TensorWave is a U.S.-based AI infrastructure startup founded in 2023 and headquartered in Las Vegas. The company specializes in building high-density, performance-optimized cloud platforms using AMD Instinct GPUs. These platforms offer an alternative to NVIDIA-based systems and are designed for large-scale AI training and inference workloads.
At the core of TensorWave’s offering is a bare-metal and managed infrastructure built around AMD’s MI300X, MI325X, and MI355X accelerators. These GPUs are optimized for high-throughput, memory-intensive tasks and are supported by direct liquid cooling to enhance energy efficiency. The infrastructure is built to support large language model training and inference with exceptional compute performance and high memory bandwidth.
In 2025, TensorWave deployed the largest publicly available AMD GPU cluster in North America, featuring 8,192 MI325X units. This installation delivers over 21 exaFLOPS of FP8 compute power and more than 2 petabytes per second of memory bandwidth. It was specifically engineered with direct-to-chip liquid cooling to maintain stability and efficiency.
TensorWave has also signed a one-gigawatt infrastructure agreement with a major data center partner to rapidly scale its operations. The partnership includes phased deployments using microgrid and power generation systems to support massive GPU clusters across the U.S.
In May 2025, the company raised $100 million in a Series A funding round led by key industry investors. The capital will be used to expand its operations, deploy additional GPU clusters, and enhance its software stack.
TensorWave provides managed inference services for models like Llama 3, with flat-rate or usage-based pricing, auto-scaling, and high availability. The platform is compliant with SOC 2 Type II and HIPAA standards, making it suitable for sectors such as healthcare, finance, and logistics. TensorWave’s focus is on delivering scalable, cost-effective AI compute infrastructure with AMD hardware at its core.
