AI Data Center Acquisitions and Investment: Who’s Buying What in 2026

The Pulse

Private equity investment in US data centers hit 45.70 billion dollars in 2025, the highest total in at least five years, and that figure represents 72% of the entire 63.35 billion dollar data center investment market for the year. A single deal drove most of that number.

That deal was the 40 billion dollar acquisition of Aligned Data Centers, backed by a consortium that reads like a list of the most powerful capital pools in the world: MGX Fund Management, Global Infrastructure Management, Microsoft, xAI, Nvidia, Temasek Holdings, the Kuwait Investment Authority, and BlackRock. Sovereign wealth funds, hyperscalers, chipmakers, and traditional private equity are now co-investing in the same physical assets, a level of capital convergence that has no real precedent in enterprise technology infrastructure.

The question this activity answers is not whether AI infrastructure investment is real. It clearly is, at a scale exceeding the Apollo program and the interstate highway system combined. The more useful question is who is actually buying, what they are buying, and why the ownership structure of AI infrastructure is shifting so dramatically toward private capital and away from straightforward corporate capital expenditure.

Core Significance

Why it matters:

• Private markets are becoming structurally essential to AI infrastructure financing, not just a supplementary capital source:  Private infrastructure funds raised a record 221 billion dollars in 2025 with average fund sizes jumping to 1.8 billion dollars, and as of May 2026 there were 695 infrastructure funds actively fundraising, targeting an aggregate 555 billion dollars in additional capital specifically for this asset class.
• Hyperscale deals remain the preferred structure for institutional capital, but the terms underlying those deals are shifting in ways that reveal real risk pricing:  Investors favor hyperscale tenant deals for easier credit underwriting compared to neocloud or GPU-as-a-Service tenants, and are increasingly hedging exposure to sub-investment-grade tenants through parent guaranties and hyperscaler credit wrappers rather than relying on tenant credit alone. [Ropes Gray data center investment 2026 AI demand power constraints]
• Deal activity across the entire AI infrastructure value chain reached a scale with no comparable precedent in technology history:  Data center transactions hit a record 61 billion dollars in 2025, AI-related mergers and acquisitions exceeded 146 billion dollars in disclosed deals, and hyperscaler debt issuance to fund the gap between front-loaded capital spending and back-loaded revenue reached 108 billion dollars in 2025 alone. [TMT IB Guide AI investment cycle infrastructure buildout MA]

Deep Context: Why hyperscalers are buying power plants, not just signing power contracts

The most consequential structural shift in AI infrastructure investment through 2025 and into 2026 is the move from Power Purchase Agreements toward direct ownership of power generation assets. A Power Purchase Agreement is a financial contract to buy renewable energy from an off-site project, useful for meeting carbon commitments but structurally unable to solve the physical problem of delivering power through an already congested grid.

Google’s acquisition of Intersect Power in December 2025 marked the clearest signal of that shift, moving from the earlier December 2024 model, a partnership with Intersect Power and TPG Rise Climate to catalyze 20 billion dollars in co-located generation investment, toward outright ownership of generation assets. Direct ownership gives a hyperscaler control over physical power supply that a purchase agreement simply cannot guarantee, a meaningful distinction given that 62% of data centers are currently power-constrained. [EnkiAI hyperscaler energy 2026 race build private grid]

As covered in our AI data center power consumption report, power availability, not chip supply or capital, has become the primary bottleneck constraining how fast AI infrastructure can actually come online. That constraint is precisely why the ownership structure of power generation itself has become as strategically important to hyperscalers as the data centers those power plants feed.

Private equity has built a genuinely new asset class around this dynamic

Firms including KKR, Energy Capital Partners, and TPG Rise Climate are forming multi-billion-dollar partnerships specifically to fund the joint construction of data centers and the dedicated power generation required to run them, treating the combination as a single investable asset class rather than two separate infrastructure categories. Goldman Sachs projected US utilities would need to invest 50 billion dollars in new generation to support data centers, a figure now being matched or exceeded by individual private equity partnerships targeting the sector alone.

