20 Deals, $11.5 Billion: AI's Hunger for Power Is Rewriting the Energy Playbook

In a single week, more than $11.5 billion in capital moved into AI-related energy and data center infrastructure across 20 separate deals. That figure alone would have been headline-worthy a year ago. Today, it barely keeps pace with demand.

The numbers tell a clear story. Every major AI company — Meta, Mistral, the US military — is racing to secure power and compute capacity. But the more interesting pattern sits beneath the surface: the deals are no longer just about building bigger boxes. They're about rethinking how energy reaches those boxes in the first place.

The Mega-Deals: Where Billions Are Landing

Digital Realty's $3.25 billion hyperscale data center fund was the week's largest single commitment, signaling that institutional appetite for data center exposure remains strong despite rising construction costs. The fund targets the hyperscale segment — facilities purpose-built for the handful of companies whose AI workloads demand warehouse-scale compute.

Two $2 billion deals followed close behind, each revealing a different dimension of the trend. Meta committed $2 billion to fund energy infrastructure for its Louisiana AI data center, a deal notable because it's explicitly about power, not servers. Meanwhile, the US Army tapped Carlyle and KKR to develop $2 billion in AI-focused data centers, marking one of the clearest signals yet that sovereign compute capacity has become a national security priority.

India's ambitions showed up through Airtel's Nxtra, which is raising $1 billion to scale data center capacity to 1 GW. And Adani Group is reportedly in discussions with Google and Meta to set up additional data center facilities in the country — though that deal remains in early talks.

AI Companies Are Now Energy Companies

Perhaps the most telling deal this week: Mistral raised $830 million in debt specifically for data center build-up. A French AI lab, founded barely two years ago, is now borrowing nearly a billion dollars — not to train models, but to house the hardware that trains them. The line between AI company and infrastructure company is dissolving.

Data centers captured 86% of the week's capital, with $10 billion flowing directly into facilities and supporting infrastructure. Grid and power deals added another $471 million, renewables contributed $600 million (driven entirely by Zelestra's $600 million Texas solar portfolio backed by Meta), and energy-as-a-service platforms pulled in $550 million through Budderfly's debt facility expansion.

The concentration in data centers is expected. What stands out is the growing activity at the edges — the deals working to solve power delivery, grid management, and energy generation for these facilities.

The Grid Problem Nobody Can Ignore

Every new data center needs power. A lot of it. And the grid wasn't built for this.

ThinkLabs raised $28 million in its Series A with a pitch that would have seemed niche three years ago: helping power grids manage data center demand. Octopus Energy acquired a majority stake in Uplight to scale digitalized grid flexibility across North America. Emerald AI raised $25 million on the premise that data centers themselves can become flexible grid assets — drawing power when it's cheap and abundant, scaling back when the grid is strained.

These aren't massive deals by this week's standards. But they point to a structural shift: the AI industry is creating demand for an entirely new category of grid management technology.

Primoris Services' $422 million acquisition of PayneCrest Electric, an electrical contractor, fits the same pattern. When the companies that build power infrastructure start getting acquired at premium multiples, the market is pricing in sustained demand.

Debt Is Doing the Heavy Lifting

Equity raises grabbed the largest share of capital at $5.2 billion across nine deals. But debt financing punched above its weight — four deals totaling $2.1 billion, including Mistral's $830 million, Zelestra's $600 million solar financing, Budderfly's $550 million facility, and EPG's $100 million in Series B+ financing for global AI data center growth.

The willingness of lenders to extend large credit facilities to data center and energy infrastructure projects reflects a bet on predictable, long-duration revenue streams. When Meta signs a power purchase agreement, the counterparty's cash flows become bankable. That's drawing in a class of capital — infrastructure debt, project finance — that doesn't typically chase tech companies.

The Frontier: Space Data Centers and Geothermal Drilling

At the other end of the spectrum from Digital Realty's billions, a handful of startups are proposing radical alternatives to the "build a bigger building" approach.

Starcloud raised $170 million at a $1.1 billion valuation for space-based data centers. The thesis: orbit offers unlimited cooling and access to uninterrupted solar power. It's speculative, but the valuation suggests investors are taking it seriously.

Telura emerged from stealth with €4 million to develop geothermal energy through electric impulse drilling. And ODC closed a $45 million Series A to build a global distributed compute grid, betting that spreading workloads across smaller facilities closer to renewable energy sources is more efficient than concentrating them in mega-campuses.

Foresight Energy Infrastructure Partners invested in Mirai Power, a German battery storage developer, while Amazon acquired 1,300 acres near the Columbia River in Washington state — a location chosen, likely, for its proximity to cheap hydroelectric power.

What This Tells Us About Q2

Three patterns from this week's data are worth watching.

First, the geographic spread is accelerating. India ($1 billion from Nxtra alone plus Adani's exploratory talks), France (Mistral's $830 million), Germany (Foresight's battery storage bet), and Nebius building one of Europe's largest data centers — the AI infrastructure buildout is no longer a US-only story.

Second, the capital stack is diversifying. Equity, debt, government partnerships, and project finance are all active. That breadth of capital sources suggests the market views AI infrastructure as a durable asset class, not a speculative bet.

Third — and this is the one to watch most closely — the ancillary deals are growing. Grid management, battery storage, solar portfolios, geothermal startups. For every dollar going into a data center, an increasing fraction is going into the systems that keep it running. If the first phase of AI infrastructure was about real estate and servers, the second phase is about electrons.

The AI industry's power problem is real, it's getting bigger, and the market has decided it's investable. Twenty deals and $11.5 billion in one week says the capital agrees.