The capital behind artificial intelligence is becoming as consequential as the technology itself. Sovereign wealth funds and public pension plans are taking a more direct role in financing the data centers, power supplies and investment platforms on which the industry depends.
New research released November 12 by PwC projects that sovereign wealth funds will hold $22.4 trillion in assets by 2030, while public pension funds will reach $36.3 trillion. Together, that would put $58.7 trillion under the stewardship of institutions capable of committing capital across decades.
For family offices assessing AI infrastructure investments, the significance extends beyond the size of those portfolios. These institutions are increasingly helping determine which projects get built, who operates them and how other investors gain access.
From Fund Investors to Direct Owners
PwC’s study, based on a comparative analysis of the world’s 350 largest public investors, describes a shift toward more hands-on ownership. Its accompanying analysis identifies approaches that include directly owning assets, developing investment platforms with partners and creating businesses or infrastructure where established assets do not yet exist.
The report cites Global SWF figures showing roughly $66 billion of state-investor capital flowed into AI and digitization in 2025. That figure covers a broader category than data centers alone, an important distinction when assessing how much investment has reached physical infrastructure.
One example is the AI Infrastructure Partnership, established by BlackRock, Global Infrastructure Partners, Microsoft and Abu Dhabi-based MGX. BlackRock announced in 2025 that the Kuwait Investment Authority would join. The initiative was created to mobilize $30 billion in equity capital, with the potential to reach $100 billion including debt financing.
Those are fundraising and investment ambitions, rather than evidence that the full amount has been deployed. They nonetheless illustrate the scale of the partnerships forming around AI’s physical requirements.
The Investment Case Runs Through the Power Grid
An AI application may reach users in seconds. Delivering the electricity and computing capacity behind it takes considerably longer.
The International Energy Agency’s 2026 analysis of energy and AI projects global data-center electricity consumption will rise from 485 terawatt-hours in 2025 to approximately 950 TWh in 2030, representing around 3% of global electricity demand. Within that total, electricity consumption at AI-focused data centers is projected to triple.
The agency also identifies constraints that capital alone cannot immediately resolve: grid connections, energy equipment, advanced chips, permitting and local opposition. For an investor, the practical questions begin with when a facility can receive power, which customer has committed to use it and who bears the cost if construction runs late.
These considerations create a wider investment universe than ownership of server buildings. Power generation, electrical equipment, cooling systems and connectivity may all participate in the buildout. Each carries its own economics, competitive pressures and risks.
What Family Offices Should Examine
For family offices, large institutional partnerships could create opportunities to invest alongside experienced operators. Competition could also make attractive assets more expensive. Access to a prominent investment platform should begin a diligence process, rather than conclude it.
A useful starting point is the underlying revenue agreement. Is demand supported by a signed contract with a financially strong customer, or by a forecast of future leasing? How long does the contract last, and can it be canceled? Does the investor own the building, the computing equipment, the power supply or a combination of all three?
Financing deserves equal attention. A project’s expected return may depend on refinancing debt at favorable rates, completing construction within budget or selling to another institutional buyer. An investment presented as long-duration infrastructure can still face near-term funding pressure.
Portfolio overlap is another concern. A family may already hold substantial AI exposure through technology shares, venture funds and private credit. Adding a data-center investment can introduce a different legal structure without necessarily introducing a different source of economic risk.
The relevant exercise is to trace the customers and cash flows across the portfolio. Several apparently separate holdings may ultimately depend on the same technology companies sustaining their capital expenditure.
Long-Term Capital Still Needs Price Discipline
Sovereign funds and pension plans bring considerable resources to complex projects, but their mandates are not interchangeable with those of a family office. A government-linked investor may also pursue domestic development or strategic objectives. A family’s investment committee must account for distributions, future capital calls, taxes and the ability to exit.
AI infrastructure has an evident appeal: an expanding industry requires physical assets that cannot be built overnight. The return to an investor, however, will depend on the price paid, the contracts secured and the capital required after the initial commitment.
For family offices, PwC’s findings make the market worth studying closely. The strongest investment case will be found in the details of a particular asset, not simply in the company it keeps.
