AWS
Microsoft Azure
Google Cloud
Some affect the platforms you can build on, others change how you manage commitments and governance, and a few deadlines are closer than they may seem.
In this update, we break down the Azure announcements, pricing and commitment changes, and upcoming lifecycle milestones that FinOps and cloud teams should have on their radar.
The short answer
For FinOps teams closing out June, prioritize three things:audit affected RIs ahead of July 1,
assess the cost implications of June’s platform previews, and
review the updated Product Terms if you deploy Azure AI services.
One-year RIs for Av2, Amv2, Bv1, D, Ds, Dv2, Dsv2, F, Fs, Fsv2, G, Gs, Ls, and Lsv2 will no longer be available for purchase or renewal. One- and three-year RIs for Dv3, Dsv3, Ev3, and Esv3 will also become unavailable for new purchase or renewal.
Existing reservations continue to receive their reservation discounts through the remainder of their original terms. The change does not currently apply to customers in China.
For Build announcements, the lesson is not “buy the new service.” HorizonDB is in public preview, and new platform capabilities should be evaluated for architecture, compatibility, availability, and pricing before being treated as cost-saving measures.
What changed in Azure in June 2026?
June’s Azure updates fall into three categories: new platform capabilities, commitment and lifecycle changes approaching important deadlines, and governance updates.| Development | June relevance | FinOps relevance |
|---|---|---|
| HorizonDB | Public preview announced at Build | New database platform; verify pricing and eligibility |
| Azure Linux | Public preview highlighted at Build | Potential OS licensing opportunity |
| Azure RIs | May-announced purchase and renewal change approaches July 1 | Review affected reservations |
| Product Terms | Azure AI terminology and cross-cutting terms updated June 1 | Review governance and procurement records |
| Blueprints | Phased retirement begins July 31 | Start governance migration |
| NVv3 and NVv4 | September retirement approaching | Coordinate workload migration and commitment decisions |
Microsoft Build: HorizonDB is the database announcement to watch
At Microsoft Build, Azure HorizonDB entered public preview. Microsoft describes it as a fully managed, PostgreSQL-compatible database that combines transactional data, vector search, and AI capabilities in a single platform for AI-powered applications.For FinOps teams, the important point is that HorizonDB is a separate Azure database service. Teams evaluating it should model HorizonDB’s pricing and eligibility separately rather than assuming pricing arrangements for other Azure database services automatically apply.
Because HorizonDB was in public preview in June, cost modeling should remain conditional. Confirm the service’s current pricing, supported regions, availability, and commitment eligibility before including it in a long-term savings forecast. Read Microsoft Blog for PostgreSQL.
Microsoft also highlighted new PostgreSQL security and migration capabilities, including Defender for Cloud integration and discovery and assessment tooling for PostgreSQL migrations. These are modernization capabilities rather than immediate billing reductions.
For FinOps teams, the practical question is whether a migration changes the workload’s total cost of ownership. That requires comparing the full database configuration, storage, compute, networking, operational requirements, and migration costs rather than treating a new managed service as inherently cheaper.
Also read: GCP June 2026 FinOps Updates
Azure Linux enters public preview
Azure Linux entered public preview as part of Microsoft’s Build 2026 announcements. Microsoft positions Azure Linux as a platform built for Azure, with an emphasis on standardization, security, and operational consistency.Moving from RHEL or SUSE to Azure Linux may reduce OS licensing costs where the workload is compatible, but it is not automatically a saving. Validate application compatibility, support requirements, security controls, dependencies, and operational ownership first.
Azure Linux does not eliminate VM compute, storage, networking, support, or application costs.
A sensible June evaluation is to inventory VMs with paid Linux licensing, identify workloads that could run on Azure Linux, and then compare the recurring license cost against migration and operational requirements.
For workloads with significant software dependencies or compliance requirements, the migration cost may outweigh the potential licensing benefit.
