
A Las Vegas company most people outside the tech world have never heard of just locked in the biggest deal of its young life. TensorWave, a neocloud firm based in the city, has agreed to supply OpenAI with 457 megawatts of AI infrastructure capacity, enough electricity to power around 375,000 homes, under a six-year agreement built around hundreds of thousands of AMD Instinct MI455X GPUs.
The agreement, first reported by Data Center Dynamics and later covered by the Las Vegas Review-Journal, caps roughly two years of development between the two companies, according to the newspaper. The capacity will run as dedicated, single-tenant infrastructure built exclusively for OpenAI, according to Data Center Dynamics. Operations are expected to begin in the first half of 2027, with full deployment expected by early 2028, per the same outlet. Neither company has disclosed the contract's dollar value or named the specific data-center sites involved.
TensorWave's Biggest Bet Yet
Data Center Dynamics calls the agreement TensorWave's largest deployment contract to date, and the site also reports that TensorWave is the first provider publicly named under OpenAI's existing partnership with AMD. That broader AMD-OpenAI arrangement, announced in October 2025, calls for deploying 6 gigawatts of AMD GPUs. Piotr Tomasik, TensorWave's co-founder, president and chief operating officer, said the company's slice represents around 8 percent of that larger AMD-OpenAI deal, according to the Review-Journal.
Tomasik framed the deal as more than a business win for his company, telling the Review-Journal it marks a step toward making Las Vegas a tech destination for workers and business. TensorWave operates as what's known in the industry as a neocloud company — it acquires GPUs, deploys them into data centers, interconnects them and maintains them, then offers that computing power to other companies for large-scale processing. Clients pay based on how often and how much they use the GPUs, the newspaper reports.
What's Actually Inside the Machines
At the center of the deal are GPUs — chips built to perform many tasks simultaneously and chew through large amounts of data, which is why they're essential for training AI models and are also used for tasks like generating images and video playback. A single GPU can cost thousands of dollars, underscoring the scale of what TensorWave is committing to deploy.
The hardware itself comes from AMD, which is trying to chip away at Nvidia's dominance in AI infrastructure. AMD's Helios rack-scale systems, which will house the MI455X chips in this deployment, are pitched by the chipmaker as outperforming Nvidia's Vera Rubin NVL72 rack by 15 percent in peak FP4 performance and by 50 percent in high-bandwidth memory capacity, according to Data Center Dynamics. AMD also claims Helios can deliver up to 30 percent more tokens per dollar than Nvidia's comparable system, the outlet reports.
A Fast-Growing Vegas Operator
TensorWave's growth has been rapid even before this contract. The company plans to grow its capacity from 35 megawatts to over 450 megawatts and expand its workforce from 225 employees to more than 1,000 by the end of 2027, per the Review-Journal's reporting. The company secured $350 million in a Series B funding round in June, with the capital set to support expansion of its AI infrastructure footprint, according to Data Center Dynamics. TensorWave also signed a separate 20-megawatt capacity agreement with TecFusions in January for data centers in Pennsylvania and Arizona, the same outlet reports.
OpenAI's own leadership framed the deal in terms of necessity rather than novelty. The company's chief technology officer for compute said that as demand for AI grows, expanding access to reliable, large-scale compute is critical to developing and delivering more capable models, according to Data Center Dynamics. TNW likewise notes that this is the first time TensorWave has been publicly named as a provider under OpenAI's AMD partnership, a detail that signals how closely the industry is watching which companies get picked to build out the AI boom's physical infrastructure.
A state board-packet document filed with the Nevada Governor's Office of Economic Development shows TensorWave applying for the state's Sales & Use Tax Abatement and lists a 2 percent Sales Tax Abatement for two years, a 50 percent Modified Business Tax Abatement for four years, and a 50 percent Personal Property Tax Abatement for ten years. Still unresolved publicly: the financial value of the OpenAI contract and the exact locations where the new capacity will be built.









