dstack
dstack is an open-source orchestration platform that standardizes GPU workload management across GPU clouds, Kubernetes, and on-prem clusters, serving AI/ML engineers, researchers, and platform teams with fleet, training, and inference abstractions plus a hosted GPU marketplace.
- Company typePrivate
- Founded2022
- HeadquartersMunich, Germany
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What dstack does
dstack is an open-source, AI-native orchestration layer headquartered in Munich with a US corporate entity (dstack Inc.) that standardizes how GPU workloads run across heterogeneous AI compute environments. Its core platform unifies fleet management, dev environments, training/batch tasks, and inference services through four abstractions — Fleets, Dev Environments, Tasks, and Services — backed by a YAML-based configuration, CLI, UI, and HTTP API. The platform operates across 15+ GPU cloud providers (AWS, GCP, Lambda, Nebius, Crusoe, RunPod, Azure, Verda, DigitalOcean, JarvisLabs, CloudRift), Kubernetes clusters, and on-prem SSH-accessible infrastructure, and supports NVIDIA, AMD, Tenstorrent, and TPU accelerators with ML-native features such as Prefill-Decode disaggregation via NVIDIA Dynamo and SGLang/Shepherd Model Gateway.
The company operates a hybrid revenue model anchored by a free open-source self-hosted core installed via pip or uv, a usage-based hosted GPU marketplace called dstack Sky (H100 at $1.90-$5.99/hr, B200 at $4.00-$9.00/hr), and an enterprise self-hosted tier with SSO, air-gapped setup, and dedicated support sold via direct sales. Its primary buyers are AI/ML engineers, AI researchers, platform/infrastructure teams, and AI-building enterprises; named customers include Electronic Arts (100+ AI projects), Graphsignal, Toffee (reporting 2-3x GPU cost reduction and 50%+ orchestration overhead cut), ETRI, Metamorphic, and Hot Aisle. The company was founded in 2022, has 1-10 employees, holds 2.2k GitHub stars, and is led by CEO and founder Andrey Cheptsov, with seed funding from Rheingau Founders and Tensor Ventures.
dstack firmographics
Firmographics- Name
- dstack
- Legal name
- dstack Inc.
- Website
- https://dstack.ai
- Company type
- Private
- Founded year
- 2022
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- dstack is an open-source orchestration platform that standardizes GPU workload management across GPU clouds, Kubernetes, and on-prem clusters, serving AI/ML engineers, researchers, and platform teams with fleet, training, and inference abstractions plus a hosted GPU marketplace.
- Ownership category
- akta.pro rank
dstack industry classification
Industry- Product category
- AI Infrastructure Orchestration
- NAICS
- Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182)
- SIC
- Services-Prepackaged Software (7372)
- akta.pro primary industry
- AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers) (HDAAAAAG)
- akta.pro secondary industry
- Model Hosting, Serving & Inference Platforms (HDAAACAB)
Keywords
Where dstack is headquartered
LocationHeadquarters
- HQ city
- Munich
- HQ country
- Germany
- HQ region
- Europe
Markets served
dstack business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Infrastructure, Marketing or Sales, Operations
Revenue model
- Open-source (Community): Free open-source version with self-hosting capability. Users install via pip/uv and run dstack server on their own infrastructure. Revenue is indirect through community growth, contributions, and ecosystem building.
- dstack Sky (GPU Marketplace): Hosted GPU marketplace providing access to GPU compute without requiring users to have their own cloud accounts. Provides a hosted and managed option alongside user's own cloud accounts and on-prem clusters.
- Enterprise (Self-hosted): Paid enterprise tier offering SSO, air-gapped setup, dedicated support, and enterprise-grade features for self-hosted deployment within customer's infrastructure.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Others | Open-source free tier |
| Subscription | Annual | Enterprise tier |
| Usage-based | Pay-as-you-go | dstack Sky GPU marketplace |
Go-to-market motion2 records
Distribution channels3 records
Marketing channels6 records
dstack product offering
Product offeringCore offering
dstack is an open-source, AI-native orchestration layer that standardizes how GPU workloads run across heterogeneous AI compute — multiple GPU clouds (AWS, GCP, Lambda, Nebius, Crusoe, RunPod, Azure, Verda, DigitalOcean, JarvisLabs, CloudRift), Kubernetes clusters, and on-premises clusters — across NVIDIA, AMD, Tenstorrent, and TPU accelerators. It exposes four primary abstractions — Fleets, Dev Environments, Tasks, and Services — managed through a CLI, UI, and HTTP API, with YAML-based configuration and a container-first architecture. Monetization is delivered through the open-source core, the dstack Sky hosted GPU marketplace, and the dstack Enterprise self-hosted subscription tier.
