Azure Pricing Changes in 2026: Every Update That Affects Your Bill

The single most consequential Azure pricing event of 2026 is not a rate change at all — it is a purchasing-availability cutoff. Starting July 1, 2026, Microsoft stopped accepting new purchases and renewals of Reserved VM Instances (RIs) for a long list of legacy VM series, pushing affected customers toward Azure savings plans for compute (Microsoft Learn, 2026). For teams on AWS, this is the closest analog to Graviton-driven migration pressure — except here the trigger is a commitment-program change rather than new hardware. For a side-by-side view of where Azure versus AWS undercuts each other, see our AWS vs Azure pricing comparison.

Existing Reserved VM Instances are not cancelled and do not lose their discount — Microsoft is explicit that "July 1, 2026 does not terminate existing RIs." The risk is entirely forward-looking: when an affected RI expires, there is no renewal option, and without a savings plan or modernized VM family in place, that workload falls back to full pay-as-you-go pricing with no advance warning banner in the portal.

On the AI side, Azure OpenAI Service pricing in 2026 continues to track Microsoft's stated commitment to price parity with OpenAI's own API for applicable models, meaning OpenAI-side reductions (including steep cuts tied to newer GPT-5.x model releases) generally flow through to Azure customers on the same models (Azure OpenAI Service Pricing).

Three confirmed changes have landed through 2026 so far. Check back or subscribe for updates as Microsoft announces more — Azure typically publishes reservation and retirement changes months in advance via Azure Updates and Microsoft Learn, but rarely surfaces them in the portal for affected customers proactively.

Q1 2026 (January – March)

Q2–Q3 2026

• Blob Storage minimum billable object size — paused, not cancelled: Microsoft had planned a 128 KiB minimum-billable-object-size policy for the Cool, Cold, and Archive access tiers, rolling out July 1, 2026 for new storage accounts and July 1, 2027 for existing ones. Microsoft paused the rollout on June 8, 2026, before it took effect, and stated it would publish a revised approach and timeline later. Cool, Cold, and Archive billing did not change on July 1, 2026 as a result.
• Azure Smart Tier for Blob Storage: reached general availability in April 2026, automatically moving individual blobs between Hot, Cool, and Cold tiers based on observed access patterns — a lower-maintenance alternative to manually configured lifecycle policies for teams that have not tuned tiering rules recently.

The RI retirement makes 2026 the year most Azure teams need to actually understand the difference between the two commitment models, rather than defaulting to whichever one they set up years ago. Reserved VM Instances lock in a discount for a specific VM series, size, and region for a 1- or 3-year term, with estimated savings of roughly 36–72% versus pay-as-you-go — the top end based on a 3-year commitment on specific SKUs (Azure Reserved VM Instances).

Azure savings plans for compute instead commit to a fixed hourly spend, applied automatically across eligible compute usage — VM family, size, and region can all change without losing the discount. Estimated savings run roughly 11–65% depending on term and commitment level (Azure Savings Plans). The trade-off is straightforward: RIs win on raw discount depth for genuinely static workloads; savings plans win on flexibility for anything that changes shape over its commitment term. For a broader breakdown of commitment models across providers, see our reserved vs spot vs on-demand comparison.

Microsoft's own guidance for the July 2026 transition offers a self-service trade-in path — existing RIs on affected series can be exchanged for savings plan commitments without waiting for natural expiration, which is worth evaluating if you are already planning a VM modernization cycle.

The VM series losing RI purchase eligibility on July 1, 2026 — Av2, Amv2, Bv1, D, Ds, Dv2, Dsv2, Dv3, Dsv3, Ev3, Esv3, F, Fs, Fsv2, G, Gs, Ls, and Lsv2 — are almost all previous-generation families with newer replacements already available (Dv5/Dsv5, Ev5/Esv5, Fv2 successors, and so on). The retirement is effectively Microsoft nudging the installed base toward hardware it would recommend anyway.

For teams with RIs expiring after July 1, 2026, Microsoft's own decision guide lays out three paths: transition to a savings plan, modernize to a current VM generation (often improving price-performance in the process), or do nothing and accept pay-as-you-go rates — explicitly framed as something that "should be a conscious decision, not an accidental outcome."

Practically, this means auditing reservation inventory is now a 2026 priority even for teams that treat RIs as fire-and-forget. Cost Management + Billing surfaces VM family and expiration date per reservation — a five-minute filter that avoids a silent bill spike months later.

Storage carries the two clearest 2026 Azure changes: one already complete, one paused before it landed. Unmanaged Disks are gone as of March 31, 2026 — not a pricing change, but a hard availability cutoff that stops affected VMs outright (Microsoft Learn).

The Blob Storage minimum-billable-object-size policy is the more instructive story. Microsoft announced a plan to bill any object under 128 KiB in the Cool, Cold, or Archive tiers as if it were 128 KiB — a change that would have meant a million small files under a few KB each getting billed at many multiples of their actual size. The policy was scheduled for new accounts on July 1, 2026, but Microsoft paused it on June 8, 2026, before rollout, and has not yet published a revised timeline. Cool, Cold, and Archive billing was unaffected on July 1 as a result of the pause.

That pause is worth tracking rather than ignoring. If you run workloads with many small objects in cooler tiers — logs, thumbnails, config files — this remains a plausible future change, and it is far cheaper to consolidate small objects proactively than to react after a revised policy actually lands.

Separately, Azure Smart Tier reached general availability in April 2026, automating blob movement between Hot, Cool, and Cold based on observed access patterns. For accounts with stale or never-configured lifecycle rules, this is a lower-effort way to capture tiering savings than hand-writing policy rules.

Azure OpenAI Service pricing is structured around three models: standard pay-as-you-go per-token billing, Provisioned Throughput Units (PTUs) for reserved capacity, and a Batch API offering a 50% discount on standard rates for completions returned within 24 hours (Azure OpenAI Service Pricing). Rates differ across the three available deployment types — Global, Data Zone (EU or US), and Regional — so the same model can carry different per-token prices depending on which deployment option you choose.

Microsoft has stated Azure is committed to pricing parity with OpenAI's own API for applicable Azure OpenAI offerings. Practically, that means when OpenAI cuts direct API pricing on a given model generation, Azure customers on the equivalent Azure-hosted model generally see the same reduction reflected on the Azure pricing page, rather than lagging behind on an older rate card.

Prompt caching — billing cached (repeated) input tokens at a reduced rate rather than the full input-token price — was scheduled to begin billing on August 21, 2026. Before that date, cached tokens were not charged separately; teams that assumed this was a permanent free feature should confirm current billing status against the pricing page for their specific model and deployment before budgeting.

For PTU-based deployments, minimum commitments range from roughly 15–50 PTUs depending on the model and region, with hourly, monthly, and annual reservation terms available. Teams spending consistently above the on-demand break-even point should model PTU reservation costs against actual sustained usage — not projected peak usage, which tends to overstate the PTU count actually needed.

AKS pricing in 2026 remains structured around three cluster tiers. The Free tier has no financially backed uptime SLA and is billed only for the underlying node compute; Standard and Premium tiers add a per-cluster hourly management fee (approximately $0.10/hour, roughly $72/month per cluster) in exchange for a backed SLA — 99.95% with availability zones, 99.9% without.

As with AWS EKS's flat per-cluster fee, the AKS Standard/Premium fee rewards cluster consolidation. Teams running separate clusters per environment (dev, staging, prod) on Standard or Premium tiers accumulate the management fee per cluster regardless of node count — using namespaces and RBAC to isolate environments within fewer clusters avoids stacking that fee unnecessarily for non-production workloads that do not need an SLA-backed control plane.