Cardinal Data Power raises $70M Series A to build gigawatt-scale AI data centres
What's the deal? Cardinal Data Power (CDP), a private developer of powered data centre campuses, has raised roughly $70 million in a Series A round. The financing was led by Hood River Capital ManagementDealroom has a profile for this one. Try Dealroom →, with a $20 million preferred equity investment from Empery DigitalDealroom has a profile for this one. Try Dealroom → (NASDAQ: EMPD) for an approximately 8% stake.
What's the endgame? CDP builds campuses for AI and high-performance computing tenants, focusing on behind-the-meter projects that pair secured power generation with large land positions. The round funds its inaugural West Texas campus, where it has signed a letter of intent for a 750 MW Phase I development.
Why now? The West Texas site spans more than 3,500 acres and is designed to deliver first power in 2027. It will ramp to roughly 1 GW of gross capacity by 2029, with expansion beyond 5 GW planned over time.
CDP was founded by the Lyda Hunt Allred and Barbara Crow Hunt families and their family offices, Cavallo Holdings and Stratford Bridge Holdings. The Dallas-based developer is an affiliate of Hunt Properties, and its principals bring decades of energy infrastructure experience. It has secured reciprocating engine generation capacity through an agreement with an independent power producer.
Why it matters: "Cardinal Data Power is uniquely positioned to bring generation, transmission and management of power solutions to the insatiable energy demands of the AI Infrastructure build-out," said Ryan Lane, co-chief executive officer of Empery Digital. The West Texas project is the first in a pipeline of large-scale developments.
The signal: The round reflects how the AI boom is turning power access into the binding constraint for data centre growth. Developers that can secure generation and land at scale — rather than just build racks — are drawing capital from both financial and strategic backers.
Read more: Wall Street Online
