Other
Microsoft
Microsoft is retiring every version of Microsoft HPC Pack, its on-premises HPC cluster and job-scheduling software; End of Support is 27 August 2027, after which there are no updates, security fixes, or technical support, and customers are directed to Azure Batch or another Azure service for parallel and HPC scheduling
departs 27 Aug 2027
Quick answers
- Is it true that Microsoft is retiring all versions of HPC Pack — End of Support is 27 August 2027; migrate HPC scheduling to Azure Batch?
- Yes. Microsoft has officially announced it, effective 27 Aug 2027. Microsoft HPC Pack — the free cluster manager and job scheduler used to run high-performance and parallel compute workloads on Windows (and hybrid Azure) clusters — is being **retired in all versions**, with **End of Support on 27 August 2027**.
- When does it take effect?
- 27 Aug 2027. Verified against the vendor's own announcement.
- What should I use instead?
- Microsoft directs HPC Pack customers to **Azure Batch** for managed parallel and HPC job scheduling, or to another appropriate Azure service, while warning that no alternative guarantees one-to-one feature parity.
What this means for you
Microsoft HPC Pack — the free cluster manager and job scheduler used to run high-performance and parallel compute workloads on Windows (and hybrid Azure) clusters — is being **retired in all versions**, with **End of Support on 27 August 2027**. From the retirement announcement (29 September 2026) HPC Pack enters a one-year, retirement-only support period: Microsoft will still ship security/CVE fixes and answer questions about documented, supported configurations, but **non-security fixes, new features, new-OS and new-hardware validation, and migration consulting are all out of scope**. After 27 August 2027 there are no updates, fixes, or support at all — existing clusters keep running but become permanently unsupported, which for regulated or security-sensitive HPC environments is effectively a hard deadline. Anyone operating an HPC Pack head node and compute cluster — on-prem, in Azure VMs, or hybrid burst-to-Azure — needs a scheduler migration plan, and Microsoft is explicit that no alternative guarantees one-to-one feature parity.
Where to go
Microsoft directs HPC Pack customers to **Azure Batch** for managed parallel and HPC job scheduling, or to another appropriate Azure service, while warning that no alternative guarantees one-to-one feature parity. Start the assessment now rather than near the date: catalogue what your cluster actually depends on — the job scheduler and queues, MPI workloads, node/pool autoscaling, on-prem licensing, and any scripts or submission tooling built against the HPC Pack APIs — because those are the pieces that do not port automatically. For cloud-native re-hosting, Azure Batch provides managed pools, scheduling and autoscale; for lift-and-shift of Windows/MPI workloads, evaluate Azure CycleCloud or a third-party scheduler (Slurm, PBS) depending on your stack. Whatever the target, plan for re-testing job submission, scaling behaviour and performance before cutover, and complete the move before the 27 August 2027 End of Support date so you are never running an unsupported, unpatched scheduler in production.
- Azure Batch — Microsoft's recommended managed parallel and HPC job scheduling service
- Azure CycleCloud or a third-party scheduler (Slurm, PBS) — for lift-and-shift of existing HPC/MPI workloads
Microsoft is retiring all versions of Microsoft HPC Pack, its cluster manager and job scheduler for high-performance and parallel compute. The official retirement notice (published 29 September 2026) sets End of Support at 27 August 2027 and starts a one-year, retirement-only support period now.
What “retirement-only support” means until 27 August 2027
Microsoft has drawn the scope tightly. In scope during the retirement year:
- Security/CVE response and necessary security updates
- Guidance from existing documentation and validated configurations
- Assistance with supported components and documented functionality
Out of scope — starting immediately:
- Non-security fixes or product improvements
- New features, custom code, or design changes
- Validation of new operating systems, hardware, dependencies, architectures, or integrations
- Migration planning, execution, or consulting
After 27 August 2027, HPC Pack receives no updates, fixes, or technical support. Existing deployments keep running but are permanently unsupported.
What to do
Microsoft’s recommended replacement for managed parallel and HPC scheduling is Azure Batch, or another appropriate Azure service — with the explicit caveat that no alternative guarantees one-to-one feature parity. Because migration planning and execution are out of scope for the retirement-year support, this is on you to scope early.
Start by cataloguing what your cluster actually relies on — the scheduler and job queues, MPI workloads, autoscaling, on-prem licensing, and any submission tooling written against HPC Pack’s APIs. Then choose a target: Azure Batch for cloud-native managed scheduling and autoscale, or Azure CycleCloud / a third-party scheduler such as Slurm or PBS if you need to lift-and-shift existing Windows or MPI workloads. Re-test job submission, scaling and performance before cutover, and finish the move before 27 August 2027 so you are never left running an unpatched, unsupported scheduler in production.
Entry changelog
- — last reviewed.
- — Entry created from Microsoft's official Azure retirement notice (Azure Updates id 570046, published 29 September 2026). The Azure Updates page is JavaScript-rendered and does not fetch server-side, so the End of Support date (27 August 2027), the all-versions scope, the one-year retirement-support terms, and the recommended target (Azure Batch) were verified against Microsoft's server-rendered release-communications API (id 570046). No detect fingerprint: HPC Pack is installed cluster software with no package/image/runtime identifier a source repository would contain (SOURCING §6).
Put this countdown in your README
Building on Microsoft? Drop this badge into a project’s README or docs — it shows the days left and
refreshes on its own every night from this page. 
[](https://lastcall.dev/entries/microsoft-hpc-pack-retirement)Spot an error or a changed date? Tell us and it’s corrected the same night.