What QumulusAI does

QumulusAI, Inc. is a U.S.-based AI cloud infrastructure company founded in 2019 in Georgia and now headquartered in Atlanta's Georgia Tech Tech Square district. The company operates a vertically integrated stack spanning owned power generation, Tier 3+ modular data centers (targeting sub-1.1 PUE), and a GPU-as-a-Service (GPUaaS) cloud delivered on the latest NVIDIA platforms including B300, B200, H200, H100, RTX Pro 6000, and L4 across more than 150 MW of available U.S. colocation capacity.

The product suite is delivered as three tiers — QumulusAI Cloud (subscription-based shared GPUs scaling in single-GPU increments), QumulusAI Cloud Pro (dedicated 1:1 GPUs in 2/4/8 GPU nodes), and QumulusAI Cloud Pure (full bare-metal clusters) — and is differentiated by a workload-optimized inference architecture that reduces AI inference cost by approximately 20% versus standard reference builds, plus a managed Kubernetes offering powered by vCluster technology. Customer-facing technology is supported by proprietary frameworks including the FACTS diagnostic (Flexibility, Access, Cost, Trust, Speed) and the 'Hyperspeed Compute' category positioning co-developed with HyperFRAME Research.

QumulusAI generates revenue primarily through multi-year GPU-as-a-Service subscription contracts (e.g., $124M+ in three-year deals with Hyperbolic and another inference platform covering 1,280 NVIDIA Blackwell GPUs) supplemented by on-demand marketplace consumption across 5 GPUaaS platforms. Distribution combines direct enterprise sales, channel partnerships (Shadeform, Runpod), the QAI Moon joint venture with Moonshot Energy and Connected Nation Internet Exchange Points (targeting 25 to 125 U.S. university campus sites), and an emerging edge inference business. As of late 2025, approximately 800 active GPUs were deployed across U.S. colocation facilities with a 2026 roadmap targeting more than 23,000 NVIDIA Blackwell GPUs, supported by $616M+ in non-dilutive capital across the USD.AI, ATW Partners, and TFC facilities, ahead of a planned direct listing on Nasdaq under ticker QMLS.

Name

QumulusAI

Legal name

QumulusAI, Inc.

Website

https://qumulusai.com

Company type

Private

Founded year

2019

Operating status

Operating

Headcount range

11–50 employees

Short description

QumulusAI is a U.S.-based AI cloud infrastructure company delivering GPU-as-a-Service on a vertically integrated stack of owned power generation, modular data centers, and NVIDIA GPU clusters. It serves AI startups, inference platforms, enterprises, and research institutions across a hyper-distributed U.S. footprint via direct sales, channel partners, and the QAI Moon joint venture.

Ownership category

akta.pro rank

Product category

AI Cloud Infrastructure

NAICS

Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182), Computer Systems Design and Related Services (5415), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518)

SIC

Electronic Computers (3571), Services-Prepackaged Software (7372), Computer Communications Equipment (3576)

akta.pro primary industry

AI Compute Cloud & GPU-as-a-Service (HDAAAAAK)

akta.pro secondary industries

AI Datacenter Power, Cooling & Racks (liquid cooling, power delivery) (HDAAAAAH), GPU-Accelerated & AI Training/Inference Servers (HDACABAG), Hybrid Cloud Compute & Virtualization Stack (HCI/VMs/Kubernetes) (HDABAMAF)

Keywords

• GPU cloud services
• AI infrastructure
• GPU-as-a-Service
• Bare metal compute
• Data center hosting

Headquarters

HQ city

Marietta

HQ country

United States

HQ region

North America

Offices9 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-as-a-Service Subscriptions: Long-term recurring revenue from multi-year GPU-as-a-Service subscription contracts (e.g., 3-year terms with Hyperbolic and another AI inference platform totaling $124M; 2-year contracts with Shadeform for 680 H200 GPUs). Includes upfront customer commitments structured into long-tail recurring payments.
2. On-Demand GPU Cloud Consumption: Usage-based consumption revenue from shared GPU access and on-demand marketplace listings (listed on 5 GPUaaS marketplaces). Approximately 800 active GPUs were deployed across U.S. colocation facilities as of late 2025.

Pricing tiers

Go-to-market motion1 record

Distribution channels6 records

Marketing channels8 records

Core offering

QumulusAI provides GPU-as-a-Service (GPUaaS) for large-scale AI workloads through three product tiers — QumulusAI Cloud (shared GPUs), Cloud Pro (dedicated GPUs), and Cloud Pure (bare metal clusters) — running on a hyper-distributed network of sub-50MW Tier 3+ modular data centers. The offering spans the latest NVIDIA GPU platforms (B300, B200, H200, H100, RTX Pro 6000, L4) and includes behind-the-meter natural gas power agreements and integrated managed Kubernetes via vCluster for inference, fine-tuning, and training workloads.

