Runpod
Runpod is a 2022-founded AI Developer Cloud platform offering GPU compute (Pods, Serverless, Clusters) and a model marketplace (Hub) to over one million developers at companies including OpenAI, Anthropic, Perplexity, Cursor, and Replit, with $120M ARR and usage-based per-second billing.
- Company typePrivate
- Founded2022
- HeadquartersSan Francisco, United States
- Headcount101–250
- GTM typeB2B
- OfferingSoftware
What Runpod does
Runpod is an AI Developer Cloud platform founded in 2022 by Zhen Lu and Pardeep Singh (former Comcast developers) that aggregates GPU compute capacity across 28 partner datacenters and exposes it through three core primitives: Pods (dedicated single-node GPU instances supporting 30+ GPU SKUs), Serverless (auto-scaling API endpoints with sub-200ms cold starts via FlashBoot), and Clusters (multi-node InfiniBand/RoCE v2-connected GPU environments scaling up to 10,000+ GPUs). The platform is complemented by Runpod Hub, a model and template marketplace for one-click deployment of open-source AI models (Qwen, FLUX, Seedream, Wan, SORA, Kling, Whisper, Granite), and Runpod Flash, an open-source MIT-licensed Python SDK that eliminates Docker containerization from serverless GPU development. The platform serves over one million developers across 31 global regions at companies including OpenAI, Anthropic, Perplexity, Cursor, Replit, Wix, Zillow, Databricks, and Civitai, with named case studies in healthcare (Aneta), creative production (TOOL), generative media (Coframe, Glam Labs, Segmind, Civitai), and conversational AI (Scatter Lab).
The business is monetized through usage-based per-second GPU billing across Pods, Serverless, and Clusters, plus subscription revenue from storage tiers and Reserved Cluster contracts (1mo/3mo/6mo/12mo/12mo+) for guaranteed capacity up to 10,000+ GPUs, and transaction fees on Public Endpoints (per-request, per-token, per-megapixel, or per-second of video). Go-to-market is product-led growth via self-serve signup with no credit card required, supplemented by enterprise sales for Reserved Clusters, channel partnerships with a16z speedrun and Santa Cruz Works, and a developer community motion centered on Discord and organic word-of-mouth (900,000+ developers reached without a salesperson). Runpod is SOC 2 Type II, HIPAA, and GDPR compliant, enabling regulated workloads in healthcare, biotech, and the EU.
The company has raised $120M total: a $20M seed in May 2024 co-led by Intel Capital and Dell Technologies Capital (with angels Julien Chaumond and Nat Friedman), followed by a $100M growth investment in June 2026 led by Summit Partners at a $1.0 billion valuation. Runpod reported $120M in annual recurring revenue as of January 2026, grew its developer base from 500K to 1M+ between January and June 2026, and has processed 20+ billion inference requests on its Serverless platform. The company operates a remote-first model with headquarters in Moorestown, NJ, an East Coast office in Charlotte, NC (opened November 2024), and team presence in San Francisco, Canada, Europe, and India, with plans to scale toward a team of several hundred.
Runpod firmographics
Firmographics- Name
- Runpod
- Legal name
- Runpod Inc.
- Website
- https://runpod.io
- Company type
- Private
- Founded year
- 2022
- Operating status
- Operating
- Headcount range
- 101–250 employees
- Short description
- Runpod is a 2022-founded AI Developer Cloud platform offering GPU compute (Pods, Serverless, Clusters) and a model marketplace (Hub) to over one million developers at companies including OpenAI, Anthropic, Perplexity, Cursor, and Replit, with $120M ARR and usage-based per-second billing.
- Ownership category
- akta.pro rank
Where Runpod is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices3 records
Markets served
Runpod business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Infrastructure, Technology or R&D, Personnel, Operations, Marketing or Sales
Revenue model
- GPU Pod rentals (Cloud GPUs): Usage-based revenue from renting dedicated single-node GPU instances across 30+ GPU SKUs in 31 global regions. Billed by the second (effectively millisecond precision) with no egress fees. Customers can pay on-demand or commit to reservations for discounts. Drives the majority of platform consumption.
