Wall Street Perspective on Data Center Markets

Big Tech data centers are the hottest assets in the world… but is it hard for developers to cash out? Here’s the paradox reshaping capital markets: • Amazon, Microsoft, and Google pre-lease billions in new hyperscale capacity years before construction ends. • Developers deliver fully leased, mission-critical facilities… • Yet when it’s time to sell? Buyers vanish. Why? • Stabilized hyperscale data centers = massive $3B+ price tags • Locked into 10–15 year leases → limited upside for buyers • Only 7% of investors target stabilized “core” assets (CBRE) The result: Developers are reinventing exits with debt securitizations instead of equity sales. → $13.4B in ABS + SASB data center deals closed in H1 2025, double last year (JLL) → Blackstone/QTS → $1.5B CMBS refinance in Atlanta + Richmond → DataBank → $1B ABS backed by Atlanta, NY, Virginia facilities Meanwhile, creative equity plays are emerging: • Forward takeouts (buyers fund development + commit to buy at stabilization) • Hyperscaler purchase options (Amazon/Microsoft buying back facilities years into leases) • Minority stake sales (developers recycle capital while staying in the operator seat) The bigger shift? Data centers went from niche infrastructure to 13% of the SASB market in just 4 years (Goldman). What this signals: → Liquidity is flowing into bonds, not asset sales. → Core buyers are thin, but new funds (Blue Owl + Qatari SWF just raised $3B) are being built to fill the gap. → The future of data center finance may look more like Wall Street than Main Street. Are securitizations the permanent exit strategy for hyperscale developers, or will a new wave of core buyers finally step in? Full story: https://lnkd.in/gTJ-vupT

The scale and financing of the global data-center build-out between 2025 and 2028 highlight a structural shift in how AI infrastructure is being funded. Morgan Stanley projects nearly $2.9 trillion in cumulative spending over this period, with tech companies themselves accounting for approximately $1.4 trillion. This level of self-funding is unprecedented and underscores both the strategic importance of AI infrastructure and the financial strain it is placing on balance sheets. What is equally notable is the growing reliance on external financing. Private credit is expected to provide roughly $800 billion, while private equity and related investors will contribute an additional $350 billion. Meanwhile, corporate bond markets remain a significant pillar, with about $200 billion in issuance, and asset-backed securitizations adding another $150 billion. This diversification of funding sources illustrates that the AI data-center arms race is no longer solely a big-tech undertaking—it is increasingly becoming a system-wide capital deployment channel involving the entire credit spectrum. This is reflected in recent bond-issuance patterns. According to Bank of America, the largest US AI companies—Alphabet, Amazon, Meta, Microsoft, and Oracle—have sharply accelerated issuance in 2025. While issuance over 2020–2024 remained relatively stable, 2025 has seen a surge, with the bulk of it concentrated between September and November. This aligns with rapidly rising capex budgets and the growing need to finance data-center expansions, power procurement, and accelerated hardware refresh cycles. The combined picture reveals a sector increasingly dependent on debt to sustain its growth trajectory. With interest rates still elevated and capex requirements compounding, the long-term sustainability of this funding model will depend heavily on whether AI-driven revenues scale fast enough to justify both the leverage and the pace of infrastructure expansion. Sources: Bank of America and Morgan Stanley

According to Bank of America, securities backed by data centers, fiber, and towers could grow nearly 46% to $115B by the end of 2026. And data centers already dominate the space, representing 61% of today’s $79B market. The deals are getting bigger, too. Meta's $29B Louisiana expansion was financed by PIMCO and Blue Owl Capital, while J.P. Morgan and MUFG just arranged $22B for Vantage Data Centers. These transactions aren’t isolated... they mark the beginning of an arms race to fund the AI backbone, with capital flowing at unprecedented scale. AI adoption is forcing hyperscalers like Microsoft, Alphabet Inc., Amazon, and Meta to ramp up cloud and digital infrastructure spend. What’s equally interesting is the capital markets side. Risk premiums on digital infra ABS have tightened across data centers, fiber, and towers, leaving investors weighing relative value against MBS, CMBS, and consumer loan ABS. Morningstar and BofA strategists note that while future spread tightening may be limited, these securitizations still offer an attractive comparative yield. This is a window into how securitization markets evolve to fund new categories of “real economy” infrastructure — and how private lenders, banks, and asset managers are competing to structure and underwrite the next wave of investments. Full Bloomberg piece here: https://lnkd.in/esuJUrMw

