<img height="1" width="1" style="display:none;" alt="" src="https://px.ads.linkedin.com/collect/?pid=8691193&amp;fmt=gif">
High-performance computing on Azure

Azure HPC at supercomputer scale. Built and run by ALIANDO.

Run simulations, optimization solvers and model training on InfiniBand-connected Azure clusters with up to 368 cores per node. We design the architecture, migrate your workloads and operate the platform, so your engineers get answers instead of queue time.

30-minute scoping call with a Technical Account Manager · No obligation

  • Azure Expert MSP
  • 14 Microsoft specializations
  • 32 countries, 5 continents
  • 1,000+ client organizations
Example jobSlurm on Azure CycleCloud

Aerodynamics simulation

Running on 32 nodes

VM series
HBv5
Nodes
32
Cores
11,776
InfiniBand
800 Gb/s per node
Capacity appears when the job is submitted. When the queue is empty, the cluster scales back down, and so does the bill.
192days left

Azure retires HC-series and HBv2-series HPC VMs on May 31, 2027

After that date, remaining HC and HBv2 virtual machines are deallocated and stop working. New 1-year and 3-year purchases already ended on April 2, 2026. Quota requests, benchmarking and cutover all take time. Move to HBv5, HBv4 or HX before the deadline makes the decision for you.

Source: Microsoft Learn retirement notices for HC-series and HBv2-series virtual machines.

Book Your HPC Assessment
The bottleneck

Your next breakthrough is stuck in a queue

Fixed capacity, variable demand

An on-premises cluster is sized for an average week. Peak weeks wait in the queue. Quiet weeks leave paid-for hardware idle.

Refresh cycles that drain capital

Every hardware generation means another capital request, a procurement cycle and a data-center project before a single job runs faster.

Scarce HPC skills

Schedulers, MPI, InfiniBand and parallel storage each demand specialist knowledge. Most teams depend on one person who holds it all.

Azure removes the capacity ceiling. ALIANDO removes the complexity.

What you can run

Three demanding workloads. One platform built for all of them.

Each workload stresses a different part of the machine. We match it to the Azure hardware designed for exactly that stress.

Simulation

Computational fluid dynamics, finite element analysis, weather and climate, molecular dynamics and energy research. Tightly coupled MPI jobs that need memory bandwidth and a low-latency interconnect.

Runs best on

HBv5 and HBv4 · InfiniBand · Azure Managed Lustre

Optimization solvers

Mixed-integer and linear programming, scheduling, routing, supply-chain planning and portfolio risk. Large models that need memory and fast cores, plus scenario sweeps that need thousands of runs in parallel.

Runs best on

HX-series and HBv4 · up to 1,408 GB of memory per node · Azure Batch for parallel runs

Model training

Distributed deep-learning training, fine-tuning and large-scale inference. Multi-node GPU jobs that need fast GPU-to-GPU networking and storage that keeps pace with checkpoints.

Runs best on

ND-series with NVIDIA H100, H200 and Blackwell GPUs · InfiniBand · Azure Managed Lustre

Also a strong fit: genomics · semiconductor design (EDA) · seismic and reservoir modeling · financial risk · rendering. Not sure where your workload lands? Book Your HPC Assessment

The platform

Purpose-built Azure HPC infrastructure, assembled the right way

Azure offers the building blocks of a supercomputing center. The results come from combining them correctly for your workload.

Compute

HBv5, HBv4 and HX for CPU workloads. ND-series for GPU workloads. Sized to the job in front of you, not to an annual budget cycle.

  • HBv5
  • HBv4
  • HX
  • ND-series GPU

Orchestration

Keep your scheduler with Azure CycleCloud, which has built-in support for Slurm, PBS Pro, LSF, Grid Engine and HTCondor. Or go cloud-native with Azure Batch, scheduling as a service.

  • Azure CycleCloud
  • CycleCloud Workspace for Slurm
  • Azure Batch

Storage

A managed parallel file system for throughput, enterprise file storage for shared data and object storage for long-term retention.

  • Azure Managed Lustre
  • Azure NetApp Files
  • Azure Blob Storage

Network and hybrid

InfiniBand RDMA between nodes. A private ExpressRoute or VPN path to your data center, so you can burst from the cluster you already own.

  • InfiniBand RDMA
  • ExpressRoute
  • VPN Gateway
368
cores per HBv5 virtual machine
AMD EPYC with high-bandwidth memory
800 Gb/s
InfiniBand per HBv5 node
for tightly coupled MPI at scale
1,408 GB
memory per HX-series node
for large solver and EDA models
5
schedulers built into CycleCloud
Slurm · PBS Pro · LSF · Grid Engine · HTCondor

Specifications from Microsoft Learn, retrieved November 10, 2026. Availability varies by Azure region.

Professional services

From first benchmark to steady-state operations

Azure supplies the hardware. Production results take architecture, migration and operations work. Even Microsoft's ready-to-deploy Slurm workspace leaves configuration, security, monitoring and upgrades with your team. That is the work we take on.

  1. 01

    Assess

    We inventory your applications, schedulers, data and licenses, then map each workload to the right VM family, scheduler and storage tier.

