Velda
Velda provides a serverless GPU compute platform that lets ML, AI, and HPC teams launch training and batch jobs via a single CLI command prefix (vrun), eliminating container and Kubernetes overhead through an environment-first architecture.
- Company typePrivate
- Founded2025
- HeadquartersMountain View, United States
- Headcount1–10
- GTM typeB2B and B2C
- OfferingSoftware
What Velda does
Velda is a private, Mountain View-based developer infrastructure company that provides a serverless GPU computing platform targeted at machine learning, AI, data engineering, and HPC teams. The platform is built around an environment-first architecture that replaces container image packaging with a remotely mounted root filesystem (NFS/FUSE) and on-demand, content-hash-keyed file fetching, so that workloads can launch in seconds without Dockerfiles, image pushes, or Kubernetes manifests. Interaction is driven by a single CLI prefix (`vrun` for interactive execution, `vbatch` for queued batch jobs) that abstracts cloud resource provisioning and supports advanced features such as gang scheduling, network-topology-aware instance placement, and autoscaling compute pools.
The product is delivered through three channels: a managed self-serve cloud at cloud.velda.io (with free monthly credits and in-browser VS Code), an open-source self-hosted distribution on GitHub, and an Enterprise tier offering bring-your-own-cloud deployment with SSO, RBAC, and premium support. Revenue is generated primarily through usage-based, per-second GPU billing across A100, H100, H200, and B200 SKUs (standard H100-1 at $4.40/hr, H100-8 at $35.20/hr; spot pricing substantially below standard), with no subscription fees or seat minimums on the cloud tier. Specialized product surfaces target robotics and physical AI workloads, DAG-based workflow orchestration, and HPC cluster scheduling positioned as a Slurm alternative.
Founded in 2025 by Chuan Qiu and operating with 1-10 employees on a bootstrapped capital base, Velda transitioned from proof-of-concept (as documented in its September 2025 security notice) to general availability of the serverless GPU platform in May 2026. The company supports multi-cloud deployments across AWS, Google Cloud, Azure, and Nebius, and has built an AnyCloud partner network that adds Crusoe, Verda, and Massed Compute for additional GPU supply. One publicly disclosed customer case study (Kumo.ai) reports large improvements in GPU utilization and experiment throughput, but no revenue, headcount growth, or external funding figures have been disclosed.
Velda firmographics
Firmographics- Name
- Velda
- Legal name
- Velda, Inc.
- Website
- https://velda.io
- Company type
- Private
- Founded year
- 2025
- Operating status
- Operating
- Headcount range
- 1–10 employees
- Short description
- Velda provides a serverless GPU compute platform that lets ML, AI, and HPC teams launch training and batch jobs via a single CLI command prefix (vrun), eliminating container and Kubernetes overhead through an environment-first architecture.
- Ownership category
- akta.pro rank
Velda industry classification
Industry- Product category
- Serverless GPU Cloud Computing
- NAICS
- Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (51821)
- SIC
- Services-Computer Programming, Data Processing, Etc. (7370), Services-Computer Processing & Data Preparation (7374)
- akta.pro primary industry
- AI Compute Cloud & GPU-as-a-Service (HDAAAAAK)
- akta.pro secondary industries
- AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers) (HDAAAAAG), GPU/Accelerator Compute Servers (HDABADAC)
Keywords
Where Velda is headquartered
LocationHeadquarters
- HQ city
- Mountain View
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
Velda business model
Business model- GTM type
- B2B and B2C
- Offering type
- Software
- Cost components
- Personnel, Infrastructure, Technology or R&D, Marketing or Sales, Operations
Revenue model
- GPU Compute (Usage-Based): Pay-per-second billing for active GPU compute time. Users only pay when workloads are running with no idle charges, subscription fees, or seat minimums. Supports various GPU types including A100, H100, H200, and B200. Prices range from $1.65/hour for A100-1a to $48.00/hour for B200-8a (standard pools), with discounted spot instances available.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Usage-based | Pay-as-you-go | Standard Pools - Nebius Provider |
| Usage-based | Pay-as-you-go | AnyCloud Pools - Partner Network (Nebius, Crusoe, Verda, Massed Compute) |
| Freemium | Pay-as-you-go | Self-Hosted Open Source |
| Subscription | Multi-year contract | Enterprise / BYOC |
Go-to-market motion2 records
Distribution channels3 records
Marketing channels6 records
Velda product offering
Product offeringCore offering
Velda provides a serverless GPU computing platform that lets developers launch distributed AI/ML training, batch inference, and data-processing jobs from any environment by prefixing commands with `vrun`, eliminating the need for Dockerfiles, Kubernetes manifests, or container registries. The platform supports A100, H100, H200, and B200 GPUs across multi-cloud pools (AWS, GCP, Azure, Nebius, Crusoe, Verda, Massed Compute) with per-second usage-based billing, and exposes an open-source self-hosted version plus an enterprise tier with SSO/SAML, RBAC, and BYOC.