- Serverless GPU endpoints: Usage-based revenue from auto-scaling API inference endpoints billed by active compute time (per-second), with zero idle cost when endpoints are not running. Mix of flex workers, always-on workers, and reserved workers.
- Clusters and Reserved Clusters: Multi-node GPU cluster revenue billed by the second on a pay-as-you-go basis, plus multi-month and annual Reserved Cluster contracts (1mo/3mo/6mo/12mo/12mo+) for guaranteed capacity up to 10,000+ GPUs with custom configurations, SLA-backed uptime, and discounted rates. Reserved Clusters are quote-based (not publicly disclosed).
- Public Endpoints (Hub API): Transaction-based revenue from public, pre-deployed AI model endpoints (Qwen, FLUX, Seedream, Wan, Sora, Kling, Whisper, Granite, etc.) billed per request, per token, per second of video, per megapixel, or per character depending on modality. Examples include $0.00001 per 1M tokens (Deep Cogito), $0.025 per request (FLUX.1 Kontext), $1.20 per 4s SORA 2 Pro I2V request.
- Storage and Network Volumes: Recurring storage revenue from Container Disk ($0.10/GB/mo), Volume Disk (running $0.10/GB/mo, idle $0.20/GB/mo), and Network Storage tiers ($0.05–$0.14/GB/mo standard and high-performance).
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Usage-based | Pay-as-you-go | Cloud GPU Pods — 30+ GPU SKUs, per-second billing, on-demand and reserved tiers |
| Usage-based | Pay-as-you-go | Serverless GPU endpoints — per-second worker billing with auto-scaling |
| Usage-based | Pay-as-you-go | Clusters — multi-node GPU clusters, per-second billing, up to 64 GPUs (8 nodes) |
| Subscription | Multi-year contract | Reserved Clusters — dedicated GPU capacity with guaranteed availability and SLAs |
| Subscription | Monthly | Storage tiers — Container Disk, Volume Disk, Network Storage |
| Transaction based/ take rate | Pay-as-you-go | Public Endpoints — pre-deployed AI models with per-request API pricing |
Go-to-market motion5 records
Distribution channels6 records
Marketing channels10 records
Runpod product offering
Product offeringCore offering
Runpod is a software-orchestrated AI Developer Cloud that aggregates GPU capacity across 28 partner datacenters and exposes it as three primitives: Pods (dedicated single-node GPU instances), Serverless (auto-scaling API endpoints), and Clusters (multi-node InfiniBand/RoCE v2 connected GPU clusters). The platform supports 30+ GPU SKUs across 31 global regions and is used by developers for AI training, fine-tuning, and production inference.
Product overview
Runpod is a unified AI Developer Cloud platform rather than a single product. It is architected as a platform-plus-modules offering covering the full AI development lifecycle, where the same account and identity flow from experiment to production. The core platform comprises four tightly integrated products: Runpod Pods (Cloud GPUs) for dedicated, on-demand GPU instances across 30+ GPU SKUs and 31 global regions; Runpod Serverless for auto-scaling API endpoints with sub-200ms cold starts via FlashBoot; Runpod Clusters for multi-node InfiniBand-connected training environments supporting up to 64 GPUs (or 10,000+ via Reserved Clusters); and Runpod Hub, a catalog of deployable open-source AI models, templates, and apps. Runpod Flash is an open-source Python SDK that extends the platform by letting developers deploy serverless GPU workloads without Docker. The platform is also accessible via Public Endpoints (pre-deployed AI model APIs), and storage options include Network Volumes, Container Disks, and persistent storage with no egress fees. Runpod reports over 900,000 developers served across 183 countries, $120M ARR, and 20+ billion inference requests processed.
Differentiator
Problem solved
Functional benefit
Brands
- Runpod Serverless: Serverless GPU inference endpoints with auto-scaling, FlashBoot sub-200ms cold starts, and pay-per-use pricing.
