Unshackle your AI infra with Cloudexe's HW/SW disaggregation tech

Cloudexe tech decouples GPU hardware from the execution environment, marrying the economics of on-prem/neo-cloud GPUs with maturity of hyperscaler infrastructure.

Now you can use GPUs from anywhere, be it from other hyperscalers, neo-clouds, or on-prem clusters, right inside your existing infrastructure, be it on hyperscaler public cloud or on-prem, with zero integration work

Sign up for free trial

Technology

Cloudexe brings together innovative technology, an ecosystem and a business model. Cloudexe's data plane decouples software environment from physical hardware, allowing movement of workloads out of hyperscaler cloud to other providers with zero integration work. Cloudexe's control plane aka realtime workload matchmaking finds the right GPU provider for any workload just-in-time, enabling new ways of paying for compute. Cloudexe's globe-spanning partner network brings plentiful and economical GPU supply anywhere.

Using Cloudexe is easy.

  1. Download and setup cloudexe client.
  2. Define your workload GPU needs in a config.json.
  3. Launch your workload command with cloudexe as the launcher.

Zero-integration UX

Cloudexe does away with the chores of remote login, software install, file updates, dataset uploads/downloads, model weights uploads/downloads, code syncing, OS updates, credentials management, port forwards, network tunneling, server lifecycle management.

Workload executes in the context of the client machine on which it is launched, including the filesystem, network, devices, IPC, etc.

That means less time wrangling infra, and more time for AI.

See it in action

Demos

Cloudexe works seamlessly for any GPU AI/ML workload.

See more demos.

Platform features

Cloudexe platform has following salient features. The platform can be leveraged in multiple ways.

  • Cloudexe Inc. has SOC-2 TypeII certification
  • Connection with remote GPU server is protected with SOTA encryption technology
  • Cloudexe technology is vendor/arch agnostic and works will for NVIDIA, AMD and others.
  • The technology works well for training, fine-tuning, inference, and classic ML workloads.
  • The technology integrates seamlessly with Kubernetes and Slurm.
  • The technology is Enterprise ready with features like API integrations, SSO, RBAC, HA, Dashboards/Alerts and more.

Products

Developer Cloud

  • GPU instances accessible via SSH and browser
  • No-setup, no-changes access to variety of GPUs from the same instance
  • 50%-80% cost savings in a no-commitment, all-inclusive GPU subscription
  • Best for R&D teams with unpredictable computing needs

Virtual GPU Cloud

  • Layer to consolidate distributed GPUs into a virtual pool
  • No-setup, no integration way to augment GPU fleet just-in-time
  • 50%-80% cost savings and 1.5x to 2x utilization improvement
  • Best for customers operating complex at-scale production infrastructure

Pricing

GPU Pay-per-use 3-month commit Location
H100-80GB-SXM $2.00/hr $1.00/hr US-West, US-East, CA-Central
H200-144GB $3.00/hr $2.00/hr US-West, US-East, CA-Central
A40 48GB $0.80/hr $0.50/hr US-West, US-East, CA-Central

We continuously add more supplier-partners to our network. We also offer generous trial periods, and free credits for educational and research institutions. Just ask. Please reach us at info@cloudexe.tech for details.

Sign up for free trial

About us

We are a seasoned team of elite engineers and entrepreneurs with deep experience in building GPU cloud infrastructure at companies like Intel, NVIDIA, Google and others. We are based in Silicon Valley.

Investors

Contact

Reach us at info@cloudexe.tech for corporate inquiries.