A Texas-based real estate investment trust specializing in AI infrastructure has secured its first major tenant by leveraging its unique ability to secure land, power, and permits early in the process. On August 10, Fermi Inc. (FRMI.US), which is focused on developing hyperscale AI data centers, announced that its Project Matador campus in Carson County, Texas, has signed its first binding customer lease with AI cloud provider TensorWave. Following this news, Fermi shares surged about 20% in Monday's after-hours trading.
The lease involves a total contract value of approximately $6.5 billion over a 15-year initial term, covering a facility with 222 megawatts of total power capacity. The agreement, signed between Fermi subsidiary Fermi Campus 1 LLC and TensorWave subsidiary TensorWave TEX1, LLC, includes options to expand with two additional data centers, potentially extending the partnership to over 650 megawatts across three phases. The facility is expected to begin phased delivery in the second half of 2027 and, once fully operational, could support tens of thousands of next-generation AMD Instinct GPUs for large-scale AI training and inference. The lease also includes two five-year renewal options, potentially extending the term to 25 years.
Fermi Chairman Marius Haas stated: "A lease of this size and duration is a massive vote of confidence in Fermi. In delivering this first binding agreement, we have achieved the most important goal we promised to the market in May."
TensorWave, the tenant in this deal, is a Las Vegas-based AI cloud provider that operates one of the largest all-AMD GPU clouds globally. The company completed a $350 million Series B funding round in June 2026, led by AMD Ventures and Magnetar Capital, at a valuation of $1.55 billion. Currently operating three data centers in Pennsylvania, Arizona, and Florida, TensorWave has secured over 2 gigawatts of long-term data center capacity and has signed leases totaling 500 megawatts. CEO Darrick Horton noted: "Our customers collectively require hundreds of thousands of AMD Instinct GPUs and need reliable timelines. Fermi has integrated power, land, and permits to deliver at this pace."
This transaction highlights the core logic of the current AI infrastructure market: securing power is paramount. While hyperscale companies and cloud providers compete for large-scale, reliable power supplies, many projects stall due to permit delays or lack of financing. Fermi has gained a competitive edge by locking in land, permits, and power resources early. Project Matador, located in the Texas Panhandle on a site larger than half of Manhattan, is designed to receive power behind the meter via a private grid, with planned capacity gradually increasing to about 17 gigawatts. Currently, approximately 6 gigawatts are permitted, with over $1.5 billion already invested in construction, targeting first power delivery in 2026. Compared to similar deals, Fermi's scale is particularly notable. For context, Bitcoin miner Riot Platforms, which is transitioning to AI hosting, previously signed a 20-year lease for 191 megawatts of AI data center capacity, which is about 14% smaller than Fermi's initial phase.
This deal also strengthens AMD's competitive position in AI infrastructure. TensorWave's commitment to deploying tens of thousands of AMD Instinct GPUs at the facility indicates AMD is challenging Nvidia's dominance in AI accelerators by fostering an AI cloud ecosystem. Notably, Fermi expects that part of the obligations under the lease will be guaranteed by a global AI leader, though the guarantor's identity has not been disclosed. Fermi stated that the Texas Tech University System, which is collaborating on Project Matador, supports the transaction and will work to ensure a smooth closing. The company is also advancing discussions with other hyperscalers, new cloud providers, and enterprise computing counterparties for the remaining capacity at Project Matador.
Comments