Cirrascale
Cirrascale Cloud Services operates a bare-metal neocloud providing private AI training and inference infrastructure across NVIDIA, AMD, Qualcomm, Tenstorrent, and Cerebras accelerators, serving enterprise AI teams, AI research institutions, government buyers, and AI startups in the US.
- Company typePrivate
- Founded2015
- HeadquartersPoway, United States
- Headcount51–100
- GTM typeB2B
- OfferingServices
What Cirrascale does
Cirrascale Cloud Services, LLC, headquartered in San Diego, California, is a privately held neocloud provider that delivers bare-metal, GPU-accelerated cloud infrastructure purpose-built for private AI training and inference workloads. The company operates a single multi-vendor AI Innovation Cloud that exposes every leading AI accelerator (NVIDIA HGX H100/H200/B200/B300 and RTX PRO 6000, AMD Instinct MI250/MI300X/MI325X/MI350, Qualcomm Cloud AI 100, Tenstorrent Galaxy Blackhole, and Cerebras Wafer-Scale Clusters) behind one unified control plane, with rail-optimized NVIDIA Quantum InfiniBand up to 3200Gb/server, NVMe hot-tier storage via the WEKA Data Platform, and GPU-direct storage tuning.
Its product surface extends beyond raw capacity to include the Cirrascale Inference Platform (an enterprise inference-as-a-service offering with automatic accelerator selection, dynamic multi-region workload balancing, serverless deployments, fine-tuned model support, RAG, and token-based pricing), Private Gemini on Google Distributed Cloud, Google GPAR Services for research institutions, the Qualcomm Inference Cloud with one-click model deployment, Cerebras AI Model Studio for GPT-class training, and Ai2 open-model endpoints (OLMo, Molmo, Tulu 3).
Cirrascale generates revenue through a mix of recurring bare-metal subscriptions (weekly to 36-month contract terms), usage-based inference token consumption, project-based professional services, and channel-driven sales via Google Cloud Marketplace, the NVIDIA Partner Network (where it holds the first Elite Cloud Services Provider designation for Visualization), Dell Platinum, Equinix/OrionVM colocation partners, and direct enterprise sales. Its primary customer segments are enterprise AI/ML teams, AI research institutions (notably Ai2), government and public-sector buyers reached via Google Public Sector, and AI startups served through the NVIDIA Inception program. The company traces its origins to Cirrascale Corporation, which spun off the cloud services business in early 2017, and as of 2026 employs 51-100 staff across US-West, US-East, and US-South data center regions.
Cirrascale firmographics
Firmographics- Name
- Cirrascale
- Legal name
- Cirrascale Cloud Services LLC
- Website
- https://cirrascale.com
- Company type
- Private
- Founded year
- 2015
- Operating status
- Operating
- Headcount range
- 51–100 employees
- Short description
- Cirrascale Cloud Services operates a bare-metal neocloud providing private AI training and inference infrastructure across NVIDIA, AMD, Qualcomm, Tenstorrent, and Cerebras accelerators, serving enterprise AI teams, AI research institutions, government buyers, and AI startups in the US.
- Ownership category
- akta.pro rank
Cirrascale industry classification
Industry- Product category
- AI Cloud Infrastructure
- NAICS
- Computer Systems Design and Related Services (54151)
- SIC
- Services-Computer Integrated Systems Design (7373), Computer Communications Equipment (3576)
- akta.pro primary industry
- AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers) (HDAAAAAG)
Keywords
Where Cirrascale is headquartered
LocationHeadquarters
- HQ city
- Poway
- HQ country
- United States
- HQ region
- North America
Offices3 records
Markets served
Cirrascale business model
Business model- GTM type
- B2B
- Offering type
- Services
- Cost components
- Infrastructure, Technology or R&D, Personnel, Operations, Marketing or Sales
Revenue model
- Bare-Metal Cloud Infrastructure (AI Innovation Cloud): Dedicated server services with entire servers reserved exclusively for customer usage. Customers purchase time on dedicated multi-GPU servers with configurable terms from weekly to 36-month contracts. Revenue generated through recurring subscription payments based on server configuration and term length.
- Professional Services: Custom support services and configuration add-ons provided on a per-service or time-and-materials basis (30-minute incremental billing). Includes third-party software installation, custom configurations, and technical support beyond standard service plan scope.
- Inference Platform (Token-Based): Usage-based pricing for inference services on the Cirrascale Inference Platform, where customers pay based on token consumption for deployed AI models. Enables scalable, predictable billing for high-volume inference workloads.
