Recognized As A “Leader” in the ISG Private/Hybrid Cloud – Data Centre Services 2026. Read More →

UL Logo

Your private
AI factory.

UPC AI Factory helps enterprises train, fine-tune, deploy, and operate AI
models while keeping sensitive data, applications, and model assets under their
control.
Request a demo
Your data, your tenancy
Fixed, predictable cost
Everything included, operated for you
Meet your AI Factory guide
Core Cloud

Click a tab on the dashboard and I’ll explain that part of the Factory.

Why a private AI factory

AI shouldn’t cost you control —
or a bill you can’t predict

Most teams hit the same three walls when they try to run AI at scale. The
Factory is built to remove all three.
Your data leaves your boundary
Public AI services and shared clouds mean your sensitive data, prompts, embeddings, and fine-tuned models sit on someone else’s infrastructure — a problem for regulated and security-conscious teams.
In the Factory: models, vectors, inference, and data stay inside your own tenant vDC. Route outside only if you choose.
Costs you can’t predict
Consumption-metered GPU billing turns a training run or a traffic spike into a bill shock, and next quarter’s budget becomes guesswork — right when finance wants certainty.
In the Factory: a fixed quantity of always-on capacity for a fixed term, billed as one monthly charge. Your cost base is known for the life of the term.
Building it yourself is a project
GPU scheduling, MIG, high-speed fabric, storage, networking, security, disaster recovery, model serving — standing this up and keeping it healthy takes a platform team you may not have.
In the Factory: it’s turnkey and operated for you by UnitedLayer — every Platform and Service included, updates during the term at no extra charge.

Who it’s for

Built for two kinds of team
Whether you’re running AI for your own business or building an AI business
for your customers, the Factory is the foundation.
Run AI privately
Enterprises running AI in-house
For regulated and security-conscious organizations — financial services, healthcare, government, and any team that can’t send data to a public AI service. Get a private AI cloud where models, vectors, and data never leave your tenancy, at a cost you can budget for the term.
  • Data sovereignty, isolation, and compliance
  • In-tenant model serving, RAG, and fine-tuning
  • Predictable, fixed cost — no metered surprises
  • No platform team to hire; operated for you
Build an AI business
Providers monetizing GPU capacity
For service providers, MSPs, and neoclouds that want to sell AI and GPU capacity. Multi-tenancy, chargeback, rate-card modelling, and margin visibility are built in — so you can launch your own always-on or per-hour offering on top of a fixed, known cost base.
  • Multi-tenant isolation per customer or business unit
  • Chargeback, showback, and rate-card modelling
  • Margin visibility on every resold workload
  • Your commercial model, your prices, your brand
What’s in it for your team

Value for every seat at the table

The Factory isn’t one team’s tool — each role gets something concrete.
CIO / CTO
A defensible AI platform
A complete AI cloud you can stand up fast, at a cost you can defend to the board — without a multi-year build.
Platform & infrastructure
Turnkey, not from scratch
GPU scheduling, Kubernetes, storage, and networking delivered and operated — so you don’t build and babysit it yourself.
AI & ML teams
Ship models in-tenant
An OpenAI-compatible router, an open-weight registry, a vector DB, and GPU workbenches — ready to serve and build.
Security & compliance
Data never leaves
Zero Trust, per-tenant isolation, disaster recovery, and an immutable cyber vault — everything stays inside your boundary.
FinOps & finance
A cost you can plan
A known cost base for the term, plus chargeback, cost-per-million-tokens, and margin visibility where you resell.
Operations & SRE
The estate runs itself
Agentless discovery, full-stack and GPU observability, and AIOps with approval-gated remediation across every vDC.
What’s included
Everything your AI Factory needs
Four Platforms. Twenty-three Services. — the private cloud itself, its
protection, its operations, and the platform to serve and build models — is included in one inteligent platform.
UPC CORE CLOUD
Run AI with confidence
8 Services · all included
The multi-tenant private cloud: isolated vDCs, the GPU accelerator platform, and the compute, container, serverless, storage and network services built on both.
Tenancy · GPU · Compute · Container · Serverless · Storage · Network · Connect
UPC PROTECT
Protect AI without compromise
4 Services · all included
Security and data protection across the Factory and every tenant vDC.
Security · Disaster Recovery · Backup · Cyber vault
UPC OPERATIONS
Operate AI with intelligence
7 Services · all included
Running and understanding the estate: discovery, monitoring, incident handling, provisioning, cost and agentic automation.
Discovery · Observability · AIOps · Cloud Management · FinOps · AI Assistant · Agent Builder
UPC AI SERVICES
Build AI your way
4 Services · all included
Serving and building models inside the tenant boundary.
Router · Registry · Vector Database · Workbench
Architecture
Engineered for security,
built to scale
Modular by design, future-proof for mission-critical workloads—
where security, speed, and flexibility power trusted AI at scale.

