AI-Driven Data Center Demand: Current State

Introduction

The demand for data center capacity is not in question. The constraint is. The four largest hyperscalers, Amazon, Microsoft, Google, and Meta, have guided to combined capital expenditure of more than $630 billion in 2026, up roughly 60% from a record $378 billion in 2025, with around three-quarters of it, close to $450 billion, aimed directly at AI infrastructure. That is one of the largest coordinated capital programs in corporate history. And yet vacancy in the primary data center markets sits at a record-low 1.4%, because the binding constraint has shifted. The scarce resource is no longer capital or even land. It is electricity. Understanding the current state of the sector means holding both facts at once: unprecedented demand, and a supply chain that physically cannot keep up.

The scale of hyperscaler spending is difficult to overstate, and the year-over-year jumps are the real story. Amazon guided to roughly $200 billion of capex in 2026 against $125 billion the prior year; Google to $175 to $185 billion, up from $91 billion; Meta to $115 to $135 billion, up from $72 billion; and Microsoft to $110 to $120 billion, up from $90 billion.

What makes AI capex different from the prior cloud build is intensity. Training and inference workloads pack far more computing power into each rack, pushing power density from the roughly 5 to 10 kilowatts per rack of traditional facilities toward 100 kilowatts and beyond. A given building now consumes a multiple of the electricity it would have a few years ago, which is why the demand story is measured in gigawatts rather than square feet and why the hyperscale segment, rather than retail or wholesale colocation, drives the market.

These commitments are large enough to dent the spenders' own cash flows. Amazon's free cash flow is projected to turn negative as the build accelerates, a sign of how completely AI infrastructure has reordered capital allocation at the largest technology companies. For a real estate banker, the takeaway is that the demand side of the data center market is anchored by a handful of the most creditworthy tenants on earth committing to spend at a pace they themselves admit cannot be fully absorbed, the tenant base profiled in the hyperscaler credit universe.

The capex numbers have outgrown even the hyperscalers' own cash flows, and the financing response has become one of the biggest stories in the sector. Amazon's free cash flow is turning negative as the build accelerates, and the spend now routinely exceeds what operating cash can cover, which has pushed the industry away from funding everything on the corporate balance sheet toward off-balance-sheet vehicles, joint ventures, and private credit.

The landmark transaction is Meta's Hyperion campus in Richland Parish, Louisiana, financed in October 2025 through a roughly $27 billion private-credit deal, the largest in history. The structure is the lesson: a special-purpose vehicle arranged by Morgan Stanley issued about $27 billion of A+-rated debt plus $2.5 billion of equity, anchored by PIMCO at around $18 billion and BlackRock, with Blue Owl Capital owning 80% of the vehicle and Meta only 20%. That lets Meta build a five-gigawatt campus without consolidating the full cost onto its balance sheet, protecting its credit profile and equity multiple while reaching a capital pool far larger than its own cash flow. It is exactly the kind of structuring a real estate banker is now asked to run.

Off-balance-sheet SPV financing

A structure in which a data center campus is built inside a separate special-purpose vehicle, funded largely by outside debt and equity so the bulk of the project cost never appears on the sponsor's own balance sheet. Meta's roughly $27 billion Hyperion vehicle, 80% owned by Blue Owl, is the defining example.

The pattern repeats across the sector. Stargate, the OpenAI, Oracle, and SoftBank joint venture announced in early 2025, targets $500 billion of data center investment, with its first site already live in Abilene, Texas and private managers like Blue Owl committing billions to individual campuses. Behind all of it sits the explosive growth of private credit, a roughly $2.1 trillion market projected to reach $3.5 trillion by 2030, which has found in AI infrastructure an asset class large enough to absorb its capital at scale. Data center asset-backed securitization is scaling alongside it, with JPMorgan projecting $30 billion to $40 billion of annual issuance in 2026 and 2027.

The financing has also reached down to the riskier players. Neoclouds, the AI-specialist providers such as CoreWeave that rent out GPU capacity, increasingly fund their chips with GPU-backed debt, borrowing against the hardware itself, a structure that holds only as long as the GPUs retain value and the contracts behind them perform. Insurers and pensions hunting for yield and long duration are pouring into the senior tranches of these structures, which is why the boom is now stress-testing parts of the insurance industry as much as the technology sector. For the banker, the takeaway is that the data center build is no longer just a real estate or technology story; it is a structured-finance story, and the firms that can assemble SPVs, syndicate private debt, and pair hyperscaler credit with institutional capital are the ones capturing the fees.

