Data Centers Became the Hottest Asset in April: $70B+ Bet on AI Compute Power

Forty-two major data center deals closed in April, totaling over $70 billion. Not all at once—staggered through the month, but with relentless momentum. The sheer scale redefines what we mean by infrastructure investment.

Five years ago, a $1 billion data center deal was exceptional. Today, it barely breaks the top ten.

The Numbers That Matter

Microsoft alone contracted for 30,000 high-end GPUs across multiple facilities. That's not a purchase order—it's a declaration of intent to dominate AI compute. Google, not to be outdone, signed up for capacity at European data centers and fast-tracked its first French facility. Amazon's AWS committed to additional Mississippi facilities and a $430 million play in India.

Then there's Blackstone. Through its QTS subsidiary, the mega-fund is raising $4.6 billion—a staggering amount for infrastructure that, just three years ago, would have been funded by regional utilities or smaller REITs.

This is not a market correction or a cyclical spike. This is structural. Hyperscale data centers—the kind needed to train and deploy the largest AI models—require different power configurations, cooling systems, and physical scale than traditional colocation facilities.

Geography Tells the Story of Competition

The United States leads with 33 major data center announcements in April, but the distribution is revealing. AWS is building aggressively in Virginia and Mississippi. Microsoft is buying land across multiple states—3,200 acres in one transaction alone. Google is pushing into Europe, explicitly stating plans for French facilities despite regulatory headwinds.

Asia is not being ignored. TikTok announced €1 billion for its second European facility. Digital Realty committed S$7 billion to Singapore. ESR, a major Asian data center operator, secured $850 million in additional equity to expand regional capacity.

The geographic play reveals the real bet: no single region will dominate AI infrastructure. The winners will be those with globally distributed footprints and local power relationships. Microsoft's strategy—land in the US, capacity in Europe, hyperscale in Asia—is becoming the template.

Why Data Centers Matter More Than Models

There's a narrative that AI is about algorithms. That's backwards. The algorithm is a commodity. What's scarce is the infrastructure to run it at scale.

OpenAI's $14 billion in facility investments. Amazon's Mississippi expansion. Google's European push. These are not financial engineering—they're capital allocation decisions that will define which companies control the bottleneck: electricity, land, and cooling.

A transformer model trained on 1 trillion tokens requires specific hardware, specific power delivery, and specific thermal management. Build the wrong facility, and you're competitive by 18 months. Build the right one, and you own a moat.

The Financing Question

What's notable is HOW these deals are being financed. Switch raised $768 million in asset-backed securities—a structure typically used for mature, predictable assets. Blackstone is using traditional institutional capital. Digital Realty is tapping sovereign wealth funds.

This is critical: institutional investors are treating data center capacity as equivalent to telecommunications infrastructure or toll roads. That's a shift. Five years ago, data center investing was venture-backed and speculative. Now it's boring, essential infrastructure being funded like airports and power plants.

What Happens Next

The April deals represent capacity that will be live by 2027-2028. That's when we'll see whether current estimates of AI compute demand hold up. If they do, expect another wave of capital commitments in H2 2026. If demand softens, you'll see a reversal—infrastructure built on unsustified assumptions.

But here's what's not going to reverse: the concentration of data center ownership. The winners—Microsoft, Amazon, Google, Blackstone—are now so far ahead in terms of capacity and financing relationships that new entrants will struggle to compete. This is winner-take-most infrastructure.

The AI boom, at least for now, is fundamentally a real estate story.