TensorWave announced a signed, six-year agreement on November 16, 2026, to provide OpenAI with 457 MW of dedicated AI infrastructure across multiple U.S. data-center sites. The company expects the first capacity in the first half of 2027 and the full deployment by early 2028. TensorWave’s announcement dates the deal; its company release describes the planned systems and workloads.
What TensorWave agreed to provide
The agreement covers infrastructure dedicated to a single customer: OpenAI. TensorWave says the planned capacity will support large-scale AI training and inference, the processes used to train AI models and generate their outputs.
The plan spans multiple U.S. data-center sites. TensorWave says the systems will use AMD Instinct MI455X GPUs in AMD Helios rack-scale systems, connected by high-performance Ethernet. These are the components and workloads specified for the planned build.
When the capacity is expected
TensorWave expects to bring the first capacity online in the first half of 2027 and complete the deployment by early 2028. The schedule describes a staged rollout, with initial capacity preceding the full build.
OpenAI announced a broader partnership with AMD on October 6, 2025, covering 6 GW of GPU capacity across multiple generations. That agreement provides context for the companies’ wider compute relationship; TensorWave’s announcement sets out its own 457 MW project. OpenAI’s partnership announcement describes the broader commitment.
TensorWave’s earlier AMD strategy
In a July 2026 interview, TensorWave co-founder and CEO Darrick Horton described the company’s earlier decision to focus on AMD:
“We saw an incredible bottleneck and scarcity in the market back in 2023, and AMD presented a credible general-purpose path forward.” — Darrick Horton, TensorWave co-founder and CEO, in a July 2026 interview
The interview predates the OpenAI agreement and offers Horton’s perspective on TensorWave’s pre-deal strategy.