TensorWave (tensorwave.com), a Las Vegas-based provider of AMD-exclusive AI and HPC cloud infrastructure, has raised $350 million in Series B funding at a $1.55 billion valuation. The platform runs large-model training and inference on AMD Instinct GPUs using the ROCm stack, offering enterprises an open alternative to NVIDIA lock-in. The capital will expand global AMD clusters and data center capacity.
AMD Clouds Capitalize on NVIDIA Constraints
The timing aligns with hyperscalers planning over $600 billion in AI infrastructure spending during 2026. Oracle Cloud announced a deployment of 50,000 AMD AI chips in October 2025 as an NVIDIA alternative. TensorWave's pure-play AMD approach targets customers seeking to avoid single-vendor dependency.
Supply Shortages Limit AI Workload Scale
Enterprises face constrained access to high-performance GPUs amid explosive generative AI demand. The AI infrastructure market stands at $75.4 billion in 2026 and is projected to reach $498 billion by 2034, growing at 26.6% CAGR. Power availability and ecosystem lock-in compound the challenges for teams running frontier models.
Pure-Play AMD Stack Avoids CUDA Lock-In
TensorWave operates large-scale clusters with MI325X and MI355X GPUs and supports unmodified CUDA workloads through partners like Spectral Compute. This delivers performance gains without requiring code changes. The company emphasizes 24/7 support and enterprise compliance as differentiators from broader cloud providers.
As CEO Darrick Horton noted:
"CUDA is not the only path forward, and the conversation is finally catching up."
Magnetar and AMD Ventures Co-Lead Round
Magnetar Capital and AMD Ventures led the round, joined by Nexus Venture Partners and Maverick Silicon. The investor mix combines growth capital, hardware ecosystem validation, and AI-native operational expertise, reinforcing the thesis around diversified AI compute.
Record Capex Accelerates Specialized Providers
Hyperscaler AI spending is expected to exceed $600 billion in 2026, creating tailwinds for dedicated infrastructure players. While competitors focus primarily on NVIDIA GPUs, TensorWave's exclusive AMD positioning and open ROCm ecosystem address demand for alternatives. Recent Oracle AMD deployments further signal mainstream acceptance.
Ex-Skunk Works Founder Drives Ecosystem Play
CEO Darrick Horton, a Forbes 30 Under 30 honoree with a background in plasma physics at Lockheed Martin Skunk Works, has positioned TensorWave as a leading AMD advocate. The team emphasizes hands-on engineering and partnerships to build production-ready open infrastructure.
10x Capacity Expansion Planned
With 8,192 MI325X GPUs live and customers including Fireworks AI and Luma AI, TensorWave intends to scale capacity by 10x while adding MI355X deployments.