What TensorWave does

TensorWave is an AMD-exclusive AI cloud infrastructure provider founded in December 2023 and headquartered in Las Vegas, Nevada. The company delivers GPU compute services exclusively on AMD Instinct accelerators (MI300X, MI325X, MI355X, and roadmap MI455X with HBM4) running the ROCm software stack, paired with direct liquid cooling, RoCEv2 high-speed networking, UEC-Ready networking, and software-defined parallel file system storage. Its portfolio includes bare-metal GPU servers, dedicated clusters, managed Slurm and managed Kubernetes orchestration, high-speed network storage, and an Enterprise Suite that layers ScalarLM, its open GPU-agnostic orchestration software, for unified LLM training, fine-tuning, and inference.

The platform serves AI labs and research institutions running distributed multi-node training, enterprises deploying production inference and agentic AI workloads, and vertical specialists such as telecommunications operators through the GSMA Open Telco AI initiative. Named customers include Meta, Moreh, Modular, Luma, and Higgsfield AI. TensorWave sells on a usage-based, pay-as-you-go model with pricing starting at $1.71/GPU HR for MI300X, $2.25/GPU HR for MI325X, and $2.95/GPU HR for MI355X, supplemented by managed-service subscriptions and enterprise security and support tiers.

The company has raised approximately $493 million in cumulative funding (a $43M SAFE in October 2024, a $100M Series A in May 2025, and a $350M Series B at a $1.55B valuation in June 2026) led or co-led by Magnetar and AMD Ventures, with participation from Nexus Venture Partners, Maverick Capital, Maverick Silicon, Translink Capital, Javelin Venture Partners, Granite Partners, and Western Frontier. It operates three data centers in Arizona, Florida, and Pennsylvania with combined capacity of 30,000 AMD Instinct processors, is expanding to 2 gigawatts of capacity from approximately 500 megawatts, and reached $100M+ in annual recurring revenue within 16 months of launch.

Name

TensorWave

Legal name

TensorWave Inc.

Website

https://tensorwave.com

Company type

Private

Founded year

2023

Operating status

Operating

Headcount range

101–250 employees

Short description

TensorWave is an AMD-exclusive AI cloud platform that provides GPU compute, managed orchestration, and storage services on AMD Instinct accelerators for AI labs, enterprises, and research institutions.

Ownership category

akta.pro rank

Product category

AI Cloud Infrastructure

NAICS

Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518), Computer Systems Design and Related Services (54151)

SIC

Services-Computer Integrated Systems Design (7373), Services-Computer Processing & Data Preparation (7374)

akta.pro primary industry

GPU-Accelerated & AI Training/Inference Servers (HDACABAG)

akta.pro secondary industries

AI Server Systems & HGX/Accelerator Platforms (HDAAAAAB), Model Deployment, Serving & Inference Platforms (HDAAABAF), AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers) (HDAAAAAG), AI Datacenter Power, Cooling & Racks (liquid cooling, power delivery) (HDAAAAAH)

Keywords

• GPU cloud services
• AI compute infrastructure
• AMD GPU clusters
• AI training infrastructure
• Cloud GPU hosting

Headquarters

HQ city

Las Vegas

HQ country

United States

HQ region

North America

Offices4 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Infrastructure, Technology or R&D, Personnel, Operations, Marketing or Sales

Revenue model

1. GPU Compute Infrastructure: Cloud-based GPU compute services billed on a per-hour/per-GPU basis for AI training, fine-tuning, and inference workloads. Customers can purchase bare metal servers, clusters, or container-based access with flexible deployment options.

Pricing tiers

Go-to-market motion1 record

Distribution channels3 records

Marketing channels6 records

Core offering

TensorWave operates an AMD-exclusive AI cloud platform providing GPU compute infrastructure for AI training, fine-tuning, and inference workloads. The platform is built on AMD Instinct accelerators (MI300X, MI325X, MI355X, MI455X) paired with ROCm software, direct liquid cooling, high-speed RoCEv2 networking, managed Slurm/Kubernetes orchestration, and the ScalarLM software layer.