Product overview
dstack is an open-source, AI-native orchestration platform for heterogeneous AI compute. The core platform (dstack) unifies GPU provisioning and workload management across GPU clouds (AWS, GCP, Lambda, Nebius, Crusoe, RunPod, Azure, DigitalOcean, and others), Kubernetes, and on-premises clusters through four primary abstractions: Fleets (infrastructure provisioning and management), Dev Environments (interactive development), Tasks (training and batch processing), and Services (model inference deployment). dstack Sky provides a hosted GPU marketplace, while dstack Enterprise offers self-hosted deployment with SSO and air-gapped setup. The platform integrates with major serving frameworks (SGLang, vLLM, TensorRT-LLM, NVIDIA NIM) and training tools (TRL, Axolotl), and enables AI agents to autonomously orchestrate infrastructure through fine-grained primitives and skill definitions.
Differentiator
Problem solved
Functional benefit
Brands
- dstack Sky: Hosted GPU marketplace where users can access compute, or bring their own clouds. Managed hosting service.
- dstack Enterprise
Products and services
- dstack (Core Platform) Open-source AI-native orchestration platform that standardizes how GPU workloads run across GPU clouds, Kubernetes clusters, and on-prem clusters. Provides CLI, UI, and HTTP API for managing Fleets, Dev Environments, Tasks, Services, and Volumes. Targeted at ML engineers, AI researchers, and platform teams building and operating AI workloads.
- dstack Sky Hosted GPU marketplace providing pay-as-you-go access to GPU compute across multiple providers without requiring users to have their own cloud accounts. Targeted at startups, individual developers, and teams needing quick GPU access.
- dstack Enterprise Self-hosted deployment of the dstack orchestration platform with SSO, air-gapped setup, and dedicated enterprise support. Targeted at mid-market and large enterprises that require on-premises or dedicated infrastructure deployments.
Quantifiable outcome
- 50%+ reduction in GPU-cloud orchestration overhead
- +2 more outcomes
Companies that use dstack
Customer profileNamed customers5 records
Segments4 records
Ideal customer profiles3 records
dstack technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration13 records
AI capability4 records
Feature7 records
dstack partnerships and signals
Strategic signalPartnerships
Four partnerships are on record, tiered core.
- DigitalOceancoredstack is integrated into the DigitalOcean AI Ecosystem as a development framework integration alongside LangChain and LiteLLM. DigitalOcean's AI Partner Program includes collaborations with AMD, NVIDIA, OpenAI, DeepSeek, Meta, and Mistral. The ecosystem targets startups, developers, and enterprises building AI applications.
- Oracle Cloud Infrastructure (OCI)coredstack integrates with Oracle Cloud Infrastructure, enabling the use of open-source LLMs (Meta's Llama 2 and Llama 3) on OCI. dstack simplifies training and deployment of LLMs in the cloud, providing enterprise control and customization.
- AMDcoreAMD is a hardware provider in the DigitalOcean AI Ecosystem alongside NVIDIA. dstack natively supports AMD GPUs (MI300X, MI30x) with ROCm 6/7 support, making AMD a credible alternative to NVIDIA for teams willing to standardize on ROCm images.
- NVIDIAcoreNVIDIA is a hardware provider in the DigitalOcean AI Ecosystem. dstack has native support for NVIDIA GPUs and integrations with NVIDIA Dynamo for PD disaggregation. dstack was featured at NVIDIA GTC 2025.
Scale indicators2 records
Recent moves6 records
Expansion highlights5 records
dstack competitors and assessment
Company assessmentBroad incumbents
- Anyscale (Ray): Commercial Ray platform for distributed AI compute and orchestration; broader in scope than dstack (general distributed compute) but addresses overlapping distributed training and serving workloads across heterogeneous infrastructure.
- Domino Data Lab: Enterprise MLOps platform with multi-cloud GPU orchestration for training and inference; competes for the same enterprise platform-team buyer persona dstack targets with self-hosted enterprise and Fleets capabilities.