AI Use Cases

AI Inference
Fine-Tuning
RAG
Agentic AI

Core Platform

AI Observability

  • Observe
  • Govern
  • Automate
  • Security

UPC Core Cloud

Multi-tenant private cloud foundation for accelerated AI workloads

Tenancy
GPU
Compute
Container
Serverless
Storage
Network
Connect

UPC Protect

Security and data protection across the Factory and every tenant vDC

Security
Backup
Disaster Recovery
Cyber Vault

UPC Operations

Discovery, monitoring, incident handling, cost and agentic automation

Discovery
Observability
AIOps
Cloud Management
FinOps
AI Assistant
Agent Builder
Performance Management
Agentic ITSM

UPC AI Services

Serving and building models inside the tenant boundary

Router
Registry
Vector
Workbench
Skills

Guardrails

  • Zero Trust
  • Encryption
  • Compliance
  • Data Residency

Infrastructure Foundation

Core
Protect
Operations
AI Platform
How it’s licensed
One licence.
Everything included.

The Factory is delivered as a private cloud against a fixed quantity of always-
on capacity, on a fixed term — billed as a monthly recurring charge, not
metered by consumption.

01
One licence, nothing gated
Every Platform and Service is included in the Factory licence. No editions, no feature-gating, no capability withheld for a higher tier.
02
Operated for you
Delivered under a single technology licence and operated by UnitedLayer for the term. You get the Factory — not a staffing problem.
03
Fixed, always-on capacity
A dedicated quantity of capacity reserved for you on a fixed term — always on, always yours, never a scramble for GPUs.
04
Predictable monthly charge
Billed as a monthly recurring charge, not by consumption. Your cost base is known for the life of the term.
05
Updates included
New releases and capabilities delivered during the term are included at no additional charge — the Factory improves under you.
06
Build your model on top
Use the Factory’s tooling to run your own always-on or per-hour commercial model — resell capacity with chargeback, rate-cards, and margin visibility.
Platform 1 · UPC Core Cloud

The multi-tenant private cloud

The vDC construct, the accelerator platform, and the compute, storage and
network services built on both. Hover any capability to see what it does.
TenancyUPC:AIF:CORE:TENANCYIncluded
Isolated virtual data centers
Isolated virtual data centers with allocated compute, memory, storage and accelerator resources, delegated administration and role-based access control — with isolation enforced across compute, storage, networking, identity and GPU.
Capabilities
Isolated vDCsAllocated compute & memoryAllocated storageAccelerator resourcesDelegated administrationRole-based access controlIsolation: compute/storageIsolation: network/identity/GPU
GPUUPC:AIF:CORE:GPUIncluded
A tenant-assignable accelerator platform
Physical accelerators aggregated into tenant-assignable resource pools with dedicated, vGPU and MIG partitioning, high-performance fabric attachment, and driver, firmware and health lifecycle maintained by UnitedLayer.
Capabilities
Tenant-assignable poolsDedicated GPUvGPU partitioningMIG partitioningHigh-performance fabricDriver & firmware lifecycleGPU health lifecycleFully managed
ComputeUPC:AIF:CORE:COMPUTEIncluded
Software-defined compute, HA by default
Software-Defined Compute with live migration, High-Availability clustering, affinity and anti-affinity placement, dynamic resource scheduling, and virtual-machine lifecycle and template management — physical-to-virtual consolidation on general compute, dedicated allocation on GPU hosts.
Capabilities
Software-defined computeLive migrationHA clusteringAffinity / anti-affinityDynamic resource schedulingVM lifecycle & templatesP2V consolidationDedicated GPU hosts
ContainerUPC:AIF:CORE:CONTAINERIncluded
Managed Kubernetes with GPU scheduling
Managed Kubernetes delivering dedicated per-tenant clusters, integrated private container registries, enterprise ingress controllers, and native service mesh — integrated with the GPU operator toolchain (e.g. NVIDIA GPU Operator) for fractional (vGPU, MIG) and dedicated physical GPU scheduling into AI, inference and fine-tuning workloads.
Capabilities
Managed KubernetesPer-tenant clustersPrivate registriesEnterprise ingressService meshGPU operator toolchainFractional GPU (vGPU/MIG)Dedicated GPU scheduling
ServerlessUPC:AIF:CORE:SERVERLESSIncluded
Event-driven, GPU-accelerated functions
Managed Serverless with an event-driven compute engine and automatic scaling — API endpoints, scheduled tasks and asynchronous event triggers, with dynamic GPU and fractional GPU acceleration for burst AI inference and automated data pipelines.
Capabilities
Event-driven engineAutomatic scalingAPI endpointsScheduled tasksAsync event triggersDynamic GPUFractional GPUBurst AI inference
StorageUPC:AIF:CORE:STORAGEIncluded
Software-defined performance storage
Software-Defined Performance Storage with built-in redundancy and High Availability — networked to Compute, Container and Serverless as virtual block devices, and presented to other hosts, including bare metal, as NFS and SMB/CIFS file shares.
Capabilities
SD performance storageBuilt-in redundancyHigh availabilityVirtual block devicesTo compute/container/serverlessNFS sharesSMB/CIFS sharesBare-metal access