- Cerebras AI Model Studio: Cloud access to Cerebras Wafer-Scale Clusters for training GPT-class models, offered with model-as-a-service pricing and optional free trial evaluation periods.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Subscription | Monthly | Service Term Length Options |
| Subscription | Monthly | Multi-GPU RTX Cloud Servers |
| Usage-based | Pay-as-you-go | Inference Platform Token-Based Pricing |
| Subscription | Monthly | Private Gemini Endpoint |
Go-to-market motion1 record
Distribution channels4 records
Marketing channels6 records
Cirrascale product offering
Product offeringCore offering
Cirrascale provides bare-metal cloud infrastructure purpose-built for AI training and inference workloads, offering dedicated GPU and accelerator capacity across NVIDIA, AMD, Qualcomm, Tenstorrent, and Cerebras hardware. The company delivers its AI Innovation Cloud platform, an inference optimization layer, and managed services for private deployment of large language models, with dedicated single-tenant and multi-tenant configurations and no data transfer fees.
Product overview
Cirrascale Cloud Services is a specialized private AI cloud and managed services provider offering the AI Innovation Cloud as its core product platform for AI training and inference workloads. The platform is the only cloud service enabling access to every leading AI accelerator (NVIDIA, AMD, Qualcomm, Tenstorrent, Cerebras) in one unified environment. Supporting products include the Cirrascale Inference Platform for enterprise model deployment with dynamic workload balancing and RAG support, Private Gemini on Google Distributed Cloud for secure on-premises AI, Google GPAR Services for research and government sectors, and specialized offerings including NVIDIA Inception support, Multi-GPU Compute, Cloud Storage Solutions (WEKA-integrated), and high-speed Networking (NVIDIA Quantum InfiniBand). The company also offers AI ecosystem partnerships with validated integrations for Kubeflow, H2O AI, IBM Power AI, and One Convergence (DKube).
Differentiator
Problem solved
Functional benefit
Products and services
- AI Innovation Cloud
- Cirrascale Inference Platform
- Private Gemini
- Google GPAR Services
- Cerebras AI Model Studio
- Qualcomm Inference Cloud
- Ai2 Model Endpoints
- Private AI
- Cloud Storage Solutions
- Networking
Quantifiable outcome
- 8x faster time to accuracy at half the cost compared to traditional cloud providers when using Cerebras AI Model Studio
- +3 more outcomes
Companies that use Cirrascale
Customer profileNamed customers4 records
Segments4 records
Ideal customer profiles4 records
Cirrascale technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
Integration6 records
AI capability11 records
Feature5 records
Cirrascale partnerships and signals
Strategic signalPartnerships
Eleven partnerships are on record, tiered core, major and minor.
- TenstorrentcoreCirrascale Cloud Services added Tenstorrent Galaxy Blackhole servers to its AI Innovation Cloud, marking Tenstorrent's entry into broad commercial deployment. Enables enterprise and research teams to deploy production AI systems through Cirrascale's bare-metal delivery model.
- EquinixmajorEquinix listed as deployment partner for Tenstorrent Galaxy Blackhole AI compute system, integrating infrastructure into Equinix colocation and data center services.
- OrionVMminorOrionVM listed as deployment partner for Tenstorrent Galaxy infrastructure integration into AI platforms and services.
- GooglecoreCirrascale partners with Google Public Sector to deliver Private Gemini deployments on Google Distributed Cloud, Google GPAR Services for research institutions, and Private Gemini Endpoint available on Google Cloud Marketplace. Enables enterprise and government customers to run Gemini AI with data staying behind their firewall.
- AMDcoreCirrascale offers AMD Instinct Series Cloud including MI250, MI300X, MI325X, and upcoming MI350 Series GPUs. Collaboration with AMD to deliver tailored cloud environments built for unique model architectures and performance needs.
- QualcommcoreCirrascale offers Inference Cloud powered by Qualcomm AI Inference Suite and Qualcomm Cloud AI 100 instances. Includes AI Playground for developer experimentation and pre-configured applications for chatbots, summarization, code development, image generation, transcription, and RAG.
- CerebrascoreCerebras AI Model Studio hosted on Cerebras Cloud @ Cirrascale enabling customers to train GPT-class models on Cerebras Wafer-Scale Clusters including Andromeda AI supercomputer. Offers 8x faster time to accuracy at half the cost of traditional cloud providers.
- NVIDIAcoreNVIDIA Partner Network Cloud Provider offering NVIDIA GPU Cloud services including HGX H100, H200, B200, B300, RTX PRO 6000, and other NVIDIA GPU offerings. Also NVIDIA Elite Cloud Services Provider for Visualization and NVIDIA Cloud Partner with WekaIO certification.
- Dell TechnologiesmajorCirrascale is a Platinum Partner of Dell Technologies, deploying Dell's industry-leading storage and hardware technologies across its AI Innovation Cloud for reliable, high-performance foundation for AI and deep learning workloads.
- WekaIOmajorWEKA Data Platform integrated for removing storage bottlenecks for deep learning training datasets with massively parallel access to NVMe drives. WEKA certified as high-performance data store solution for NVIDIA Cloud Partners.