The most urgent June FinOps item is the July 1 RI change.
The change itself was announced by Microsoft on May 4, 2026. Starting July 1, 2026, new purchases and renewals will no longer be available for specified Azure Reserved VM Instances. June is therefore the final planning window before the change takes effect.
Microsoft’s transition guidance identifies the affected terms as:
One-year RIs: Av2, Amv2, Bv1, D, Ds, Dv2, Dsv2, F, Fs, Fsv2, G, Gs, Ls, and Lsv2.
One-year and three-year RIs: Dv3, Dsv3, Ev3, and Esv3.
In particular, Dv3, Dsv3, Ev3, and Esv3 are not being retired as a result of this RI change. Existing RIs remain valid and continue receiving their reservation discounts through the end of their original terms.
The change also does not currently apply to customers in China. Read Transition guide for retired Azure Reserved VM Instances.
What should teams do in June?
Start with the reservation inventory.Identify affected VM families and reservation expiry dates in the Azure portal.
Separate workloads that are stable from workloads already scheduled for migration, resizing, or retirement.
For stable workloads that are still eligible for an affected RI term before July 1, compare the economics of the available RI purchase or renewal against the alternatives.
For workloads expected to change, evaluate Azure Savings Plan for Compute or migration to a supported VM family.
Confirm the transaction and resulting reservation or plan status after purchase or trade-in.
Do not treat the deadline as a reason to maximize commitment coverage. A reservation creates value only when the workload remains eligible and sufficiently utilized. Base the decision on current usage, expected workload stability, and planned architecture changes.
For a deeper comparison, see Usage.ai’s guide to Azure Savings Plans versus Reserved Instances.
Microsoft Product Terms changed the Azure AI framework
Microsoft’s June 1 Product Terms update made several changes to Azure’s AI terminology and cross-cutting terms.It introduced a Microsoft Azure AI Services definition covering certain cross-cutting requirements for Azure services incorporating AI.
It renamed Azure Direct Models to Foundry Models sold by Azure and replaced several model-provider-specific provisions with a consolidated reference to documentation.
It clarified the scope of models classified as Non-Microsoft Products and confirmed that certain third-party models remain subject to the Enterprise AI Services Code of Conduct.
The Product Terms also clarified Azure Hybrid Benefit language for Azure Local, specifying that only one Azure Local SKU is available for Azure Hybrid Benefit.
Teams using Azure Local should verify their exact licensing position against the June 2026 Microsoft Product Terms changes rather than assuming that Azure Hybrid Benefit treatment for Azure VMs applies identically to Azure Local.
Azure Blueprints enters a phased retirement
Another June lifecycle update affects governance rather than direct cloud pricing.Microsoft’s current retirement schedule places full Azure Blueprints retirement on January 31, 2027, with phased restrictions beginning July 31, 2026.
Microsoft recommends migrating existing blueprint definitions and assignments to Azure Deployment Stacks and, where appropriate, Template Specs. Read more about Azure Blueprints.
The phased timeline is:
| Date | Change |
|---|---|
| July 31, 2026 | New blueprint definitions and versions can no longer be created |
| October 31, 2026 | Existing definitions can no longer be modified; new assignments can no longer be created |
| December 31, 2026 | Existing assignments can no longer be modified |
| January 31, 2027 | Azure Blueprints is retired |
The concern is operational: governance automation that depends on Blueprint definitions, assignments, or locks needs a migration plan before those capabilities are restricted or removed.
This is a planning item, not a July 1 emergency.
Also read: AWS in June 2026 Updates.
Another lifecycle item: NVv3 and NVv4
Two GPU VM families are scheduled for retirement on September 30, 2026: NVv3 and NVv4. Microsoft recommends migrating affected workloads to newer VM series before the retirement date.For NVv3, one- and three-year RI purchases ended on November 2, 2025. Microsoft currently recommends migrating NVv3 workloads to newer options such as NVadsA10_v5 or NVads_V710_v5.