Differentiator

Problem solved

Functional benefit

Brands

• QumulusAI Cloud: Shared GPU access product for inference, fine-tuning, and experimentation; elastic scaling in single-GPU increments with transparent, usage-based pricing.
• QumulusAI Cloud Pro
• QumulusAI Cloud Pure
• QAI Moon

Products and services

• QumulusAI Cloud (Shared GPUs) Subscription-based shared GPU pool providing elastic access to NVIDIA GPUs in single-GPU increments for inference, prototyping, and fine-tuning. Available on-demand or as reserved capacity with transparent, usage-based pricing. Targeted at AI developers and teams needing flexible GPU access without dedicated hardware commitments.
• QumulusAI Cloud Pro (Dedicated GPUs) Dedicated 1:1 GPU compute with guaranteed access and flexible scaling in 2, 4, or 8 GPU increments for model training, optimization, and continuous deployment. Targeted at startups and teams scaling production AI workloads requiring reliable, exclusive compute.
• QumulusAI Cloud Pure (Bare Metal Clusters) Full bare-metal clusters with no virtualization layer, delivered as exclusive 1:1 nodes in single or multi-node increments with reservations beginning at one month. Targeted at hyperscale training workloads and enterprises needing maximum performance, predictability, and control without hypervisor overhead.

Quantifiable outcome

• Approximately 20% reduction in AI inference costs versus standard reference architectures
• +3 more outcomes

Named customers3 records

Segments5 records

Ideal customer profiles5 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Integration4 records

Feature9 records

Partnerships

15 partnerships are on record, tiered core networking partner, core hardware partner, core channel partner, flagship strategic partnership, flagship strategic jv partner, flagship deployment partner, supporting infrastructure partner, core technology partner, flagship customer / co-marketing partner, minor channel partner and core analyst / research partner.

• Cisco (Cisco Nexus One)core networking partnerTechnology or Integration · 11 June 2026Cisco Nexus One powers the cluster fabric for the Lenovo and Supermicro bare-metal server deployments supporting Hyperbolic and another AI inference platform, delivering secure, high-performance AI networking for $124M+ in customer commitments covering 1,280 NVIDIA Blackwell GPUs.
• Lenovocore hardware partnerTechnology or Integration · 11 June 2026Lenovo supplies bare-metal servers equipped with NVIDIA Blackwell GPUs deployed by QumulusAI under $124M+ in customer subscriptions to Hyperbolic and another AI inference platform.
• Supermicrocore hardware partnerTechnology or Integration · 11 June 2026Supermicro supplies bare-metal servers equipped with NVIDIA Blackwell GPUs deployed by QumulusAI under $124M+ in customer subscriptions to Hyperbolic and another AI inference platform.
• Shadeformcore channel partnerChannel Partner/ Reseller/ Distributor · 28 May 2026Shadeform is a unified GPU platform connecting AI teams with compute resources from over 30 cloud partners. Two-year contract for deployment of two NVIDIA H200 clusters totaling 85 nodes (680 GPUs) at QumulusAI's Kansas City location serving two high-growth AI inference platforms. Shadeform brings committed multi-year clients to QumulusAI's hyperdistributed data center footprint.
• vCluster (vCluster Labs)flagship strategic partnershipStrategic or Co-development Partner · 24 March 2026QumulusAI and vCluster announced a strategic partnership to enable enterprises to deploy secure, isolated Kubernetes environments on shared GPU infrastructure for AI development. Together they established the vCluster AI Lab running on QumulusAI's distributed GPU infrastructure powered by NVIDIA Blackwell B300 and RTXPRO 6000 platforms, allowing vCluster engineers to prototype features and test emerging AI workloads. Through this collaboration QumulusAI offers a managed Kubernetes solution powered by vCluster technology.
• Moonshot Energyflagship strategic jv partnerStrategic or Co-development Partner · 15 January 2026Joint venture (QAI Moon) between Moonshot Energy and QumulusAI to commercialize distributed, inference-optimized GPU-as-a-Service at national scale. Moonshot provides modular electrical infrastructure, manufacturing capacity (500,000 sq ft facility in Lewisville, TX), and AI Pods. QumulusAI provides GPU orchestration, workload delivery, and the commercial operating model. Together with Connected Nation Internet Exchange Points, deploying 25 sites initially and scaling to 125 across U.S. university campuses.
• Connected Nation Internet Exchange Points (IXP.us)flagship deployment partnerStrategic or Co-development Partner · 15 January 2026IXP.us provides carrier-neutral Internet Exchange Points and physical real estate for AI Pod colocation at 125 regional hub communities, primarily on U.S. research university campuses. Through the QAI Moon joint venture, IXP.us becomes the anchor interconnection partner for low-latency AI compute deployment. First deployment slated for July 2026 at Wichita State University.
• DE-CIXsupporting infrastructure partnerTechnology or Integration · 15 January 2026DE-CIX serves as the Internet Exchange (IX) operator for IXP.us facilities, providing redundant 400G IX ports on the DE-CIXaaS switch at each IXP.us site to enable ultra-low-latency AI compute deployment through QAI Moon.
• TOWARDEXsupporting infrastructure partnerImplementation/ SI/ Consulting Partner · 15 January 2026TOWARDEX provides diverse manhole and conduit access design and construction for IXP.us facilities, including the nation's first large-scale Meet-Me Street network (Hub Express System in Boston).
• Connectbasesupporting infrastructure partnerTechnology or Integration · 15 January 2026Connectbase provides system-of-record visibility for all fiber, transport, and IP transit networks entering each IXP.us site, supporting QAI Moon deployments.
• NVIDIAcore technology partnerTechnology or IntegrationPrimary GPU supplier across the entire QumulusAI platform — NVIDIA B300 SXM6, B200 SXM, RTX Pro 6000, H200 SXM, H100 SXM, H100 NVL, and L4 cards power all QumulusAI Cloud, Cloud Pro, and Cloud Pure deployments.
• Hyperbolicflagship customer / co-marketing partnerStrategic or Co-development PartnerHyperbolic is an AI cloud platform providing scalable GPU compute for AI startups, research teams, and enterprises; named 3-year customer of QumulusAI covering NVIDIA Blackwell GPUs and referenced as co-research partner in QumulusAI's joint research with HyperFRAME Research.
• Runpodminor channel partnerChannel Partner/ Reseller/ DistributorRunpod is identified as a leading GPUaaS marketplace partner used by QumulusAI for on-demand marketplace distribution since 2023.
• HyperFRAME Researchcore analyst / research partnerGTM or Marketing PartnerHyperFRAME Research (CEO Steven Dickens) co-developed research briefs with QumulusAI including 'The Hyperspeed Compute Era' and 'AI Infrastructure Has Two Bottlenecks: Access and Utilization', and has produced multiple independent analyses of QumulusAI's strategy.
• Data Air Flowsupporting infrastructure partnerImplementation/ SI/ Consulting PartnerData Air Flow provides end-to-end mechanical cooling solutions and expert commissioning services through a strategic collaboration with Moonshot on all QAI Moon deployments.