Platform 2 · UPC Protect

Security and data protection
Zero Trust security, orchestrated disaster recovery, policy-driven backup, and
immutable recovery — across the Factory and every tenant vDC.
SecurityUPC:AIF:PROTECT:SECUREIncluded
Zero Trust at the vDC boundary
Zero Trust architecture, Identity and Access Management, Privileged Access Management, secrets vault, and certificate lifecycle enforcement at the vDC boundary.
Capabilities
Zero Trust architectureIdentity & Access ManagementPrivileged Access ManagementSecrets vaultCertificate lifecyclevDC-boundary enforcement
Disaster RecoveryUPC:AIF:PROTECT:DRAASIncluded
Orchestrated DR, three tiers
Disaster-recovery orchestration — policy-driven replication schedules, retention management, automated restore validation, and failover runbook orchestration — available in Gold, Silver and Bronze tiers, operating in conjunction with Cyber Vault and Backup where subscribed.
Capabilities
DR orchestrationPolicy-driven replicationRetention managementAutomated restore validationFailover runbooksGold / Silver / Bronze
BackupUPC:AIF:PROTECT:BACKUPIncluded
Point-in-time protection, policy-driven
Scheduled point-in-time protection at system, tenant vDC, virtual-machine and file-share level — policy-driven snapshot schedules with independent retention per policy, application-quiesced snapshots for supported guests, and export volumes for third-party backup software.
Capabilities
System-levelvDC & VM levelFile-share levelPolicy-driven schedulesIndependent retentionApp-quiesced snapshotsExport volumesPoint-in-time
Cyber VaultUPC:AIF:PROTECT:VAULTIncluded
Immutable recovery, admin-proof
Immutable retention for recovery data — snapshots marked immutable cannot be deleted by any user, including platform administrators, and unlock requests are subject to a mandatory delay. Vault copies are held in a separate facility under separate administrative credentials from production.
Capabilities
Immutable retentionUndeletable by anyoneAdmin-proofMandatory unlock delaySeparate facilitySeparate credentials
Disaster recovery tiers

Recovery objectives you can
commit to

Disaster Recovery is available in three tiers; the tier subscribed and the
recovery objectives committed are stated in the Service Order Form.
Tier
RPO
RTO
Gold
15 minutes
1 hour
Silver
60 minutes
4 hours
Bronze
4 hours
24 hours
Platform 3 · UPC Operations
Running and understanding the estate
Discovery, monitoring, incident handling, provisioning, cost and agentic automation — scoped to each tenant vDC.
DiscoveryUPC:AIF:OPS:DISCOVERYIncluded
Agentless discovery and CMDB sync
Agentless discovery across physical, network, storage, VMs, tenant vDCs and GPU fabric — with dynamic grouping, dependency mapping and CMDB synchronization.
Capabilities
Agentless discoveryPhysical / network / storageVMs & tenant vDCsGPU fabricDynamic groupingDependency mappingCMDB synchronization
ObservabilityUPC:AIF:OPS:OBSIncluded
Full-stack, GPU and inference telemetry
Network, server, VM, storage and database monitoring combined with GPU utilization, thermal, power and pool-capacity telemetry, plus inference token throughput and latency by model and workload — scoped to each tenant vDC.
Capabilities
Network/server/VMStorage & databaseGPU utilizationGPU thermal & powerPool capacityInference token throughputLatency by model/workloadPer-tenant vDC scope
AIOpsUPC:AIF:OPS:AIOPSIncluded
Correlate, RCA, remediate with approval
Event correlation, noise reduction, automated root-cause analysis, bi-directional ITSM/CMDB flow, runbooks, change risk and impact analysis, and approval-gated remediation.
Capabilities
Event correlationNoise reductionAutomated RCAITSM/CMDB bi-directionalRunbooksChange risk & impactApproval-gated remediation
Cloud ManagementUPC:AIF:OPS:CLOUDMGRIncluded
Self-service provisioning, governed
Self-service provisioning from approved blueprints with role-based access, approval workflow and expiry — placement informed by capacity, policy, performance and cost, with lifecycle to retirement across the private factory and connected public cloud.
Capabilities
Self-service provisioningApproved blueprintsRole-based accessApproval workflowExpiryCost/policy/perf placementLifecycle to retirementPrivate + public cloud
FinOpsUPC:AIF:OPS:FINOPSIncluded
Cost per token, chargeback, margin
Chargeback/showback, rightsizing, anomaly detection, budgets and forecasts — including GPU-hours and cost per million tokens, cost-to-serve, and private-versus-public inference comparison resolved to the tenant vDC, plus rate-card modelling and margin visibility where capacity is resold.
Capabilities
Chargeback / showbackRightsizingAnomaly detectionBudgets & forecastsGPU-hoursCost per million tokensCost-to-servePrivate-vs-public inferenceRate-card modellingMargin visibility
AI AssistantUPC:AIF:OPS:ASSISTANTIncluded
Ask your estate anything
Natural-language query of estate state from discovery, observability, incident and topology data — with contextual recommendations and next-best-action guidance.
Capabilities
Natural-language queryEstate stateFrom discovery & observabilityIncident & topologyContextual recommendationsNext-best-action
Agent BuilderUPC:AIF:OPS:AGENTBUILDERIncluded
Build your own operational agents
A visual builder to author, test, publish and version customer-specific operational agents and multi-step workflows — executing under the same policy binding and approval gating as supplied runbooks.
Capabilities
Visual builderAuthor & testPublish & versionCustom operational agentsMulti-step workflowsPolicy bindingApproval gating
Platform 4 · UPC AI Platform Services

