Amazon reportedly revives deal with Rick Perry-backed data center venture Fermi America

The negotiations center on Project Matador, a sprawling campus that aims to generate up to 17 GW of combined on-site power. This reunion follows a collapsed tenant deal that caused Fermi's stock to drop sharply and exposed significant governance issues within the company.

New Power Agreement

On November 12, 2026, Fermi announced a new power capacity agreement covering 600 MW with an undisclosed investment-grade hyperscaler, widely believed to be Amazon. The agreement includes a $200 million upfront prepayment, which represents a significant vote of confidence from a major customer.

The deal also includes substantial commitments tied to early construction and the purchase of materials. This new prepayment is larger than the $150 million initial tenant deal that collapsed in December 2025.

Project Matador Details

Project Matador sits on land leased from the Texas Tech University system. The plan calls for a blend of nuclear, natural gas, solar, and wind generation to power the site directly. Fermi went public in October 2025 at a valuation of roughly $15 billion, but had no operational revenue at the time.

The 600 MW figure represents only a small fraction of the 17 GW Fermi says Project Matador could eventually produce. The company has faced major internal governance issues, which led to halted construction and further pressure on the share price.

Implications and Next Steps

For Fermi shareholders, a finalized agreement with a hyperscaler would provide essential funding and a credible operating path, two things the company has lacked since the December 2025 collapse. The $200 million prepayment matters more than the headline capacity number as it represents upfront cash from an investment-grade customer.

However, the risks remain significant. The counterparty on the 600 MW agreement has not been officially named, and the broader Amazon talks are reported rather than confirmed. Execution is the bigger challenge as Fermi must prove it can build generation capacity across multiple energy sources on a timeline acceptable to a hyperscaler.