That scale of private capital formation around a single infrastructure category, purpose-built to solve a specific physical bottleneck rather than to fund general-purpose real estate, is itself evidence of how central power constraints have become to the entire AI buildout, a dynamic that did not exist in any prior data center investment cycle.

Data Insights

By the numbers:

All figures from S&P Global, Goldman Sachs, and named investment banking and infrastructure research cited inline.

• The five largest hyperscalers are projected to spend approximately 602 to 700 billion dollars combined on capital expenditure in 2026, with roughly 75% directed specifically at AI infrastructure:  That spending level represents approximately 1.9% of US GDP, a figure that research comparisons place above the combined historical cost of the Apollo program, the interstate highway system, and the broadband expansion era. [Intellectia AI infrastructure investment 2026 600 billion hyperscaler]
• Applied Digital’s April 2026 lease agreement illustrates the scale individual data center transactions have reached even outside the largest hyperscaler deals:  The company signed a 7.5 billion dollar total contracted value lease with a new US-based investment-grade hyperscaler tenant covering 300 megawatts of critical IT load at its 430 megawatt Delta Forge 1 campus, bringing the company’s total contracted lease revenue across its portfolio to more than 23 billion dollars.[Applied Digital hyperscaler tenant Delta Forge 1 lease 2026]
• 74 new AI-ready data center facilities broke ground in 2026 alone, spread across 28 states, cumulative to more than 690 billion dollars in committed capital expenditure from six major hyperscalers:  Microsoft alone pledged 80 billion dollars for fiscal 2025, while Amazon Web Services committed 100 billion dollars across new US regions through 2028, illustrating that the buildout is geographically dispersing well beyond the traditional data center hubs of Virginia and a handful of other established markets. [ValueAdd VC AI buildout tracker 690 billion 74 data centers]

Table 1: Major AI data center investment and acquisition activity, 2025 to 2026

Table 2: Where AI infrastructure capital is coming from

The Business Case: What the ownership shift means for enterprises buying AI infrastructure capacity

The practical consequence of this capital convergence for enterprises purchasing AI compute capacity, whether through cloud contracts or direct colocation leases, is that the underlying data center they are contracting with is increasingly owned by a consortium of financial investors rather than a single operating company with a straightforward corporate balance sheet.

That ownership structure matters for contract durability and pricing. A facility financed through a special purpose vehicle with private credit backing has different default and continuity risk characteristics than one financed on a hyperscaler’s own corporate balance sheet, a distinction increasingly relevant given that S&P Global data shows 3,988 US data centers currently exist, with more than a quarter still in the planning stage rather than actually under construction.

As covered in our Hardware-as-a-Service report, the broader enterprise shift toward consumption-based infrastructure procurement applies directly here. Enterprises signing long-term capacity commitments should increasingly evaluate the financial structure and capital backing behind a data center operator, not just the technical specifications of the facility, since ownership structure is now a genuine input into contract risk assessment.

Expert Nuance: The financing gap itself is becoming the central risk investors are pricing

The single most important number in the entire AI data center investment story may not be a spending total, but a mismatch. Goldman Sachs Research’s own credit strategists flagged that the growth in hyperscaler capital expenditure estimates is meaningfully outpacing the growth in actual data center construction, a gap that Goldman explicitly states warrants close monitoring as an indicator of long-term financing needs. [Goldman Sachs private markets data center financing role 2026]

That gap is precisely why the asset-backed securities market for data centers, currently around 25 billion dollars, faces projected take-out financing needs approaching 300 billion dollars, an order-of-magnitude expansion that the current private credit and structured finance ecosystem has not yet been tested at scale to absorb.