Scale indicators15 records

Recent moves10 records

Expansion highlights6 records

Direct peers

• Vultr: Independent cloud provider offering GPU instances alongside general compute; competes with QumulusAI for AI startup and mid-market GPU workloads via cloud marketplace listings.
• Runpod: GPU-as-a-Service marketplace and cloud provider; QumulusAI lists its NVIDIA GPUs on Runpod's marketplace as a distribution channel, making Runpod both a partner and a direct neocloud competitor.
• Lambda Labs: GPU cloud and neocloud provider offering shared, dedicated, and on-demand NVIDIA GPU instances plus on-prem hardware; cited alongside CoreWeave as QumulusAI's primary neocloud benchmark.
• Applied Digital: AI cloud infrastructure operator building large-scale GPU data centers; QumulusAI's CEO Mike Maniscalco was previously CTO at Applied Digital, and the companies share a vertically integrated hosting-and-power heritage.
• Crusoe Energy: Neocloud specializing in GPU compute powered by stranded/clean energy; directly comparable to QumulusAI's behind-the-meter natural gas power strategy and AI cloud offering.
• CoreWeave: Largest independent neocloud providing NVIDIA GPU compute (H100, H200, B200) on a vertically optimized stack; directly competes with QumulusAI for AI training and inference customers and is the explicit benchmark in QumulusAI's analyst coverage.
• Crusoe Cloud: Crusoe's managed AI cloud offering delivering NVIDIA GPU instances with a vertically integrated energy-and-data-center stack; same vertically integrated neocloud category as QumulusAI.

Emerging players

• Together AI: AI cloud platform optimized for open-source model inference and training; overlaps with QumulusAI's target inference-platform customer segment (e.g., Hyperbolic) and competes for the same workloads.
• Hyperbolic: AI cloud platform providing on-demand GPU compute for AI builders and researchers; a flagship QumulusAI customer but also an emerging neocloud competitor addressing the same inference workloads.

Broad incumbents

• Oracle Cloud Infrastructure (OCI): Hyperscaler with a dedicated focus on large-scale NVIDIA GPU clusters (including OCI Supercluster); competes with QumulusAI for enterprise AI workloads but as part of a much broader cloud portfolio.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks7 records

Key highlights7 records

Customer concentration

• LinkedInlinkedin.com/company/qumulusai
• Xx.com/QumulusAI

Revenue estimate

Valuation estimate

Number of profiles

Profiles7 records

Subsidiaries1 record

Funding overview

Funding rounds4 records

Investors1 record

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 QumulusAI