Products and services
- Runpod Cloud GPUs (Pods) On-demand dedicated GPU instances deployed in under 30 seconds across 31 global regions. Supports 30+ GPU SKUs from RTX A5000 to B300, billed by the second with no egress fees. Used for development, training, fine-tuning, batch jobs, and long-running workloads requiring direct control over the container, storage, GPU type, and runtime environment.
- Runpod Serverless Auto-scaling GPU API endpoints for containerized inference workloads. Workers scale from zero to thousands on demand via FlashBoot with sub-200ms cold starts and zero idle cost. Supports flexible runtimes (Python, Node.js, Go, Rust, C++), GitHub-based deployment, and custom Docker images.
- Runpod Clusters Multi-node GPU environments with 1,600-3,200 Gbps InfiniBand/RoCE v2 east-west bandwidth for distributed training and large batch workloads. Self-service provisioning scales up to 64 GPUs per cluster (or 10,000+ via Reserved Clusters), with Slurm orchestration support, per-second billing, and tenant-isolated network fabrics.
- Runpod Hub Catalog of preconfigured AI repositories, templates, models, and open-source AI apps deployable on Runpod with one click. Includes public endpoints for ready-to-use AI models (text, image, video, audio), GitHub integration, automated build/test pipelines, and a publishing workflow for open-source contributors.
- Runpod Flash Open-source MIT-licensed Python SDK that eliminates Docker containerization from serverless GPU development. Uses @Endpoint decorator with typed GPU enums, supports queue-based batch jobs, load-balanced HTTP APIs, custom Docker images, and existing endpoints. Includes flash build, flash deploy, EndpointJob API for async work, cross-endpoint function calls, and coding agent integration for Claude Code/Cursor/Cline.
Quantifiable outcome
- $120M ARR (as of January 2026)
- +14 more outcomes
Companies that use Runpod
Customer profileNamed customers24 records
Segments6 records
Ideal customer profiles4 records
Runpod technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration7 records
AI capability14 records
Feature10 records
Runpod partnerships and signals
Strategic signalPartnerships
Four partnerships are on record, tiered core, flagship and minor.
- vLLMcoreRunpod partnered with vLLM to enhance AI inference performance, leveraging vLLM's PagedAttention algorithm and providing GPU resources for testing. The collaboration emphasizes Runpod's commitment to the open-source community and improving AI infrastructure and accessibility.
- OpenAIflagshipRunpod named OpenAI's official infrastructure partner for the Model Craft Challenge Series. Runpod and OpenAI will distribute up to $1M in compute credits supporting the first challenge, Parameter Golf, which engaged 1,100 researchers.
- a16z speedruncoreRunpod joined a16z speedrun as an official infrastructure partner, giving founders in the program access to Runpod GPU infrastructure as part of speedrun's $5M+ perks package. Founders building through speedrun can now build with Runpod starting day one.
- Santa Cruz WorksminorSanta Cruz Works has partnered with Runpod to provide GPU-first cloud infrastructure, facilitating AI startups in accessing scalable compute resources rapidly and efficiently, thereby enhancing innovation and production capabilities within the Santa Cruz ecosystem.
Scale indicators15 records
Recent moves7 records
Expansion highlights6 records
Runpod competitors and assessment
Company assessmentDirect peers
- CoreWeave: CoreWeave is a vertically integrated GPU cloud provider offering on-demand H100/H200/B200 instances, serverless inference, and clusters at massive scale. It is the most directly comparable neocloud competitor to Runpod, targeting overlapping AI developer and enterprise workloads.
- Lambda Labs: Lambda provides dedicated GPU instances and clusters (on-demand and reserved) for AI training and inference, with a similar per-GPU pricing model and AI-developer focus. It competes head-to-head with Runpod on Pods and Clusters workloads, particularly among AI researchers and startups.
- Crusoe Energy: Crusoe operates large-scale GPU datacenters powered by stranded energy, offering AI cloud services including on-demand GPUs and clusters. It targets enterprise-scale AI training and inference workloads in direct competition with Runpod's Reserved Clusters offering.