Everyone's chasing data center land. Almost everyone is missing the real constraint. It's not fiber. It's not even land. It's power. U.S. Interior Secretary Doug Burgum said at the Prologis conference: "To win the AI arms race against China, we've got to figure out how to build these artificial intelligence factories close to where the power is produced, and just skip the years of trying to get permitting for pipelines and transmission lines." Translation: The next generation of data centers won't be built where the land is cheap. They'll be built where the power is available. Three implications for dirt investors: 1. Nuclear Proximity = New Premium: Amazon already signed deals with Dominion Energy near the North Anna nuclear power station in Virginia and expanded partnerships with Talen Energy at the Susquehanna nuclear plant. Sites within transmission distance of existing nuclear facilities just became exponentially more valuable. 2. Warehouse Conversions Accelerate: If Prologis is eyeing their 6,000 buildings for data center conversion, every industrial site with surplus power capacity needs re-evaluation. What looks like a struggling warehouse today might be a data center tomorrow. 3. Grid Capacity > Geographic Desirability: Constellation Energy CEO Joseph Dominguez noted that data economy customers "want to run their systems 24-7" with "firm pricing so that they know the price for energy for 20 years". Long-term power contracts are becoming the new land entitlements. But here's what nobody's talking about: The same power constraints driving this opportunity are also creating massive project risks. According to a recent CoStar analysis, data centers will account for up to 60% of total power load growth through 2030. But there's a timing mismatch: data centers take 2-3 years to build, while power system upgrades take 8 years. That gap is forcing developers to either wait or find sites with existing capacity. The Community Resistance Factor Data Center Watch estimates $64 billion in data center projects were blocked or delayed over a recent two-year period. There are now 142 activist groups across 24 states organizing against data center development. Northern Virginia alone-the nation's largest data center market-has 42 activist groups fighting projects. Reasons cited: water consumption, higher utility bills, noise, decreased property values, loss of open space. Translation for land investors: Sites with existing power capacity + community support just became exponentially more valuable than sites with just land and zoning. The power infrastructure thesis isn't just about finding available capacity. It's about finding that capacity in counties that actually want data centers. Not every market will roll out the welcome mat. Are you evaluating community sentiment alongside power infrastructure access?

Data centers, compute, and energy have become a bottleneck and a cash cow. Companies that once discussed software margins in earnings calls now debate cooling technologies and power procurement. The numbers tell a story of infrastructure at an inflection point: Data centers consuming 460 TWh in 2022 (pre-ChatGPT) will exceed 1,000 TWh by 2026 and the global data center market size is projected to approach $1T within 7 years. Behind every earnings mention are two core realizations: 1) Data center infrastructure is foundational to everyone’s AI aspirations Meta increased CapEx by $5B to $72B, citing "substantial internal demand for GPU resources." Microsoft warns AI demand will exceed supply through 2025. Dell raised AI server guidance to $20B. Google is acquiring stakes in crypto miners for GPUs. Hyperscalers are going nuclear with Google signing with Kairos Power for 500MW, Amazon buying Talen Energy's 960MW campus, and Microsoft partnering with Constellation. Every tech giant's earnings call now reads like infrastructure procurement because when one GPT query burns 10x the energy of a Google search, training frontier models requires city-scale power, and AI ambitions die without compute. 2) There is a SH*T TON of money to be made across the data center value chain CyrusOne raised $9.7B specifically for AI infrastructure. Blackstone paid $1B for a Pennsylvania gas plant. Traditional utilities like PPL now build generation exclusively for data centers. Power isn't infrastructure anymore – it's the business model. Cooling specialists like Submer and Green Revolution tackle 300% power increases from new chips. Edge players like Armada can deploy modular centers anywhere. AI-native infrastructure companies like VAST Data ($9.1B valuation) rebuild the stack from scratch. Nscale raised $1.1B in another play from crypto miners turned infra provider. The gold rush extends everywhere, with NVIDIA projecting that the AI infrastructure market will hit $4 trillion by 2030 and a $1T+ buildout underway – every layer of the stack is capturing value. And... with that… coming soon… the full CB Insights’ data center value chain report.

I’ve been writing for months in my weekly notes about the unintended consequences of the exponential build-out in data centres, and this week those concerns have moved firmly into the political mainstream. Bernie Sanders’ call for a moratorium on new data centres tied to the AI arms race is not an isolated comment. It reflects growing pressure from the real economy. The AI boom has come with very real costs: electricity demand has surged, power prices have jumped, grid constraints are tightening, and the average American household is increasingly footing the bill through higher utility costs at a time when budgets are already stretched. Trump was elected on a clear mandate to prioritise Main Street over Wall Street. Yet so far, the gains have largely accrued to asset prices and equity multiples, while households continue to struggle with energy bills, rents and the broader cost of living. With midterms approaching at the end of next year, that imbalance looks politically fragile. Sanders’ intervention raises some uncomfortable questions that markets appear complacent about. How socially and politically sustainable is the assumption of exponential growth in AI compute and data centre capacity? How resilient is the US equity rally if rising energy costs start to act as a de facto tax on households and small businesses? And more strategically, can the US realistically compete with China in the AI race if domestic politics begin to constrain power availability, permitting and infrastructure investment? Markets tend to extrapolate straight lines. Politics rarely does. Anyone assuming limitless AI growth, unlimited electricity and zero political backlash may need to revisit those assumptions sooner rather than later.