    You getWorkload-fit and sizing recommendation

  2. 02

    Prove

    We stand up a proof-of-concept cluster and benchmark your real jobs, so you see time and cost per job before you commit budget.

    You getBenchmarked time and cost per job

  3. 03

    Build

    Landing zone, identity, networking, InfiniBand-ready images, parallel storage, license servers and quota, delivered as code.

    You getProduction-ready HPC environment

  4. 04

    Migrate

    We move job scripts, queues and data, set up bursting from your existing cluster where it pays off, and plan cutover around your project deadlines.

    You getMigrated workloads and runbooks

  5. 05

    Operate

    Managed operations from an Azure Expert MSP: monitoring, patching, scheduler upgrades, capacity planning and cost governance.

    You getManaged HPC with usage and cost reporting

What we take off your plate

  • Scheduler migration: Slurm, PBS Pro, LSF, Grid Engine, HTCondor
  • MPI and InfiniBand tuning
  • License-server design for commercial solvers and simulation codes
  • vCPU quota and regional capacity planning
  • Parallel storage design with Azure Managed Lustre and Azure NetApp Files
  • Hybrid bursting over ExpressRoute or VPN
  • Cost governance: autoscaling, budgets and Azure purchasing options
  • Security and identity aligned to your policies

Not sure cloud HPC will beat your cluster?

Good. Prove it before you commit. We benchmark your real jobs on Azure and show you time and cost per job, next to what you run today.

Why ALIANDO

A global Microsoft partner with the depth for hard problems

ALIANDO was formed in 2025 by the merger of Henson Group and myCloudDoor, two established Microsoft partners. Today we serve more than 1,000 client organizations worldwide.

32
countries
across five continents
14
Microsoft specializations
cloud, AI and security
1,000+
client organizations
served worldwide
6
Partner of the Year awards
from Microsoft

Azure Expert MSP

A Microsoft designation for managed service providers, earned through a rigorous audit by an independent third party and re-checked every year. It is why we can operate your HPC platform, not only build it.

Specializations that matter for HPC

Infra and Database Migration, Kubernetes, AI Platform, Cloud Security and Identity and Access Management are among our 14 Microsoft specializations.

Access to Microsoft funding

As an Azure Expert MSP, we can help you apply for Microsoft funding programs that offset assessments, proofs of concept and deployments. Microsoft decides eligibility.

Questions, answered

Azure HPC: frequently asked questions

What is high-performance computing (HPC) on Azure?

HPC on Azure uses large numbers of CPU or GPU virtual machines, connected by low-latency InfiniBand networking, to solve problems too large for ordinary servers. Capacity is added and removed on demand, so compute stops being a fixed ceiling.

Which Azure virtual machines are built for HPC?

HBv5, HBv4 and HX for CPU workloads, and ND-series for GPU workloads. HBv5 offers up to 368 cores and 800 Gb/s of InfiniBand per virtual machine. HX offers up to 1,408 GB of memory for large models.

Can we keep our current scheduler and job scripts?

Yes. Azure CycleCloud has built-in support for Slurm, PBS Pro, LSF, Grid Engine and HTCondor, so existing scripts, queues and policies can move with minimal change. For new cloud-native workloads, Azure Batch provides scheduling as a service.

Do we have to move everything to the cloud at once?

No. Many teams start by bursting peak demand from an on-premises cluster to Azure over ExpressRoute or VPN, then move more workloads as the results prove out.

What happens to HC-series and HBv2-series virtual machines?

Microsoft retires both on May 31, 2027. Any remaining virtual machines are deallocated and stop working. Microsoft recommends moving to HBv5, HBv4, HX or HBv3, depending on the workload.

How do commercial solver and simulation licenses work in Azure?

It depends on the vendor. Not every vendor offers pay-as-you-go licensing, and you may need a license server in Azure or a connection to your on-premises one. We design this with you during the assessment.

How do we keep cloud HPC costs under control?

Clusters autoscale with the job queue, so you pay for what runs. We add budgets and per-team limits, choose the right Azure purchasing options and benchmark cost per job before production.

Who is ALIANDO?

ALIANDO is a global Microsoft partner formed in July 2025 by the merger of Henson Group and myCloudDoor. It is an Azure Expert MSP with 14 Microsoft specializations, operating in 32 countries across five continents.

Start now

Book your HPC assessment

192 days remain before HC-series and HBv2-series retirement, and every week on a constrained cluster is a week of results you do not get back. Thirty minutes is enough to find out where you stand.

  1. 01Scoping call. Thirty minutes on your workloads, schedulers and deadlines.
  2. 02Assessment. A workload-fit and sizing recommendation for Azure.
  3. 03Proof of concept. Your real jobs, benchmarked for time and cost.

Your contact

Michael Betancourt Gelves

Technical Account Manager · ALIANDO

michael.betancourt@aliando.com

Tell us what you need to run

Four details are enough to prepare a useful first call.

No obligation. We use your details only to respond to this request. Read our privacy policy.

Microsoft and Azure are trademarks of the Microsoft group of companies. NVIDIA and AMD names are trademarks of their respective owners. Technical specifications and retirement dates are taken from Microsoft Learn as of November 10, 2026 and can change.