For NVv4, one- and three-year RI purchases also ended on November 2, 2025. Microsoft recommends newer options including NVads_V710_v5, with NVadsA10_v5 and NGads_V620 also identified for certain workloads.
There is one regional exception worth noting: NVv3 workloads in Azure UAE North have an extension until January 31, 2027. Other NVv3 workloads remain on the September 30 retirement schedule.
For teams operating either series, migration planning should happen alongside commitment planning. A workload expected to move to a different VM family should not be treated as a stable candidate for a long-term commitment simply because its current utilization is predictable.
What June means for Azure FinOps teams
June’s updates point to a common FinOps pattern: product changes and commitment decisions should be managed on the same timeline.Start with the July 1 RI change. Then separate three kinds of work:
Cost decisions: review affected reservations and model commitment alternatives.
Architecture decisions: evaluate HorizonDB and Azure Linux based on workload requirements, compatibility, pricing, and total cost rather than headline cost assumptions.
Governance and lifecycle decisions: update AI terminology and controls, begin the Blueprints migration, and plan migrations away from NVv3 and NVv4.
How Usage.ai fits
For teams reviewing Azure commitment coverage, Usage.ai can complement native Azure pricing and reservation tools as a commitment-management layer. First establish which usage is stable and eligible; then determine how much commitment to carry.Usage.ai’s Azure Savings Plan coverage can help teams evaluate commitment opportunities without treating a provider maximum discount as an expected saving. For the underlying mechanics and current Azure commitment options, see the Azure Savings Plan guide.
June 2026 Azure update: what to do next
For teams closing out June, review your Azure estate in this order:Reservations: identify affected RI families and expirations before July 1.
Usage stability: distinguish predictable base usage from workloads likely to migrate, resize, or scale down.
New platforms: evaluate HorizonDB and Azure Linux as preview technologies, not guaranteed cost reductions.
Governance: review the June Product Terms changes if your organization uses Azure AI or Foundry Models.
Lifecycle: put Blueprints, NVv3, and NVv4 migration work on the appropriate planning calendar.
How Usage.ai Fits
Azure provides native tools for rightsizing workloads and purchasing Reservations and Savings Plans. These options can reduce the cost of stable usage, but long-term commitments can create risk when workloads change and actual usage no longer matches the original commitment.At Usage.ai, we help teams manage the commitment layer as usage changes. With Flex Insured Commitments, teams can get the 72% savings of a 1- or 3-year commitment with none of the commitment.
If an eligible Flex Commitment costs more than equivalent pay-as-you-go usage, we provide cashback protection to help cover the difference. See our cashback documentation for details.
Our goal is to help manage the commitment layer after you have established an appropriate Azure compute baseline, particularly when usage is stable enough to benefit from commitment pricing but may change over time.
Automate Azure Reservations and Savings Plans with Usage.ai and get cashback protection without the risk of long-term commitments.
Talk to an Expert
Frequently asked questions
Does the July 1 RI change retire the affected Azure VM families?
No. The July 1 change removes new RI purchase and renewal availability for specified VM series. Some affected families, including Dv3, Dsv3, Ev3, and Esv3, remain active VM products.
When was the Azure RI change announced?
Microsoft announced the change on May 4, 2026. The new purchase and renewal restrictions take effect July 1, 2026, making June the key planning period for affected customers.
Does the July 1 RI change apply to Azure customers in China?
Not currently. Microsoft’s May announcement states that the change does not currently apply to customers in China.
Should I move an affected RI to an Azure Savings Plan?
Not automatically. Compare the workload’s stability, migration plans, eligibility, expected usage, and commitment economics. Microsoft identifies Savings Plans as one transition option, but the right choice depends on the workload.
Is Azure HorizonDB generally available?
No. HorizonDB entered public preview in June 2026. Pricing, availability, and eligibility should therefore be verified before using it in a production commitment model.