S&P Global’s own market analysis frames the underlying uncertainty plainly: a handful of hyperscalers and a handful of frontier AI labs are all investing as though they will have a business in this space a decade from now, a bet that may prove correct but has not yet been validated by a full economic cycle, meaning today’s financing structures are being built on capital expenditure projections rather than demonstrated long-term revenue. [S&P Global private equity investment data center 5 year high]

Strategic Outlook

1. Watch whether Blackstone’s planned public company for stabilized data center assets successfully launches, since it would create the first dedicated public vehicle for this specific asset class:  With Blackstone describing its data center portfolio as exceeding 150 billion dollars already built plus 160 billion dollars in pipeline development, a successful public listing would provide the clearest available signal of how public capital markets price mature, income-generating AI infrastructure assets separately from the growth-stage capital expenditure story. [Intellectia AI investment supercycle 2026 hyperscaler spending]
2. Expect continued consolidation of power generation ownership by hyperscalers as the power-constraint bottleneck persists:  Following Google’s Intersect Power acquisition, other hyperscalers facing similar generation-capacity constraints are likely to pursue comparable direct-ownership transactions rather than relying solely on Power Purchase Agreements, particularly in regions where grid interconnection queues extend multiple years.
3. The asset-backed securities financing gap is the single number most worth tracking for systemic risk in this cycle:  A market currently sized around 25 billion dollars facing a projected 300 billion dollar take-out requirement represents genuine untested territory, and how smoothly that financing scales, or does not, will be a more reliable signal of AI infrastructure’s financial health than any individual capital expenditure announcement.

Key Question Answered

Who is actually buying AI data centers in 2026, and why has private capital become so central to the buildout?

The buyer base has consolidated into three overlapping categories. Hyperscalers themselves, Amazon, Microsoft, Google, Meta, and Oracle, are spending a combined 602 to 700 billion dollars in 2026 capital expenditure, with roughly 75% directed at AI infrastructure. Private equity and infrastructure funds, led by firms like Blackstone with more than 150 billion dollars already deployed, have become the second major buyer category, raising a record 221 billion dollars in 2025 alone. The third category, sovereign wealth funds and strategic co-investors, joined the largest deals directly, illustrated by the 40 billion dollar Aligned Data Centers acquisition backed by a consortium spanning MGX, Microsoft, xAI, Nvidia, Temasek, the Kuwait Investment Authority, and BlackRock simultaneously.

Private capital has become central because hyperscaler capital expenditure, however large, is not growing fast enough on its own balance sheet to fund the full scale of the buildout, and because off-balance-sheet structures, special purpose vehicles backed by private credit, let hyperscalers convert what would otherwise be capital expenditure into operating lease expense while keeping the debt off their own corporate books. The power generation side of the buildout has followed a parallel logic, with direct asset ownership, illustrated by Google’s Intersect Power acquisition, increasingly replacing simple purchase agreements as hyperscalers seek guaranteed control over an increasingly scarce physical resource.

The Takeaway

The AI data center investment story in 2026 is no longer primarily a story about how much money big tech companies are spending on their own balance sheets. It is a story about the emergence of an entirely new asset class, one that has pulled in sovereign wealth funds, private equity giants, chipmakers, and traditional infrastructure investors simultaneously, all co-investing in the same physical facilities.

That convergence reflects genuine confidence in AI infrastructure as a durable, income-generating asset class, but it also concentrates a specific kind of risk. The financing structures being built today, asset-backed securities, private credit facilities, and consortium equity deals, are sized against capital expenditure projections that assume continued exponential AI demand growth, a bet that has not yet been tested through a full economic cycle or a meaningful AI demand slowdown.

For enterprises and investors watching this space, the practical signal worth tracking is not the next headline capital expenditure number. It is whether the private credit and asset-backed securities market can actually absorb the scale of financing this buildout requires, a gap Goldman Sachs itself has flagged as warranting close monitoring, and whether the ownership consortiums forming around individual mega-deals like Aligned Data Centers prove to be a stable long-term financing model or an early symptom of a market moving faster than its financing infrastructure can safely support.