Product overview

TensorWave is an AMD-exclusive AI cloud platform providing infrastructure-as-a-service for AI training, fine-tuning, and inference at scale. The portfolio is organized around Bare Metal GPU infrastructure as the core offering, supported by AMD Instinct accelerator products (MI300X, MI325X, MI355X, MI455X), managed orchestration services (Managed Kubernetes, Managed Slurm), and High-Speed Network Storage. The Enterprise Suite layers ScalarLM software on top of the hardware foundation to provide unified LLM lifecycle management. Additional products include AI Inference, AI Model Fine-Tuning, AI Model Training solutions, Observability for fleet monitoring, Expert Support services, and a comprehensive Security Suite covering encryption, compliance, and access governance. The platform emphasizes open architecture without vendor lock-in, leveraging ROCm ecosystem compatibility and direct liquid cooling for sustained performance.

Differentiator

Problem solved

Functional benefit

Products and services

• Bare Metal GPU Infrastructure Full-machine GPU ownership providing predictable performance, consistent networking, and direct hardware access for AI training, fine-tuning, and inference workloads on AMD Instinct accelerators.
• AMD Instinct MI300X AMD Instinct MI300X air-cooled accelerator with 192GB HBM3 memory, optimized for large language models with high memory capacity and strong compute performance, starting at $1.71/GPU HR.
• AMD Instinct MI325X AMD Instinct MI325X liquid-cooled accelerator with 256GB HBM3E memory, delivering exceptional bandwidth and compute density for large-scale training and production-grade inference, starting at $2.25/GPU HR.
• AMD Instinct MI355X AMD Instinct MI355X high-density accelerator with 288GB HBM3E memory and up to 2.8x training speedup, purpose-built for frontier model development and massive-scale inference workloads, starting at $2.95/GPU HR.
• AMD Instinct MI455X Next-generation AMD Instinct MI455X accelerator built on Helios rack-scale architecture with 432GB HBM4 memory, delivering approximately 40 PFLOPS FP4 dense performance with up to 72 GPUs per rack.
• Managed Kubernetes GPU-optimized Kubernetes service preconfigured with device and network operators for large-scale AI training, fine-tuning, and inference, handling control plane setup, upgrades, and security patching.
• Managed Slurm HPC job scheduling service for large-scale training and research on dedicated GPU cloud clusters, eliminating cluster setup, tuning, and maintenance overhead.
• High-Speed Network Storage Software-defined parallel file system providing a unified data layer across GPU-dense clusters with high-throughput, low-latency petabyte-scale flash storage for training, inference, and checkpointing.
• Enterprise Suite Full-stack production AI platform combining high-performance AMD GPU clusters, managed services, enterprise security, and the ScalarLM software layer for unified LLM workload orchestration.
• ScalarLM Open GPU-agnostic software layer for production LLM workloads that manages training, inference, scheduling, and optimization across clusters without vendor lock-in.
• AI Inference High-performance inference infrastructure built for modern AI workloads with more memory, longer contexts, higher throughput, and minimal latency on AMD Instinct accelerators.
• AI Model Fine-Tuning Single-node fine-tuning infrastructure using high-memory AMD Instinct GPUs for targeted model updates without complex multi-node clusters.
• AI Model Training Distributed AI training infrastructure with data-first architecture featuring RoCEv2 fabric, fast checkpointing, and managed Slurm for large-scale model development.
• Observability GPU monitoring and performance metrics platform providing real-time visibility into GPU health, usage, network performance, storage, and Slurm job insights.
• Expert Support White-glove technical support offering hands-on assistance from engineers for stability and performance optimization, with sub-30-minute response times.
• Security Suite Enterprise-grade cloud security including encryption, access governance, infrastructure protection, and third-party assurance for GPU infrastructure and customer workloads.

Quantifiable outcome

• 1.5-2x higher inference throughput with Modular collaboration
• +4 more outcomes

Named customers5 records

Segments3 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability4 records

Feature5 records

Partnerships

Five partnerships are on record, tiered core and supporting.

