Meta shares AI infrastructure costs through $2 billion asset sale – Image: Xpert.Digital
Meta relies on partner financing for AI infrastructure: Billion-dollar deal for data centers
The strategic shift in AI infrastructure
Meta Platforms has made a significant strategic shift in financing its artificial intelligence infrastructure. The company is selling $2 billion worth of data center assets to attract external partners to fund the extensive infrastructure needed for AI development. This decision reflects a fundamental change among tech giants, which have traditionally self-funded their expansion but are now facing the rapidly rising costs of AI data centers.
The new strategy is evident in Meta's quarterly report, in which the company announced that a plan to divest certain data center assets was approved in June, reclassifying $2.04 billion in land and construction projects as "held for sale." These assets are to be transferred to third parties within the next twelve months for the purpose of jointly developing data centers.
Related to this:
Financing the AI revolution
Spending on AI infrastructure is reaching unprecedented levels. Meta has raised its capital expenditure forecast for 2025 to between $66 billion and $72 billion, an increase of approximately $30 billion over the previous year. These massive investments are part of an industry-wide trend, with the four largest technology companies—Meta, Amazon, Alphabet, and Microsoft—projected to spend a combined total of up to $364 billion in their respective fiscal years of 2025.
Susan Li, Meta's CFO, confirmed in an analyst call that the company is actively exploring opportunities to collaborate with financial partners on the joint development of data centers. These partnerships are intended to help fund the massive capital expenditures planned for the coming year. While Meta still plans to finance the majority of its capital expenditures internally, certain projects could attract "significant external financing" and offer greater flexibility should infrastructure requirements change.
The superclusters of superintelligence
Mark Zuckerberg has unveiled ambitious plans to create a "superintelligence" that will require hundreds of billions of dollars in investment in AI data centers. At the heart of this strategy are two groundbreaking projects called Prometheus and Hyperion, designed to deliver industrial-scale computing power through superclusters.
Prometheus, a one-gigawatt data center, is scheduled to go online in 2026, making it one of the first AI infrastructures of this scale. The center will be built in New Albany, Ohio, and is expected to have over 500,000 graphics processing units (GPUs). Hyperion, the even more ambitious project, will be built in Louisiana and can scale up to five gigawatts over several years. Zuckerberg describes these facilities as so large that one of them would cover "a significant portion of the ground area of Manhattan.".
The challenges of AI infrastructure
The development of these massive AI data centers presents significant technical and logistical challenges. The energy intensity of AI workloads far exceeds that of traditional data centers. According to research data from the International Energy Agency, data center electricity consumption will rise to 945 terawatt-hours by 2030, roughly equivalent to Japan's annual electricity consumption. AI workloads already account for 24 percent of server power consumption and 15 percent of total data center energy demand.
Water scarcity presents another critical challenge. One of Meta's data centers in Newton County, Georgia, has already led to water shortages in some households. These environmental impacts are increasing the pressure on technology companies to find more sustainable solutions for their AI infrastructure.
Talent acquisition and market dynamics
Alongside its infrastructure investments, Meta has poured billions into recruiting leading AI talent. The company offers AI researchers compensation packages that can reach up to $200 million over four years, a hundred times what their peers earn. This aggressive talent acquisition is part of Meta's strategy to compete with rivals like OpenAI, Google, and Anthropic.
The new division, Meta Superintelligence Labs, led by Alexandr Wang, the former CEO of Scale AI, focuses on developing foundational models and conducting fundamental AI research. The company has recruited prominent researchers from organizations such as OpenAI, Google DeepMind, and Anthropic, including Shengjia Zhao, a co-creator of ChatGPT, who serves as the team's chief scientist.
The economics of AI infrastructure
Global spending on AI infrastructure is showing explosive growth. According to McKinsey analysis, an estimated $5.2 trillion in capital expenditures will be needed for AI data centers by 2030. This figure reflects the sheer scale of the investments required to meet the growing demand for AI computing power.
