
OpenAI has signed a six-year agreement for 457 megawatts of AMD-powered computing capacity from US cloud provider TensorWave. The deal, announced on 16 November, adds another substantial source of infrastructure for the ChatGPT developer as competition for electricity, data-centre space and AI processors intensifies.
According to TensorWave’s announcement, the dedicated capacity will be spread across multiple US data-centre sites and built around AMD Instinct MI455X graphics processors in Helios rack-scale systems. The companies did not disclose the contract value, the timetable for bringing each site online or the amount of capacity already available.
A megawatt figure describes a power allocation, not a count of operational GPUs or an immediate increase in ChatGPT performance. Delivering it requires secured sites, grid connections, cooling equipment, networking and installed processors. The six-year term signals a long-running supply arrangement, while the absence of a deployment schedule means readers should not treat all 457 megawatts as live today.
For OpenAI, the arrangement diversifies both its cloud suppliers and its processor supply. The company has relied heavily on Nvidia-powered infrastructure across the AI industry, but this deal places AMD hardware at the centre of a significant new deployment. TensorWave specialises in AMD-based AI computing, giving the chipmaker another route into large-scale training and inference workloads.
The announcement also sits within OpenAI and AMD’s wider infrastructure relationship. Capacity contracted through a specialist cloud provider can complement direct chip agreements and other data-centre projects. It does not mean OpenAI is replacing all of its existing suppliers. Large AI developers increasingly assemble computing portfolios across several partners because no single project can reliably meet every requirement or opening date.
The difficult work now is turning a contract into usable compute. Power has to arrive at the right sites, Helios systems must be installed and supported, and software workloads need to run efficiently on AMD’s platform. Those steps will determine the commercial value of the agreement more than the headline capacity alone.
The scale is nevertheless notable. A 457-megawatt commitment illustrates how the AI industry’s demand is increasingly measured in electricity as much as in processor shipments. It also raises the familiar questions of financing, grid capacity and local costs. TechBooky recently examined OpenAI’s pledge to cover costs associated with its own data-centre developments. The TensorWave contract is a separate supplier arrangement, and its site-level energy terms have not been published.
Investors and customers will be watching for named locations, delivery milestones and evidence that the promised AMD systems can operate at the required scale. Until those details emerge, the agreement is a major capacity commitment, not proof that 457 megawatts of new AI computing are already in service.