Product overview
Velda is a dev-friendly workload orchestration platform for AI, ML, and data-intensive tasks, built around a serverless GPU computing paradigm. The platform operates as a unified system with a core serverless GPU platform at its center, supplemented by specialized capabilities for robotics and physical AI, workflow orchestration, and HPC cluster management as an alternative to Slurm. Interaction is primarily through the vrun/vbatch CLI commands, which allow developers to run workloads on remote compute without container packaging. Deployment options span Velda Cloud (managed offering), Self-Hosted/Open Source (free), and Enterprise (paid, with SSO/RBAC). The compute infrastructure relies on autoscaling pools that abstract GPU allocation across multi-cloud providers.
Differentiator
Problem solved
Functional benefit
Brands
- Velda Cloud: Managed cloud platform with instant VSCode and GPU access, plus free monthly credit for individual and small teams.
- Velda Enterprise
Products and services
- Serverless GPU Platform Velda's flagship serverless GPU computing platform that enables ML teams to launch distributed AI workloads and batch jobs directly from development environments using a simple command prefix (vrun), without requiring container packaging or cluster provisioning. It supports A100, H100, H200, and B200 GPUs across multi-cloud pools with per-second billing.
- Velda Cloud Velda's managed cloud service offering instant VS Code browser access and GPU compute with free monthly credits, designed for individuals and small teams who want to start using the platform without managing their own infrastructure.
- Self-Hosted Velda (Open Source) Open-source, self-hosted Velda deployment option that provides full infrastructure control with no license fee for organizations running their own GPU clusters on their preferred cloud or on-premises environment.
- Robotics & Physical AI Product vertical on the Velda platform for GPU-accelerated development and deployment of robotic systems and physical AI applications.
- Workflow Workflow orchestration capabilities supporting DAG-based job pipelines, fan-out patterns, and complex multi-step ML workflows with dependency management on the Velda platform.
- Velda Slurm Alternative Modern HPC cluster management solution positioned as a cloud-native alternative to Slurm, featuring environment customization, cluster autoscaling, multi-cloud scheduling, and interactive development capabilities.
Quantifiable outcome
- Job startup reduced from 4-10 minutes (container-based) to seconds
- +4 more outcomes
Companies that use Velda
Customer profileNamed customers1 record
Segments4 records
Ideal customer profiles2 records
Velda technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
Integration3 records
AI capability2 records
Feature8 records
Velda partnerships and signals
Strategic signalPartnerships
Seven partnerships are on record, tiered core.
- NebiuscoreNebius is listed as the preferred provider for Standard Pools and is also part of the AnyCloud partner network. Provides H100, H200, and L40s GPU instances through Velda's platform.
- CrusoecorePart of the AnyCloud pool partner network providing additional GPU options and availability beyond the standard Nebius provider.
- VerdacorePart of the AnyCloud pool partner network for GPU compute access.
- Massed ComputecorePart of the AnyCloud pool partner network for GPU compute access.
- AWS (Amazon Web Services)coreSupported as a cloud backend for Velda deployments. Secure cloud infrastructure provider for container isolation and development environments.
- Google CloudcoreSupported as a cloud backend for Velda deployments. Multi-cloud support includes GCP with network-topology-aware instance allocation.
- AzurecoreSupported as a cloud backend for Velda deployments. Part of the secure cloud infrastructure providers.
Scale indicators1 record
Recent moves6 records
Expansion highlights6 records
Velda competitors and assessment
Company assessmentDirect peers
- Replicate: Replicate runs ML models in the cloud via a simple API and serves a similar audience of ML engineers who want to skip infrastructure. Its usage-based billing and emphasis on instant model deployment parallel Velda's serverless positioning.
