Computing power has a landlord

(Image caption) A large AI data center is located in the desert of the southwestern United States, surrounded by power transmission facilities and cooling systems, showcasing the reality of computing power moving from the "cloud" to land, electricity and infrastructure finance.

In the past, we said that AI was in the cloud.

The word "cloud" is light, seemingly weightless, devoid of land, power lines, water pipes, debts, and leases. A user opens an app, types in a sentence, and a model provides an answer in seconds. It appears to be an almost invisible capability, quietly existing behind the screen.

But AI doesn't actually live in the cloud.

It increasingly resides in Texas, Louisiana, Arizona, Ohio, and Virginia, in power-hungry data centers, and within power lines, substations, cooling systems, bond documents, and long-term leases. Now, it's also beginning to grace the books of global asset management firms like BlackRock.

A data center, a new asset contract

On July 28, Meta and BlackRock announced a partnership to develop and own a large data center campus in El Paso, Texas. The campus, with a development cost of approximately $14 billion and a designed computing capacity of 1 gigawatt, is expected to be operational by 2028. Under the agreement, funds managed by BlackRock will hold 80% of the joint venture, while Meta will retain 20%. Meta will contribute approximately $2.3 billion worth of land and assets under construction and receive approximately $1 billion in one-time distributions to adjust for the final shareholding ratio. BlackRock's investment will be financed through approximately $12.5 billion in debt. Meta will sign a lease agreement for the entire campus, with an initial term of four years and an extension option of up to 20 years through four renewals.

This news story appears to be about a tech company, but it's actually an infrastructure finance story at its core.

Meta needs computing power, BlackRock holds assets, debt investors value leases, credit, and future cash flow, and El Paso provides land, workers, electricity access, water resources, and local public spaces. These elements combined allow AI to clearly reveal its asset-heavy nature for the first time.

(Image caption) A large AI data center campus under construction in the El Paso desert, Texas, demonstrates how computing infrastructure is shifting from a cloud-based concept to a reality based on land, electricity, and heavy assets.

El Paso is not an isolated case.

This isn't the first time Meta has used this structure. In 2025, Meta partnered with Blue Owl Capital to develop the Hyperion data center campus in Richland Parish, Louisiana. According to Meta's announcement at the time, funds managed by Blue Owl held 80% of the joint venture project, while Meta retained 20%; both parties committed to contributing proportionally to support approximately $27 billion in total development costs, with Meta contributing land and assets under construction, and receiving approximately $3 billion in one-time allocations. Meta also used the facilities through leases, rather than keeping the entire project as its own asset on the company's books.

This comparison is very important.

El Paso is not an isolated financial innovation, but a template for financing AI infrastructure that is being replicated. Meta is gradually forming a new model across different states and partners: tech companies design, build, and use computing facilities, infrastructure capital holds most of the assets, the debt market provides long-term funding, and tech companies lock in computing power through leases.

In other words, AI computing power is starting to have landlords.

(Image caption) The server room inside the advanced AI data center, with its dense racks and closed-loop liquid cooling system, supports the computing power density required for large-scale model training and inference.

What did the landlord who bought the computing power buy?

In its official statement, Meta stated that the El Paso campus will support its AI technology development, with Meta serving as the initial sole tenant. The project is expected to create over 4,000 construction jobs at its peak and support approximately 300 operational jobs upon completion; currently, over 2,300 workers are on-site. BlackRock also stated that it will invest nearly $30 million through its Future Builders program to train over 12,000 electricians over three years to support Texas's growing demand for energy, infrastructure, and data center workers.

These commitments are worth recording. But in GFM's view, a more pressing question is: what changes are occurring in the financial structure of the AI industry when a company uses external capital to own a data center and then acquires computing power through long-term leases?

Previously, what attracted capital markets most to tech companies was their unconventional nature compared to traditional industrial companies. Their stories often revolved around software, advertising, social networks, search, and cloud services. They had low marginal costs, strong cash flow, limited fixed assets, and could rapidly scale up their economies of scale.