That demand has produced the tightest conditions in the property type's history. Vacancy across North American primary markets fell to a record-low 1.4% at the end of 2025. Net absorption hit a record 2,497.6 MW for the year, beating the prior record of 1,809.5 MW set in 2024, a market measured in megawatts of power rather than square feet because power, not floor area, is what tenants actually consume.

Preleasing

Preleasing is the practice of signing tenants to a data center before construction is complete, often before it has even broken ground. In the current market, hyperscale facilities are routinely fully preleased years ahead of delivery, which is why headline vacancy stays near zero: the space is spoken for long before it exists.

Northern Virginia remains the gravitational center of the market. It led all primary markets for net absorption in 2025 at 1,102.0 MW and, at roughly 4.0 GW of total inventory, is close to three times the size of the second-largest market, Atlanta, at about 1.5 GW, with Dallas-Fort Worth the third market to pass 1 GW. The concentration matters because it is precisely where the supply constraint now bites hardest, and it explains why the sector sits in the supply-constrained expansion phase described in cycle positioning across sectors.

The most important number in the sector is not a capex figure or a vacancy rate. It is four years, the average wait for a grid connection in the United States. That single fact reframes the entire investment thesis: a developer can have capital, a signed hyperscale tenant, and entitled land, and still be unable to deliver because the local utility cannot supply power for years.

Grid Interconnection

Grid interconnection is the physical and contractual connection of a facility to the electric transmission system, governed by the local utility and grid operator. For data centers it has become the gating item: securing an interconnection agreement, and the multi-year queue position that comes with it, now happens before site selection rather than after.

That queue is now visibly throttling new construction, producing a result that looks paradoxical against record demand.

The industry's response has been to stop waiting for the grid. Bring-your-own-power arrangements and on-site generation, including batteries and natural gas turbines, are becoming routine in large-scale data center planning. Power procurement has moved from an afterthought to the first item in the development pro forma, the dynamic explored in power as the binding constraint and increasingly handled alongside the energy and power coverage teams that understand generation and interconnection.

The Power Deals Behind the Headlines

The scramble for electricity has produced a wave of landmark procurement deals that now define the sector as much as the buildings do. On the nuclear side, Constellation Energy agreed to restart the shuttered Three Mile Island reactor, rebranded the Crane Clean Energy Center, under a 20-year agreement dedicating its full 835 megawatts to Microsoft, with Constellation investing roughly $1.6 billion to bring it back online. Amazon contracted with Talen Energy for up to 1,920 megawatts from the Susquehanna plant, and Meta signed a 20-year deal with Constellation for the 1.1-gigawatt Clinton plant. Beyond restarts, the same buyers are funding small modular reactors that will not deliver until late this decade. On the faster-to-deploy side, on-site natural gas is becoming standard: Meta's Hyperion campus is planned with roughly ten gas turbines, and across the market about 30% of planned new capacity is now designed to run independently of the grid, up from effectively zero a year earlier. The common thread is that hyperscalers, not utilities, are now the marginal buyers setting the price of clean power, and a platform's portfolio of energy contracts has become as valuable as its land.

The grid queue gets the headlines, but the physical supply chain is just as binding. The single scarcest component is the large power transformer, where lead times have stretched to roughly 128 weeks, about two and a half years, with generator step-up units running near 144 weeks and prices up on the order of 77% to 95% since 2019. A campus cannot energize without this gear, so the transformer order book now gates delivery as tightly as the interconnection queue, and the two delays compound rather than overlap. Industry analysts estimate that 30% to 50% of the capacity planned for 2026 will slip to 2028 as a result. Cooling is the next constraint: AI racks drawing 100 kilowatts and beyond have made air cooling insufficient, forcing a fast shift to liquid and direct-to-chip cooling that the supply base is still scaling to meet. Even the chips gate revenue, since a finished, powered building earns nothing until its GPUs arrive and are installed. The practical effect is that developers now run procurement as a strategic function, ordering long-lead transformers and turbines years ahead, sometimes before a site is fully permitted, and treating a secured equipment slot almost the way they treat a secured megawatt. The bottleneck is physical across the board, which is why capital alone cannot accelerate the build.

Because power is the constraint, new capacity follows electricity rather than population. The established giants, Northern Virginia foremost among them, are running into local opposition and grid limits, so developers are pushing into secondary markets chosen for power availability: Phoenix, Atlanta, Columbus, central Texas, and emerging sites near low-cost generation. The pattern is a deliberate migration toward wherever a developer can actually secure megawatts on a workable timeline, even if it means building far from the traditional connectivity hubs.