• ZyphracoreStrategic or Co-development Partner · 4 May 2026Zyphra Cloud launched as full-stack open AI platform powered by 15MW AMD Instinct MI355X GPU cluster on TensorWave infrastructure; offers bare-metal GPU deployments for pretraining, RL, post-training, and agentic inference
• GSMA Open Telco AIsupportingStrategic or Co-development Partner · 2 March 2026TensorWave serves as AMD's cloud partner providing 8,000 AMD GPUs to train telco-grade AI models as part of GSMA's global initiative involving 46 organizations across 35 countries
• Credo Technology GroupsupportingTechnology or Integration · 25 February 2026TensorWave partners with Credo to deploy ZeroFlap Active Electrical Cables and Optics across AMD-based AI cluster infrastructure; enables faster time to first token, improved network uptime, and 1,000x better reliability than legacy interconnect solutions
• TECfusionssupportingTechnology or Integration · 14 January 2026TensorWave selected TECfusions data center sites for combined 20MW AI infrastructure expansion: 10MW at Keystone Connect near Pittsburgh, PA and 10MW at Tucson, AZ facility; supports expansion to 1 GW capacity
• ModularcoreStrategic or Co-development PartnerCollaboration combining TensorWave's MI355X infrastructure with Modular's software stack (Mojo, MAX, Mammoth) delivering 1.5-2x higher inference throughput and 40-60% cost reduction; creates unified hardware-agnostic AI compute layer

Scale indicators10 records

Recent moves10 records

Expansion highlights6 records

Direct peers

• CoreWeave: Largest independent GPU cloud provider originally focused on Ethereum mining, now a major NVIDIA-based AI cloud. Direct competitor to TensorWave in offering large-scale GPU clusters for AI training and inference to enterprises and AI labs.
• Lambda: GPU cloud provider offering on-demand NVIDIA H100/H200 instances and dedicated clusters for AI training and inference. Competes head-to-head with TensorWave for AI lab and enterprise training/inference workloads with similar pricing models.
• Crusoe Energy: AI cloud provider that leverages stranded/renewable energy to power large-scale GPU clusters. Comparable to TensorWave in offering hyperscale-class GPU compute to AI customers, though Crusoe is NVIDIA-focused.
• Together AI: AI cloud platform focused on open-source large model training and inference, offering GPU clusters plus optimized inference stacks. Direct competitor to TensorWave in serving AI labs and enterprises with training and inference infrastructure.
• RunPod: GPU cloud provider offering on-demand and reserved NVIDIA GPU instances for AI training, inference, and serverless workloads. Competes with TensorWave in the mid-market and developer segment for GPU compute access.

Broad incumbents

• Vultr: Broad cloud infrastructure provider offering compute, storage, and bare-metal GPUs (including AMD instances). Competes with TensorWave in the AMD GPU cloud segment but as part of a wider general-purpose cloud portfolio.
• Paperspace (DigitalOcean): GPU cloud offering (now part of DigitalOcean) for AI/ML training and inference on NVIDIA GPUs. Comparable to TensorWave as a developer-focused GPU cloud, though broader cloud portfolio provides different scale dynamics.
• Amazon Web Services (EC2 P5/P4d instances): Hyperscaler offering GPU-accelerated instances for AI workloads via EC2. Competes with TensorWave for enterprise AI infrastructure spend with bundled networking, storage, and managed services.
• Google Cloud (A3 instances): Hyperscaler offering GPU instances (including NVIDIA H100/A100) and integrated Vertex AI platform. Competes with TensorWave for enterprise AI training and inference workloads with a comprehensive cloud ecosystem.

Emerging players

• Ori: Emerging GPU cloud provider offering AMD-based GPU instances (Instinct MI series) for AI workloads. Closest emerging competitor to TensorWave's AMD-exclusive positioning, though smaller in scale and customer base.

Market position

Strengths4 records

Weaknesses5 records

Competitive moat4 records

Key risks6 records

Key highlights7 records

Customer concentration

• Xx.com/TensorWaveCloud
• GitHubgithub.com/tensorwavecloud
• YouTubeyoutube.com/@TensorWaveCloud
• Facebookfacebook.com/tensorwavecloud
• LinkedInlinkedin.com/company/tensorwave
• Instagraminstagram.com/tensorwavecloud

Compliance5 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles15 records

Funding overview

Funding rounds6 records

Investors16 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about TensorWave