AI is rewriting this story.

Larger models require more chips, more chips require more power, and more power requires transmission capacity, cooling capacity, land, construction permits, and financing arrangements that can span multiple years of construction. When AI companies no longer need just engineers and code, but rather gigawatt-level computing power parks, their economic models begin to resemble those of the telecommunications, energy, logistics, and infrastructure industries.

Meta is buying more than just usage rights; it's locking in a gateway to future computing power. BlackRock isn't buying ordinary real estate; it's buying long-term, rentable, and valuable infrastructure. Debt investors see construction progress, lessee credit, lease terms, cost of capital, electricity costs, and exit strategies. This is closer to project financing than the traditional tech stock narrative.

The era of heavy assets for asset-light technology companies

A Reuters report cites data from BofA Global Research showing that AI-related bond issuance reached $270 billion by early July this year, nearly double the total for 2025. Meanwhile, the market is raising more questions about the returns, cash flow, and future operating costs of large tech companies' AI spending.

This is a very real problem.

Meta is neither Amazon Web Services nor Microsoft Azure. It currently lacks a cloud business of comparable scale to sell idle computing power to external enterprises. Its AI investments ultimately need to recoup their revenue through advertising tools, Meta AI applications, smart glasses, recommendation systems, and other product forms. This payback path isn't nonexistent, but the market needs a clearer timeline and evidence of commercialization.

From this perspective, El Paso is not an isolated project, but rather a method for large tech companies to cope with the capital pressures of AI. Instead of tying all their assets to their balance sheets, tech companies are outsourcing some of their infrastructure to external capital, acquiring usage rights through leases. This approach increases expansion speed and makes financial performance appear more flexible.

But the risk hasn't disappeared. It's simply been redistributed to different participants.

Who will bear the cost of construction overruns? Who will absorb the increased financing costs? If AI demand falls short of expectations, will the assets become idle? If grid upgrade costs exceed expectations, who will foot the bill? If local residents begin to worry about electricity bills, water resources, and public service pressures, what conditions should local governments impose on technology companies and asset holders?

(Image caption) Thousands of construction workers are working on a data center construction site, reflecting the scale of jobs created and the pace of construction at the peak of the Meta-BlackRock joint venture project.

Data center public billing

These issues are no longer just theoretical in El Paso.

Meta stated that it will cover the full cost of energy used in its data centers to avoid negative impacts on consumers; the company also stated that it will increase the availability of sufficient clean and renewable energy to match the 100% power usage of its data centers. Regarding water, Meta said the El Paso campus will use a water-saving closed-loop liquid cooling system and has committed to restoring local water resources.

These claims need to be put to the test in the coming years. For local communities, the commitment itself is not the end goal. People need to know how the commitment is measured, who audits it, how it is corrected if deviations occur, and whether residents can continue to receive transparent information throughout the project's operation.

The public nature of AI infrastructure is shifting from abstract discussion to concrete bills.

In July, New York State announced a statewide moratorium on new hyperscale data centers, suspending some state environmental permits for up to a year in order to establish a new regulatory framework. The New York State government explicitly stated that the rapid growth of data centers could impact electricity costs, natural resources, the power grid, and local communities, and that the state hopes to establish a set of standards to protect users, the environment, and communities.

This indicates that AI data centers are no longer just projects for tech companies and investors. They are impacting local job markets, power planning, water supply arrangements, road construction, and public finances. Their value to a city cannot be judged solely by investment and job creation; it must also be considered in terms of tax arrangements, electricity responsibilities, water resource impacts, the quality of permanent jobs, and whether the city incurs sunk costs if the project fails to operate as expected.

Infrastructure is never just an asset; it's also public relations. This is true for airports, highways, ports, power plants, and reservoirs, and it will be true for AI data centers as well. They serve private companies but rely on public resources; they generate revenue and change local life; they are written into financial models and appear in residents' electricity bills, water usage arrangements, and urban planning.

Why did BlackRock get involved?

BlackRock's entry into AI data centers is not surprising.