Serving and building models
inside the tenant boundary

An OpenAI-compatible router, an open-weight model registry, a per-tenant vector
database, and pre-built GPU workbenches — all within your vDC.
RouterUPC:AIF:AIPS:ROUTERIncluded
OpenAI-compatible inference, in-tenant
An OpenAI-compatible API endpoint inside the tenant vDC that routes inference across privately hosted models and, where the tenant elects it, external model providers — supporting common serving backends (e.g. vLLM, LiteLLM, TensorRT-LLM) so models move between private and public providers without application changes, with per-tenant API keys and access control.
Capabilities
OpenAI-compatible APIIn-tenant endpointPrivate model routingOptional external providersvLLM / LiteLLM / TensorRT-LLMNo app changesPer-tenant API keysAccess control
RegistryUPC:AIF:AIPS:REGISTRYIncluded
Open-weight models, privately deployed
A catalog of open-weight models for private deployment within a tenant vDC (e.g. Llama, Mistral, Falcon, Gemma, Phi, Qwen, DeepSeek families), together with customer-supplied private models — deployed versions and their serving configuration are recorded and can be rolled back.
Capabilities
Open-weight catalogLlama / Mistral / FalconGemma / Phi / Qwen / DeepSeekCustomer private modelsPrivate deploymentVersion recordingServing configurationRollback
VectorUPC:AIF:AIPS:VECTORIncluded
A per-tenant vector database
A vector-database instance deployed per tenant vDC on dedicated tenant resources (e.g. Milvus, Qdrant, Weaviate), with the distribution selected at design stage — deployed and lifecycle-managed by UnitedLayer, while index design, embedding strategy and application integration remain with the customer.
Capabilities
Per-tenant vector DBDedicated resourcesMilvus / Qdrant / WeaviateDesign-stage selectionFully managed lifecycleCustomer index & embeddingApplication integration
WorkbenchUPC:AIF:AIPS:WORKBENCHIncluded
Pre-built GPU development environments
Pre-built GPU development environments deployed from the service catalog (e.g. JupyterLab with PyTorch, TensorFlow, ONNX Runtime and CUDA toolchains), with per-tenant workspace isolation and shared dataset mounts from Core Cloud Storage.
Capabilities
Pre-built GPU envsJupyterLabPyTorch / TensorFlowONNX RuntimeCUDA toolchainsWorkspace isolationShared dataset mounts
Why the AI Factory
A complete AI cloud, in your
tenancy
Multi-tenant and isolated, GPU-first, private by default, and fully managed —
one factory that spans infrastructure, protection, operations, and the AI
platform.
Multi-tenant by design
Every tenant gets an isolated virtual data center — with isolation enforced across compute, storage, networking, identity and GPU, delegated administration, and role-based access.
GPU-first
Accelerators pooled per tenant with dedicated, vGPU and MIG partitioning on a high-performance fabric — scheduled into VMs, containers and serverless alike.
Everything in your tenancy
Models, vectors, inference and data stay inside your vDC. Route to external providers only if you choose — with per-tenant keys and access control.
Fully managed & included
UPC Core Cloud, Protect, Operations and AI Platform Services — 23 Services across four Platforms, all included and operated by UnitedLayer.

Recognized As A “Leader” in the ISG Private/Hybrid Cloud – Data Centre Services 2026

Learn More
Talk to an architect
↑