- Ai2 (Allen Institute for Artificial Intelligence)coreCirrascale built custom infrastructure for Ai2 using NVIDIA Hopper GPUs, optimized storage, and design focused on future growth. Ai2 trains open-source models like Molmo, OLMo, and Tülu 3 on Cirrascale infrastructure.
Scale indicators4 records
Recent moves7 records
Expansion highlights5 records
Cirrascale competitors and assessment
Company assessmentDirect peers
- CoreWeave: GPU-focused neocloud offering bare-metal and virtualized NVIDIA GPU compute (H100/H200/B200) for AI training and inference. Most direct comparable as a specialized AI infrastructure provider with similar customer base (AI labs, enterprises) and bare-metal positioning.
- Lambda Labs: AI cloud and GPU workstation provider operating dedicated NVIDIA GPU clusters (H100, H200, B200) for training and inference. Closely comparable in target customer (AI researchers and startups), subscription bare-metal model, and emphasis on developer experience.
- Together AI: AI cloud platform purpose-built for training and serving open-source and frontier models, with token-based inference pricing and multi-GPU clusters. Comparable business model to Cirrascale Inference Platform with usage-based inference monetization.
- Equinix Metal: Bare-metal cloud service from Equinix offering dedicated servers with automated provisioning across global IBX data centers. Directly comparable on bare-metal, unvirtualized infrastructure delivery; also a Cirrascale colocation partner via Equinix.
- Paperspace (DigitalOcean): GPU cloud offering dedicated GPU instances for ML training and inference, now part of DigitalOcean. Comparable offering for AI/ML workloads with similar developer- and SMB-focused positioning and per-month subscription pricing.
- Vultr: Cloud provider offering bare-metal servers, GPU instances (NVIDIA H100/A100), and dedicated hosting across 32+ global locations. Comparable on bare-metal, multi-region, hourly/monthly subscription pricing for AI and general compute workloads.
Broad incumbents
- Google Cloud Platform: Hyperscaler offering AI/ML infrastructure including GPUs, TPUs, and managed AI services. Broad incumbent also serving as a Cirrascale channel partner through Google Cloud Marketplace and Google Distributed Cloud, blurring competitive vs. partner dynamics.
- Amazon Web Services (AWS): Hyperscaler dominating cloud AI infrastructure with EC2 GPU instances, Trainium, and Bedrock. Primary broad incumbent whose scale, pricing, and AI service portfolio define the competitive ceiling Cirrascale's bare-metal positioning must differentiate against.
- Microsoft Azure: Hyperscaler providing GPU-accelerated VMs (ND H100/B200 series) and Azure AI services for enterprise training and inference. Broad incumbent whose OpenAI alliance and Azure-OpenAI service compete directly for AI inference workloads Cirrascale targets.
Emerging players
- Crusoe Energy: AI cloud provider operating large-scale NVIDIA GPU clusters (H100/H200) powered by stranded/clean energy. Emerging neocloud with overlapping AI training/inference customer focus and bare-metal delivery, though anchored in energy-cost differentiation.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat6 records
Key risks7 records
Key highlights7 records
Customer concentration
Cirrascale social profiles
Digital presenceCirrascale financial estimates
Financial estimateRevenue estimate
Valuation estimate
Cirrascale leadership team
Management profileNumber of profiles
Profiles4 records
Cirrascale funding detail
Funding detailFunding overview
Funding rounds7 records
Investors5 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Cirrascale M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Cirrascale
What does Cirrascale do?
Cirrascale provides bare-metal cloud infrastructure purpose-built for AI training and inference workloads, offering dedicated GPU and accelerator capacity across NVIDIA, AMD, Qualcomm, Tenstorrent, and Cerebras hardware. The company delivers its AI Innovation Cloud platform, an inference optimization layer, and managed services for private deployment of large language models, with dedicated single-tenant and multi-tenant configurations and no data transfer fees.
Is Cirrascale a public or private company?
Cirrascale is a private company. It is classified as unknown and is currently operating.
When was Cirrascale founded?
Cirrascale was founded in 2015. It employs 51 to 100 people.
Where is Cirrascale based?
Cirrascale is headquartered in Poway, United States, in the North America region.
How does Cirrascale make money?
Four revenue lines are on record. Bare-Metal Cloud Infrastructure (AI Innovation Cloud) is the primary driver. The others are professional Services, inference Platform (Token-Based) and cerebras AI Model Studio.
Who are Cirrascale's main competitors?
Direct peers on record are CoreWeave, Lambda Labs, Together AI, Equinix Metal, Paperspace (DigitalOcean) and Vultr. Broad incumbents are Google Cloud Platform, Amazon Web Services (AWS) and Microsoft Azure. Crusoe Energy is listed as an emerging player.
Does Cirrascale have an API?
No public API is recorded for Cirrascale.
What industry is Cirrascale in?
Cirrascale's product category is AI Cloud Infrastructure. Its primary akta.pro industry code is HDAAAAAG, AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers). Its NAICS code is 54151 and its SIC code is 7373.