The global asset management industry has been searching for new types of infrastructure assets that can support long-term cash flow. In the past, it was highways, ports, logistics centers, power grids, energy projects, and communication towers; now, AI data centers are beginning to be included in the same asset family.

If the lease is long enough, the lessee has strong enough credit, and the risks of electricity and operations can be measured, computing power can be packaged into a stable-yield asset by the financial market.

This is also the most historical aspect of this news story.

AI was once described as a race to see who could build the most powerful models, the fastest reasoning, and the most human-like answers. Today, the other side of this race has emerged: who can acquire electricity at the lowest cost, who can convince the debt market of future computing power needs, who can persuade local governments and communities to accept hyperscale data centers, and who can turn still uncertain AI applications into financeable long-term leases.

(Image caption) The high-voltage transmission lines and substations connecting the data center highlight the huge demand for power infrastructure and grid capacity for a 1-gigawatt computing power park.

The Asian market will face the same problem.

This also has reference value for the Asian market.

Capital from China, Hong Kong, Singapore, Japan, South Korea, and the Middle East is all looking for entry points into AI infrastructure. But what truly matters isn't simply including the phrase "AI data center" in fundraising materials, but rather the ability to clearly answer a few fundamental questions.

Who is the ultimate lessee? What is the lease term? How are electricity costs locked in? Does the debt maturity match the asset's lifespan? What conditions did the local government provide? What external costs do residents bear? If AI demand declines, how can the asset be reused?

These questions determine whether a data center is long-term infrastructure or just another financial transaction driven by the AI narrative.

Having a landlord for computing power is just the beginning.

Meta and BlackRock's El Paso project is not yet operational. Whether construction can be completed on schedule, whether financing costs are manageable, whether the lease can support the debt, whether the AI products can generate sufficient revenue, and whether the local community believes the project's benefits outweigh its costs—these are all questions that require time to answer.

But today we can already see a clear path: AI is moving beyond mere software imagination and into the world of infrastructure finance.

In this world, computing power has landlords, models have electricity meters, AI has creditors, local governments have negotiating power, and residents have the right to inquire about costs.

This is why GFM paid attention to this news.

Our focus isn't on Meta building another data center, nor on BlackRock making another investment. Our focus is on how AI is changing the way capital is organized, how it is changing infrastructure ownership, and how it is bringing the growth story of tech companies into land, electricity, debt, and public governance.

The future competition in AI will not only take place on model leaderboards, but also on power contracts, bond documents, local hearings, data center leases, and asset management portfolios.

When a technology begins to require landlords, leases, debts, and public commitments, it no longer belongs solely to engineers and product managers. It begins to permeate the financial structure of society and the daily order of public life.

(Image caption) The BlackRock Future Builders program trains electricians to cultivate technical talent for Texas’s growing energy, infrastructure and data center industries.

GFM Observation

Meta's collaboration with BlackRock serves as a reminder that the underlying competition in AI has entered a new phase. Model capabilities and chip supply remain important, but whoever can organize capital, electricity, land, long-term leases, and community responsibility into a sustainable system will determine the true boundaries of AI infrastructure.

For investors, this is no longer an era where only the total investment amount matters. It's about who holds the assets, who bears the debts, who leases the computing power, who pays for the electricity, and who bears the risks.

For cities, this is no longer an era that focuses solely on employment and tax revenue. It's about whether residents are protected, whether the power grid is resilient, whether water resources are respected, and whether commitments can be monitored in the long term.

For the media, this is no longer an era of simply chasing AI product releases. It's about seeing how technology becomes institutionalized, how capital transforms industries, and how infrastructure impacts the lives of ordinary people.

Having a landlord for computing power is just the beginning. Next, society needs to know who pays for this new house, who uses it, who benefits from it, and who will bear the bills it leaves behind.

Disclaimer

This article is for news research and public discussion purposes only and does not constitute investment, legal, tax or transaction advice; relevant data and transaction arrangements are subject to company announcements and regulatory documents.