- Weights & Biases (W&B): ML experiment tracking and platform increasingly extending into compute orchestration and inference; competes for AI researcher and platform-team mindshare with dstack on multi-cloud GPU workflows.
Emerging players
- Lightning AI: AI development platform with built-in multi-cloud compute orchestration, dev environments, and training/inference workflows; competes with dstack on the same developer-and-ML-engineer persona.
- Modal Labs: Serverless AI compute platform offering on-demand GPU orchestration and inference deployment; competes with dstack's hosted Sky marketplace and Services abstraction for developers wanting GPU access without cloud accounts.
Direct peers
- Determined AI: AI infrastructure platform for training and fine-tuning models on multi-cloud GPU clusters; overlaps with dstack's Tasks and Fleets abstractions for distributed ML workload orchestration.
- Tensorwave: AMD-focused GPU cloud and orchestration provider; overlaps with dstack's multi-cloud GPU scheduling and AMD ROCm support for AI training and inference workloads.
- SkyPilot: Open-source multi-cloud AI workload orchestrator spun out of UC Berkeley; directly competes with dstack on unified scheduling of GPU training and inference across clouds and Kubernetes with a similar developer-led, YAML-driven UX.
- Run:ai: GPU orchestration and scheduling platform acquired by NVIDIA; competes with dstack's Fleets for managing heterogeneous accelerator pools, multi-tenancy, and utilization, with tighter native Kubernetes integration.
Others
- CoreWeave: Hyperscale GPU cloud provider that is both a backend partner and ecosystem adjacent to dstack; comparable in that both target AI compute workloads across NVIDIA and other accelerators but CoreWeave sells bare infrastructure rather than orchestration.
Market position
Strengths4 records
Weaknesses4 records
Competitive moat5 records
Key risks5 records
Key highlights6 records
Customer concentration
dstack social profiles
Digital presencedstack financial estimates
Financial estimateRevenue estimate
Valuation estimate
dstack leadership team
Management profileNumber of profiles
Profiles1 record
dstack funding detail
Funding detailFunding overview
Funding rounds2 records
Investors2 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
dstack 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 dstack
What does dstack do?
dstack is an open-source, AI-native orchestration layer that standardizes how GPU workloads run across heterogeneous AI compute — multiple GPU clouds (AWS, GCP, Lambda, Nebius, Crusoe, RunPod, Azure, Verda, DigitalOcean, JarvisLabs, CloudRift), Kubernetes clusters, and on-premises clusters — across NVIDIA, AMD, Tenstorrent, and TPU accelerators. It exposes four primary abstractions — Fleets, Dev Environments, Tasks, and Services — managed through a CLI, UI, and HTTP API, with YAML-based configuration and a container-first architecture. Monetization is delivered through the open-source core, the dstack Sky hosted GPU marketplace, and the dstack Enterprise self-hosted subscription tier.
Is dstack a public or private company?
dstack is a private company. It is classified as venture growth investor backed and is currently operating.
When was dstack founded?
dstack was founded in 2022. It employs 11 to 50 people.
Where is dstack based?
dstack is headquartered in Munich, Germany, in the Europe region.
How does dstack make money?
Three revenue lines are on record. Open-source (Community) is the primary driver. The others are dstack Sky (GPU Marketplace) and enterprise (Self-hosted).
Who are dstack's main competitors?
Broad incumbents on record are Anyscale (Ray), Domino Data Lab and Weights & Biases (W&B). Emerging players are Lightning AI and Modal Labs. Direct peers are Determined AI, Tensorwave, SkyPilot and Run:ai. CoreWeave is listed as an others.
Does dstack have an API?
Yes. dstack provides an HTTP API for programmatic orchestration of GPU workloads, covering users, projects, backends, fleets, runs, gateways, volumes, logs, events, repos, files, secrets, exports, metrics, GPUs, templates, proxy, and authentication. The API is documented at https://dstack.ai/docs/reference/http/. Additionally, dstack offers a Python API plugin interface and a REST API plugin interface for extending functionality. Developer documentation is at dstack.ai/docs/reference/http.
What industry is dstack in?
dstack's product category is AI Infrastructure Orchestration. Its primary akta.pro industry code is HDAAAAAG, AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers), with a secondary code of HDAAACAB, Model Hosting, Serving & Inference Platforms. Its NAICS code is 5182 and its SIC code is 7372.