The same logic plays out internationally. The Gulf states are courting hyperscale capacity with abundant, cheap power and sovereign capital behind it, while the Nordics offer cool climates and clean grids. Europe's core markets, Frankfurt, London, Amsterdam, Paris, and Dublin, face the same power and permitting bottlenecks as Northern Virginia, and Dublin in particular has effectively paused new connections. The structural demand surge behind all of this, and why it favors the operators who control power, is laid out in data center real estate and the AI demand surge.

The capital pouring into the sector has produced the largest transactions in real estate. In October 2025, Aligned Data Centers was sold at an enterprise value of about $40 billion to a consortium that included the AI Infrastructure Partnership, whose members span Microsoft, MGX, BlackRock's Global Infrastructure Partners, Nvidia, and xAI. That single deal more than doubled the prior record, set only a year earlier by Blackstone and CPP Investments' roughly $16 billion take-private of AirTrunk in Asia. Across 2025, S&P counted 113 data center transactions worldwide worth more than $69 billion, a record by some $8 billion, and the pipeline shows no sign of slowing into 2026.

The deals share a logic. Buyers are paying for platforms that control the two genuinely scarce inputs, land with secured power and creditworthy hyperscale tenants, rather than for buildings as such, and they are increasingly led by infrastructure funds, sovereign wealth, and the chipmakers and AI labs themselves rather than by traditional real estate investors. The public markets feed the same demand: Blackstone's externally managed data center REIT, BXDC, came to market as a roughly $1.75 billion blind-pool listing precisely to give public investors access to the build, a deal tracked in the current REIT IPO pipeline. For a banker, this is where the sector's M&A and capital-markets fees concentrate, and why digital infrastructure has become a coverage priority spanning both real estate and energy and infrastructure groups.

No survey of the current state is complete without the debate over whether the build is rational, and a banker should be able to argue both sides. The arithmetic that worries skeptics is stark: the largest technology companies are set to spend on the order of $660 billion to $690 billion of capex in 2026 against direct AI revenue estimated near $51 billion, a roughly 13-to-1 gap between investment and current return. Critics point to circular financing, in which chipmakers, model developers, and cloud providers invest in one another in ways that can inflate apparent demand, and to the cash burn at neoclouds and AI labs, with OpenAI projecting steep losses even as revenue grows. The law firm Clifford Chance has flagged that something like 100 gigawatts of capacity is slated to come online between 2026 and 2030 without a clear, demand-tested picture of who will fill it.

The sharpest financial concern is a duration mismatch. The GPUs that justify these buildings depreciate fast, with useful lives plausibly as short as one to three years against the six-year schedules some operators book, while the debt financing them runs far longer. Lenders normally want an asset's life to exceed the loan's tenor; AI hardware inverts that, which is why observers draw uncomfortable parallels between today's data center securitizations and the layered structured finance that unwound in 2008.

The rebuttal matters as much as the worry. The real estate itself, the powered shell, the substation, the interconnection, and the cooling, has a useful life measured in decades, not the two or three years of the chips inside it, and it is leased to among the most creditworthy tenants on earth. Even if a generation of GPUs is obsolete quickly, the power and the building remain scarce and reusable. Whether the boom is a bubble therefore turns less on the real estate, which is genuinely supply-constrained, than on whether AI demand justifies the compute, a question that sits one layer above the property. The honest current-state read is that the buildings are real and scarce, the financing is aggressive and largely untested through a downturn, and the demand underneath the chips is the variable no one can yet prove.

The current state translates into a clear set of opportunities and risks. Powered land and shell buildings with secured interconnection command enormous premiums, because the power, not the concrete, is the scarce asset. Capital is flooding toward platforms that control sites with locked-in electricity, which is exactly why the largest M&A and capital-raising activity in real estate now runs through this sector, from the AirTrunk take-private to the BXDC listing in the current REIT IPO pipeline.

The valuation conversation has shifted with it. A site is no longer priced mainly on rent per square foot but on its secured power capacity and the credit of the tenant behind it, so a parcel with a signed interconnection and a hyperscale lease can trade at a multiple of an identical parcel still stuck in the queue. That repricing has pulled real estate bankers into territory once owned by power and infrastructure teams: underwriting a data center deal now means underwriting a power-purchase agreement, a substation timeline, and an equipment order book alongside the lease. The mandates that result, platform sales, take-privates, joint-venture recapitalizations, and the SPV financings described above, are among the largest in the entire market, which is why nearly every bank has moved digital infrastructure to the center of its real estate and infrastructure coverage.

The state of the market in 2026 is therefore a paradox that the best analysts can explain in one breath: the most capital-rich, demand-rich sector in real estate is supply-constrained not by money but by megawatts. That is why the build costs and time-to-power calculus covered in data center capex and time to power have become the decisive variables, and why power, more than anything on a traditional rent roll, will determine which projects actually get built over the next several years.