Markets Don’t Change All at Once. They Change Quietly, Then All at Once. One of the most underappreciated features of major market transitions is how subtle they feel while they’re happening. Capital rarely moves because of a single headline. It reallocates gradually, responding to deeper shifts in how value is created, how work is organised, and where productivity is emerging. Only later do those changes show up clearly in index leadership, sector performance, and investor outcomes. We’re seeing that dynamic again today. The contrast between the surge in data centre construction and the ongoing decline in office development is more than a real estate story. It’s a signal. Capital is flowing toward digital infrastructure and computing density, while legacy assumptions about physical space continue to be repriced. These are the kinds of shifts that reshape opportunities over years, not quarters. Periods like this are rarely comfortable for investors. They tend to be characterised by uncertainty, uneven participation, and frequent mispricing. But history suggests they are also the environments in which disciplined, pragmatic approaches matter most. Volatility is often framed as something to fear. In reality, it is frequently the mechanism through which long-term opportunities are created, particularly when markets are adjusting to structural change rather than cyclical noise. As we approach year-end, my focus is less on forecasts and more on understanding how capital allocation, leadership, and risk pricing may evolve as these transitions continue into 2026. Markets don’t announce regime shifts. They reveal them slowly, and then suddenly. If this perspective resonates, you can explore free sample issues of the Global Investment Letter and subscribe to our complimentary weekly market commentary, both available here: https://lnkd.in/g2mBz8fJ #investing #markets

A four-year-old data center running at 95% occupancy just failed a hyperscaler's technical qualification process. The engineering team had built it correctly... This is not a construction story. It is a workload compatibility story and most of the global data center installed base is on the wrong side of it. The facility was OCP-grade with a PUE of 1.15 and strong utilization. Then a hyperscaler submitted AI training specs, and the engineering team ran a retrofit analysis against three variables conventional metrics miss. 1. Power density per rack. Required: 40–50 kW. Designed: 8–12 kW. Retrofit costs reached nine figures, unsupported by lease returns. 2. Thermal architecture. Air cooling breaks at ~50 kW per rack. Liquid cooling requires full structural redesign, not an upgrade. 3. Interconnect fabric. GPU networking must be designed in from construction; retrofits rarely pencil against lease economics. The real risk in the data center market today is not the assets that look distressed. It is the assets that read strong on conventional metrics and carry AI-era language in their marketing while holding legacy-era specifications in their engineering. And it starts with a procurement document that screens out a four-year-old OCP-grade facility before the site visit is scheduled. Meanwhile, the market narrative has focused on the data center supply gap. And the retrofit economics? Quietly closing the window for most of the installed base before the analysis reaches the investment committee. That is not a management problem It is a structural disqualification the headline megawatt figure does not surface. Because the real question in data center underwriting today is not which assets are performing. It is which assets are qualified for the workload generating the lease rates that justify the valuation. The capital that sees the gap before it shows in occupancy will allocate differently from capital waiting for financials. A new operator category identified this gap before the established market did. Three former commodity traders in New Jersey identified the unmet workload and built for it. That is where this series ends. Read the article below. #datacenters

I keep seeing headlines like this one suggesting data center REITs could “triple your money” as AI demand explodes. Here’s the article for context: https://lnkd.in/grQvPbF6 I don’t disagree with the core thesis. AI is absolutely driving unprecedented demand for digital infrastructure. But I think the real story is more nuanced and more interesting than the headline. My perspective from working across regions and operators: • AI demand is real, but not all capacity is equal • Power access, grid constraints, and cooling design now matter more than square footage • Debt structures and lease terms will separate strong REITs from fragile ones • Location and workload alignment will define long-term returns • The next cycle rewards discipline, not just exposure My POV: Data center REITs can be a powerful way to participate in AI infrastructure growth but only if investors understand which infrastructure actually supports sustainable AI workloads. The easy money phase is behind us. The smart money phase is just beginning. If this adds value, feel free to share it. These conversations matter as capital continues to flow into AI infrastructure. #AIInfrastructure #DataCenters #Investing #DigitalInfrastructure

$17.3B in data center activity in just 4 days… but the real story isn’t growth. It’s constraint. In early Feb 2026, 21 data center projects moved forward across the US, from hyperscale campuses to retrofit upgrades. But beneath the headlines, a clear pattern is emerging 👇 Two markets are now running in parallel: ⚡The long clock — hyperscale campuses Billions in planning… but many stuck waiting on: 🚀Power availability 🚀Grid connections 🚀Zoning approvals Land is secured. Power is the bottleneck. 🔧The short clock — upgrades & retrofits This is where work is happening right now: 🚀UPS replacements 🚀HVAC upgrades 🚀Controls & critical infrastructure Live bids. Real projects. Immediate delivery. And the numbers make it clear: 🚀8 mega projects account for $16.4B of the $17.3B total 🚀The rest of the market is driving near-term MEP opportunity Texas tells the story best: It’s not one market, it’s two: 🚀One building GW-scale campuses for the future 🚀One delivering constant retrofit & infrastructure work today So what’s the real signal? This isn’t just a construction boom. It’s a power & infrastructure bottleneck playing out in real time. For MEP teams, the shift is clear: Success is no longer just about delivery. It’s about solving: 🚀Power strategy 🚀Phasing & energisation 🚀Cooling at higher densities 👉 Earlier in the lifecycle 👉Before construction even starts Bottom line: The data center market isn’t slowing down, it’s splitting in two.