastructure finance. But the returns depend on regulatory certainty, contracted demand, construction costs and the ability of customers to pay for the electricity and computing services they consume.

For the stock market, the investment cycle creates both opportunities and valuation risks. Companies supplying electricity, transformers, cooling systems, generators and grid infrastructure may benefit from higher demand. Technology companies, meanwhile, must demonstrate that infrastructure expenditure can translate into sustainable revenue growth, operating margins and free cash flow.

The central financial risk is a mismatch between the useful life of infrastructure and the speed at which technology, demand and customer economics change.

A power plant or transmission line may operate for decades. A particular generation of AI chips, servers or computing architecture may become commercially less competitive much sooner.

That difference in asset lives creates an important question for lenders and investors: who remains responsible for the infrastructure if the original commercial assumptions no longer hold?

The Expenditure’s Verdict: Who Should Bear the Cost of the AI Infrastructure Boom?

The investment in artificial intelligence is creating a new category of strategic infrastructure. Technology companies are no longer simply purchasing servers and software. They are increasingly involved in financing power generation, developing energy partnerships, securing long-term electricity supply and negotiating access to national grids.

The participation of BlackRock in Meta’s El Paso project, Google’s partnership with energy and infrastructure companies in Wyoming, and the proposed private-sector development at Paducah demonstrate how the financial structure of the AI industry is changing.

The benefits could be substantial: higher productivity, new industries, skilled employment, stronger digital infrastructure and additional investment in electricity generation.

But the financial case must be assessed beyond the headline investment figure.

A $100 billion project is not automatically a $100 billion economic benefit to the surrounding population. A data centre that creates thousands of temporary construction jobs may employ far fewer people once operational. A private investment can generate public benefits, but only if the contractual, fiscal and infrastructure arrangements distribute those benefits appropriately.

The same principle applies to electricity prices. Households should not automatically be expected to finance infrastructure built primarily to serve private commercial demand. At the same time, new generation and grid investment can benefit the wider electricity system if projects are properly designed and financed.

The most important test is therefore not how much capital the technology sector can deploy, but whether the resulting infrastructure generates returns sufficient to cover its economic costs without imposing disproportionate risks on the rest of society.

Who should ultimately finance the electricity infrastructure behind artificial intelligence: technology companies, institutional investors, utilities, governments or the consumers who depend on the same grid?

Official Sources

  1. International Energy Agency — Powering Artificial Intelligence
  2. Alphabet — Q3 2026 Earnings Call
  3. Amazon — Q3 2026 Financial Results
  4. Meta — El Paso Data Centre and BlackRock Venture
  5. Google — Wyoming Data Centre Investment
  6. Meta — Nuclear Energy Agreements
  7. U.S. Department of Energy — Paducah Partnership
  8. U.S. Energy Information Administration — October 2026 Outlook
  9. Federal Energy Regulatory Commission — Large Load Interconnections
  10. PJM — 2028/2029 Capacity Auction
  11. European Commission — Data Centre Energy Efficiency
  12. Red Eléctrica — Demand Access Capacity