Hardware spending dominates investments. In 2025, an estimated 80 percent of the $644 billion in AI spending will go toward hardware that manufacturers have upgraded with AI-enabled capabilities. Spending on AI servers is projected to increase from $135 billion last year to $180 billion, a rise of 33 percent.
New partnerships and business models
Rising costs are forcing technology companies to develop innovative financing models. BlackRock, Microsoft, and the Abu Dhabi-based investment fund MGX, for example, have formed a partnership for AI infrastructure investment, initially aiming to raise $30 billion and ultimately provide $100 billion for financing data centers and energy projects.
These partnerships are emerging against the backdrop of a changing market in which even the largest technology companies are seeking external support. Amazon plans to invest over $100 billion, Microsoft $80 billion, and Alphabet $85 billion by 2025. These coordinated investments demonstrate the scale of the race for AI dominance.
Technological innovation and efficiency
The industry is working intensively to improve the efficiency of AI systems. New developments such as DeepSeek's "Mixture of Experts" architecture, which comprises a network of smaller, specialized models, promise improved training efficiency. These innovations could help control the otherwise rapidly increasing demand for electricity.
Advances in cooling technology will also be crucial. Liquid cooling is rapidly becoming the standard for new AI data centers, as it can handle the high heat loads of the latest graphics processors. Companies like Accelsius are developing innovative cooling solutions such as NeuCool racks, which can support up to 100 kilowatts of computing power.
Energy supply and sustainability
Energy supply for AI data centers is becoming a critical challenge. Goldman Sachs predicts that global electricity demand from data centers will increase by 50 percent by 2027 and by as much as 165 percent by the end of the decade. This demand is driving investment in new power plants and the modernization of electricity grids.
Technology companies are increasingly turning to renewable energy and nuclear power. Meta, Microsoft, and others are exploring small modular reactors as a solution to their energy needs. At the same time, they are using record-breaking power purchase agreements for renewable energy to reconcile their climate goals with growing energy demands.
Geopolitical dimensions
AI infrastructure is evolving into an area of strategic national importance. The US government has launched various initiatives to strengthen its domestic AI infrastructure. The Stargate project, a joint initiative of OpenAI, Oracle, and SoftBank, plans to invest $500 billion in AI data centers in Texas.
President Trump signed an executive order leasing federal land belonging to the Departments of Defense and Energy for gigawatt-scale AI data centers and new clean energy facilities. These measures aim to accelerate the development of the next generation of AI infrastructure in America.
Market segmentation and specialization
The AI infrastructure market is becoming increasingly differentiated. Hyperscale data centers, encompassing at least 10,000 square meters of space and 5,000 servers, are being specifically designed for large AI workloads. Simultaneously, specialized edge AI data centers are emerging for applications requiring low latency.
Colocation providers are adapting their offerings to AI requirements and providing specialized cooling and energy solutions. Companies like CyrusOne, Cologix, and Digital Realty are investing billions in modernizing their facilities for AI workloads.
The future of AI infrastructure
The AI infrastructure landscape will change dramatically in the coming years. By 2028, global spending on AI infrastructure is expected to exceed $200 billion annually. These investments will drive new business models, technological breakthroughs, and societal change.
The shift towards shared infrastructure models, as demonstrated by Meta's $2 billion sale, could become the new standard. These partnerships allow companies to share risks, deploy capital more efficiently, and manage the massive investments required for the AI revolution.
Mark Zuckerberg's vision of superintelligence may be years away, but the infrastructure that will enable it is already being built. Meta's decision to bring in external partners marks a turning point in how technology companies finance and shape the future of AI. This strategic shift could serve as a blueprint for other companies and usher in a new era of collaboration in the technology industry.
Related to this:
Your AI transformation, AI integration and AI platform industry expert
☑️ Our business language is English or German
☑️ NEW: Correspondence in your native language!
Konrad Wolfenstein
I and my team are happy to be available to you as your personal advisor.
You can contact me by filling out the contact form here wolfenstein@xpert.digital:or simply call me at +49 7348 4088 965. My email address is
I'm looking forward to our joint project.