- Anyscale: Anyscale (built on Ray) provides managed Ray-based compute for AI, including autoscaling clusters. Velda explicitly integrates Ray and offers Ray AutoScaled clusters as an example, so the two compete in the same Ray-powered distributed training niche.
- Vast.ai: Vast.ai is a marketplace for renting distributed GPU compute at consumer prices. It targets cost-sensitive ML practitioners and competes with Velda's usage-based, multi-cloud pricing on similar workloads (training, inference).
- RunPod: RunPod is a cloud GPU platform offering on-demand and serverless endpoints for AI/ML/inference. It competes head-on with Velda's pay-per-second GPU model across H100/A100 tiers and targets a similar self-serve ML developer base.
- Modal Labs: Modal offers serverless GPU compute for AI/ML with code-first ergonomics (decorator-based). Like Velda, it abstracts infrastructure so developers can ship ML workloads without managing containers; both target ML engineers with usage-based GPU billing and IDE-style developer experience.
Broad incumbents
- CoreWeave: CoreWeave is a large, well-capitalized GPU cloud built for AI workloads, offering bare-metal and Kubernetes-based GPU clusters. It is a broader incumbent in the same category as Velda but with substantially larger infrastructure footprint and customer base.
- Amazon Web Services (EC2/Batch/SageMaker): AWS provides GPU compute via EC2 P/G instances, SageMaker, and Batch. As the largest cloud provider and a named Velda partner, it is a broad incumbent whose serverless ML offerings (e.g., SageMaker, AWS Batch) compete for the same workloads, and whose pricing/marketing Velda must position against.
- Lambda Labs: Lambda is an established AI cloud provider with on-demand GPU instances, clusters, and workstations. It competes with Velda on raw GPU access and an ML-developer-friendly stack, but with a more traditional IaaS posture than Velda's environment-first abstraction.
- Together AI: Together AI operates an AI cloud combining inference, training, and model APIs on GPU infrastructure. It is adjacent/overlapping with Velda on training and inference use cases, with broader model-serving product surface area.
Emerging players
- Beam.cloud: Beam offers serverless GPU for ML inference and training with code-first developer experience. It is an emerging niche player with similar product philosophy to Velda, serving AI developers who want infrastructure abstracted away.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights7 records
Customer concentration
Velda social profiles
Digital presenceVelda compliance and trust
Trust signalCompliance3 records
Velda financial estimates
Financial estimateRevenue estimate
Valuation estimate
Velda leadership team
Management profileNumber of profiles
Profiles1 record
Velda funding detail
Funding detailFunding overview
Funding rounds
Investors
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Velda 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 Velda
What does Velda do?
Velda provides a serverless GPU computing platform that lets developers launch distributed AI/ML training, batch inference, and data-processing jobs from any environment by prefixing commands with `vrun`, eliminating the need for Dockerfiles, Kubernetes manifests, or container registries. The platform supports A100, H100, H200, and B200 GPUs across multi-cloud pools (AWS, GCP, Azure, Nebius, Crusoe, Verda, Massed Compute) with per-second usage-based billing, and exposes an open-source self-hosted version plus an enterprise tier with SSO/SAML, RBAC, and BYOC.
Is Velda a public or private company?
Velda is a private company. It is classified as founder individual operated bootstrapped and is currently operating.
When was Velda founded?
Velda was founded in 2025. It employs 1 to 10 people.
Where is Velda based?
Velda is headquartered in Mountain View, United States, in the North America region.
How does Velda make money?
One revenue line is on record: GPU Compute (Usage-Based).
Who are Velda's main competitors?
Direct peers on record are Replicate, Anyscale, Vast.ai, RunPod and Modal Labs. Broad incumbents are CoreWeave, Amazon Web Services (EC2/Batch/SageMaker), Lambda Labs and Together AI. Beam.cloud is listed as an emerging player.
Does Velda have an API?
No public API is recorded for Velda.
What industry is Velda in?
Velda's product category is Serverless GPU Cloud Computing. Its primary akta.pro industry code is HDAAAAAK, AI Compute Cloud & GPU-as-a-Service, with a secondary code of HDAAAAAG, AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers). Its NAICS code is 5182 and its SIC code is 7370.