TensorWave signs six-year deal to supply 457 MW of AMD-powered capacity for OpenAI
On November 16, 2026, TensorWave, a Las Vegas cloud provider that runs only AMD hardware, said it will supply OpenAI with 457 megawatts of IT capacity across several U.S. data center sites under a six-year agreement. According to Pulse 2.0, the capacity will operate as dedicated, single-tenant infrastructure, so the racks serve OpenAI alone. The compute uses AMD Instinct MI455X GPUs inside AMD Helios rack-scale systems. The 457 MW figure measures contracted IT power capacity, not electricity that operating chips already draw.
Deliveries are expected to begin in the first half of 2027, with full deployment targeted for early 2028. Neither company has disclosed the contract value or named the data center sites. The agreement sits within the AMD and OpenAI partnership announced in October 2025, which covers up to 6 gigawatts of AMD GPUs. TensorWave is the first neocloud publicly named under that framework, and it describes the award as its largest deployment to date.
TensorWave was among the first public clouds to deploy AMD's MI300X accelerators and has since added MI325X and MI355X chips. Its June 2026 Series B raised $350 million at a $1.55 billion valuation, co-led by Magnetar and AMD Ventures. Nexus Venture Partners also participated in that round and is among the company's backers.
Why it matters
For founders, the deal signals that frontier labs will commit to large, dedicated blocks of non-Nvidia compute and that specialist clouds can deliver them. A buyer of OpenAI's scale puts AMD's rack-level systems through production demands at a size few providers have reached, which may make other AI companies more willing to evaluate AMD capacity. Operators that can secure power, grid access and long-term leases now have a clearer route to enterprise-scale contracts.
For investors, the award supports a model in which contracted megawatts and multi-year customer commitments anchor valuation, not GPU counts alone. Because the price is undisclosed, outsiders cannot yet tell what revenue the 457 MW will produce. Teams renting GPUs should watch whether AMD clusters become available in larger blocks, and should compare contract length, power commitments and delivery schedules before committing.
Key points
- 1TensorWave will provide 457 MW of IT capacity to OpenAI across multiple U.S. sites for six years.
- 2The capacity uses AMD Instinct MI455X GPUs in AMD Helios rack-scale systems, run single-tenant for OpenAI.
- 3Deliveries are expected to begin in the first half of 2027, with full deployment targeted for early 2028.
- 4Neither the contract value nor the data center sites have been publicly disclosed.
- 5