- Modal Labs: Modal offers serverless GPU compute for AI workloads with a Python-native developer experience, focusing on inference and batch jobs. It is the closest direct competitor to Runpod's Serverless product and is frequently cited in Runpod's own comparison content.
- Baseten: Baseten provides a serverless inference platform for production ML models, with autoscaling, GPU support, and a focus on ML engineering teams. It competes with Runpod Serverless for production AI inference workloads.
- fal.ai: fal.ai is a developer-focused serverless GPU platform specializing in fast inference for generative media models (image, video, audio). It directly competes with Runpod's Public Endpoints and Serverless offerings in the generative AI developer segment.
- Together AI: Together AI operates an AI cloud platform combining GPU infrastructure, serverless inference, and hosted open-source foundation models. It competes with Runpod across the full lifecycle (training, fine-tuning, inference) with a comparable developer-led GTM.
- Vast AI: Vast AI is a peer-to-peer GPU marketplace aggregating capacity from individual providers for AI compute rentals. It competes with Runpod Pods on price-sensitive GPU rental workloads, particularly among hobbyists and individual developers.
Broad incumbents
- Paperspace (DigitalOcean): Paperspace, acquired by DigitalOcean, offers on-demand GPU instances and notebooks aimed at developers and small teams. It competes with Runpod on Cloud GPUs / Pods and is positioned as a broader, more general cloud incumbent in the GPU segment.
- Amazon Web Services (EC2 GPU instances / Bedrock): AWS provides GPU instances (P4/P5 families with H100/H200), Bedrock for managed foundation model inference, and SageMaker for ML training. Runpod explicitly positions against AWS on developer experience and per-second pricing, making it the primary broad incumbent competitor.
Market position
Strengths5 records
Weaknesses4 records
Competitive moat7 records
Key risks6 records
Key highlights7 records
Customer concentration
Runpod social profiles
Digital presenceRunpod compliance and trust
Trust signalCompliance3 records
Runpod financial estimates
Financial estimateRevenue estimate
Valuation estimate
Runpod leadership team
Management profileNumber of profiles
Profiles3 records
Runpod funding detail
Funding detailFunding overview
Funding rounds5 records
Investors4 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Runpod 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 Runpod
What does Runpod do?
Runpod is a software-orchestrated AI Developer Cloud that aggregates GPU capacity across 28 partner datacenters and exposes it as three primitives: Pods (dedicated single-node GPU instances), Serverless (auto-scaling API endpoints), and Clusters (multi-node InfiniBand/RoCE v2 connected GPU clusters). The platform supports 30+ GPU SKUs across 31 global regions and is used by developers for AI training, fine-tuning, and production inference.
Is Runpod a public or private company?
Runpod is a private company. It is classified as venture growth investor backed and is currently operating.
When was Runpod founded?
Runpod was founded in 2022. It employs 101 to 250 people.
Where is Runpod based?
Runpod is headquartered in San Francisco, United States, in the North America region.
How does Runpod make money?
Five revenue lines are on record. GPU Pod rentals (Cloud GPUs) is the primary driver. The others are serverless GPU endpoints, clusters and Reserved Clusters, public Endpoints (Hub API) and storage and Network Volumes.
Who are Runpod's main competitors?
Direct peers on record are CoreWeave, Lambda Labs, Crusoe Energy, Modal Labs, Baseten, fal.ai, Together AI and Vast AI. Broad incumbents are Paperspace (DigitalOcean) and Amazon Web Services (EC2 GPU instances / Bedrock).
Does Runpod have an API?
Yes. Runpod offers a full public API for managing GPU instances and serverless endpoints programmatically, plus CLI and SDKs. The Runpod SDK is available in Python, JavaScript, and Go for programmatic API access to Pods, Serverless endpoints, and Clusters. A separate Flash CLI handles deployment and CI/CD integration. Authentication via RUNPOD_API_KEY or flash login OAuth flow. Public documentation at docs.runpod.io covers endpoints, workers, serverless configurations, GitHub integration, and storage. Developer documentation is at docs.runpod.io.