Fermi Locks $6.5B Binding Lease With TensorWave: 222MW AMD GPU Campus Opens in Texas
Fermi Inc. (NASDAQ: FRMI), operating under the Fermi America brand, announced on August 10 that it has executed a binding customer lease at its Project Matador campus in Carson County, Texas. The tenant is TensorWave TEX1, a subsidiary of AI cloud provider TensorWave Inc. The lease is Fermi’s first binding commercial commitment.
The Deal
Phase 1 delivers a turnkey 222MW data centre beginning in the second half of 2027. Revenue over the initial 15-year term: approximately $6.5 billion. The facility is designed to support tens of thousands of next-generation AMD Instinct GPUs for training and inference at scale. TensorWave has secured expansion rights for two additional data centres; if exercised, the total partnership would reach over 650MW across all three phases.
AMD’s involvement is notable. The Instinct line has accelerated from a curiosity to a credible alternative to NVIDIA’s H100 and B200 series in high-volume AI training clusters. TensorWave is betting that AMD’s cost-per-flop advantage at scale will hold through the campus lifetime.
Who Is Fermi
Fermi Inc. is a publicly traded AI data centre developer dual-listed on Nasdaq and the London Stock Exchange. Project Matador in Carson County, Texas, is its flagship campus. The company has not previously disclosed binding leases, making the TensorWave agreement its first public proof of commercial demand.
The dual listing and the Texas location follow a pattern among AI infrastructure developers targeting both US hyperscaler demand and European institutional capital. The LSE listing gives Fermi access to UK and European pension and sovereign wealth capital that is increasingly allocated to AI infrastructure as an infrastructure asset class.
TensorWave’s Position
TensorWave is an AI cloud provider focused on AMD-based inference infrastructure. The company has positioned itself as the primary non-NVIDIA alternative for enterprises running large-scale model training and inference. A 222MW single-campus commitment — before delivery — signals TensorWave’s confidence that enterprise demand for AMD GPU capacity will persist at the scale required to fill a 15-year lease.
The $6.5B figure is phase 1 revenue only, not the full contracted value if expansion options are exercised. A 650MW endpoint would put the total contract value well above $15B.
Context
The AI infrastructure lease market has established a clear pattern in 2026: operators announce capacity, hyperscalers and AI cloud providers sign long-duration leases before shovels go in the ground, and the lease revenue becomes the collateral that finances construction. CoreWeave, Hut 8, CleanSpark, and Bitdeer have all executed multi-billion-dollar leases under this structure. Fermi’s Matador deal slots into that pattern with one addition: an AMD-first, non-NVIDIA chip stack — a deliberate differentiation as demand for Blackwell clusters pushes NVIDIA lead times and pricing to levels that make AMD’s MI450X increasingly competitive on total cost of ownership.