Table of Contents
During the firm’s recent earnings call, executives repeatedly emphasized that demand for AI compute infrastructure remains strong, particularly among hyperscalers and large-scale AI operators seeking near-term access to power capacity
In sum – what to know:
Power constraints – Fermi executives said AI infrastructure deployment is increasingly limited by power availability, grid interconnection timelines and equipment availability.
Capital discipline – The company said future capital deployment will remain aligned with commercial progress and tenant agreements tied to Project Matador.
Tenant pipeline – Fermi said discussions with hyperscalers, neocloud providers and infrastructure partners continue as customers seek large-scale power capacity for AI workloads.
Fermi America executives said power availability, grid interconnection timelines, and equipment constraints are becoming central challenges for hyperscale AI infrastructure deployment, as the company focuses on tenant negotiations and disciplined capital deployment for its Project Matador campus.
During the company’s first quarter earnings call, executives repeatedly emphasized that demand for AI compute infrastructure remains strong, particularly among hyperscalers and large-scale AI operators seeking near-term access to power capacity.
“Power availability, not capital and not demand, appears to be the biggest constraint,” said Marius Haas, chairman of the board at Fermi America. “It’s clear that delays are being reported across announced projects globally, and those delays are being driven by grid interconnection timelines and equipment availability.”
Haas said the company’s strategy is designed to address those constraints through private power infrastructure and long-lead-time generation equipment.
The company also stressed a more disciplined approach to future spending as it works toward signing binding tenant agreements. Jacobo Ortiz Blanes, co-president of the Office of the CEO, said Fermi completed Phase 0 of Project Matador and would align future development activity with commercial progress.
“With this significant milestone and the conclusion of Phase 0, we pause additional site development as it was always planned until a tenant is signed,” Ortiz Blanes said. “Future capital deployment will remain disciplined and aligned with commercial progress.”
The company also said it has secured nearly $1 billion in financing commitments tied primarily to equipment financing and infrastructure development.
Executives repeatedly pointed to power infrastructure as a defining issue for the AI data center market. “The country is in a generational race for AI compute, and that race is bottlenecked by power,” Haas said. “Behind the meter, gigawatt scale, redundant private power that is delivered on the necessary timeline is not a nice-to-have for the hyperscalers and frontier model developers. It is the constraint.”
Fermi said it currently controls approximately 2.2 GW of natural gas generation equipment, including Siemens and GE turbines, and has received a 6 GW clean air permit for the project site. The company also filed for an additional 5 GW permit expansion during the quarter.
Anna Bofa, co-president of the Office of the CEO, said customer engagement around Project Matador remains active despite recent leadership changes at the company. According to Bofa, discussions with hyperscalers, neocloud providers, and enterprise compute operators are increasingly focused on power availability, delivery timelines, and commercial structures tied to large-scale AI deployments.
“Customers are not looking for a conceptual capacity, they are looking for credible near-term power, real infrastructure, secured equipment, permitting progress, land control, and a team that can execute,” Bofa said.
Executives also said the company is evaluating partnerships with established data center operators and infrastructure providers as part of its commercialization strategy.
“We are also evaluating strategic partnerships with established and respected data center operators and infrastructure partners,” Bofa said. “We view those partnerships as potential accelerators, a way to expand our execution capacity, increase customer confidence, and serve a broader set of tenants while maintaining commercial discipline.”
The company said current customer discussions span multiple potential deployment sizes. Bofa said some opportunities involve “a couple hundred megawatts,” while others are “1 G or more.”
The broader AI infrastructure market is also driving interest in alternative energy architectures beyond traditional grid connections. Industry forecasts suggest U.S. data centers could account for 9% to 17% of national electricity consumption by 2030, increasing pressure on utilities and power infrastructure. As developers seek faster and more flexible deployment models, technologies such as Battery Energy Storage Systems (BESS), microgrids and behind-the-meter generation are gaining attention as potential complements to large-scale AI campuses. The trend also reflects growing concerns around the long-term use of diesel backup systems because of fuel costs, emissions, noise and community opposition.
