Hyperfluid is software. Twenty self-service services in a single console. You install it inside your perimeter and your teams serve themselves.
The overview: what runs across your organization, service by service.
Twenty interconnected services in a single console. Your teams consume them on their own with no ticket, within the limits you set.
A console, an hfctl CLI, a Terraform provider and GitOps CRDs: four ways to drive the same declared state.
Hover a service: what it does shows up here.
The same architecture everywhere. Only the nature of the machines changes.
You add workers and GPUs as your needs grow.
Nodes are your physical machines, or your VMs on the hypervisor you already run.
Bare metal, VMware, Vates, OpenStack, Proxmox
Nodes are instances from your own account, VMs or bare metal. The contract and the bill stay in your name.
AWS, GCP, Azure, OVH, Scaleway
Nodes are your machines, with no route to the outside. Images arrive through your mirrored registry.
Mirrored image registry, no egress
Change target: the console, the catalog and your rules do not move. Only the outbound connectors depend on what your network allows.
A ticket to the data team, one more ETL to write and then maintain, and a delay nobody controls.
Both sources answer one SQL statement. The result is exposed over SQL, GraphQL, REST or MCP, and BI tools and agents plug straight into it.
Services involvedFederated lakehouse, Big data analytics, Data APIs & MCP, S3 object storage
An integration engagement to scope, or nobody to take it on, and the pile stays dead.
A pre-wired pipeline reads the documents, classifies them and files them. They become an Iceberg table, queryable like the others.
Services involvedKnowledge base building, Connectors & ingestion, Federated lakehouse
The case goes to committee, and comes back as a refusal or as a pilot on fake data.
The model runs on your own GPUs behind an OpenAI-compatible endpoint. The agent reads, read-only, the perimeter opened when it was created.
Services involvedPrivate AI platform, AI Agents, Data APIs & MCP
As many tickets as there are pieces, each with its own queue, before anything ships.
A container with its URL and HTTPS, a managed database, a cache, an event bus, a private registry and a Git forge with its runners. Created by the team, the day it needs them.
Services involvedContainers as a Service, Managed PostgreSQL, Redis-compatible cache, Managed Kafka streaming, Private container registry, Git forge & managed CI
An admin cockpit separate from the console. It is enabled at install time and is not open by default.
You know whether you can still say yes.
Nothing opens without you.
You know what the fleet costs, and what it earns.
The cluster refuses on its own what has no business running.
The cockpit has its own access control. An organization token does not get in, and an admin token does not get into any organization. The separation is in the code, not in the procedure.
And every admin action on an organization goes to that organization's audit log, attributed to the account that performed it.
The same ergonomics as the hyperscalers, on your own infrastructure. No data leaves your perimeter.
Since the Broadcom acquisition, the licence line has changed order of magnitude. The layer above it is available today: containers, storage, databases, IAM, monitoring.
Virtualization itself arrives in early 2027.
A yearly flat fee known in advance, hardware you amortize, consumption visible per resource in the console. No egress fees, no end-of-month surprise.
Your models run on your own GPUs behind an OpenAI-compatible endpoint. Each agent gets a read-only service account, limited to the data you open for it.
Cloud Act, FISA 702: infrastructure operated by a US provider stays exposed to a foreign jurisdiction, wherever it is hosted. An internal cloud does not.
A "sovereign" offering built on US technology remains subject to the Cloud Act: the provider keeps the key. Hyperfluid is built on open standards and auditable software, operated in Europe by a European team. Sovereign from infrastructure to AI.
Read the manifestoAn envelope, a licence, or a share of your sales. Never a request.
Startups and SMEs
Start your product on a managed European platform, with no infrastructure to provide.
Deploy on your infrastructure or your own cloud account (BYOCloud), keep total control.
You start at no cost, you only pay on your sales.
Building your product on Hyperfluid, or deploying it at your clients? Software vendor Integrator
Preferential pricing for the social and solidarity economy on request.
See full pricing detailsTwenty services, with AI running through them. Your models run on your own GPUs.
Your VMs run next to your containers. The old ones are taken over as they are.
The turnkey rack, GPUs included, rented on your side. You plug in power and network.
What an internal cloud is, and what it is not.
An internal cloud: 20 self-service services in a single console, installed in your datacenter, your cloud account or an air-gapped network.
A complete cloud platform installed inside an organization's own perimeter, where teams enable services themselves, with no ticket and no integration project.
No. We do not run a datacenter for you. Hyperfluid is software you install, and the contract, the bill and the quotas of the infrastructure stay in your name.
Servers, or the virtualization you already run: bare metal, VMware, OpenStack, Proxmox. Hyperfluid brings its own orchestration layer. Sizing is scoped with you.
Your teams, with 20 days of professional services at install time and priority support. On the shared offer, our team operates it.
No. You bring your existing containers, Hyperfluid handles deployment, networking, HTTPS and secret injection.
Ceph and the S3 protocol, PostgreSQL, Valkey, Kafka, Trino, Apache Iceberg, Forgejo, vLLM, OpenID Connect. No proprietary format.
Yes. No outbound connection is required to run the platform.
The shared offer is free for 6 months, then starts at €200/month based on the resource envelope. Self-hosted is an annual licence. Onde Porteuse is revenue-sharing for software vendors.
Yes. You are not reselling components to assemble: you deploy a finished platform, described in GitOps, reproducible identically at every client.
Tell us where you want to install it and what your teams need to be able to create. We answer with a scope, a sizing and a price.
Just looking for platform access? Request it directly on console.hyperfluid.cloud/signup
We have received your request and will get back to you soon.
The form could not be sent automatically.
You can contact us directly by email: contact@nudibranches.tech