Insights

The biggest Azure machines yet arrive without a temp drive

Braintree Insights | 26 August 2026

The biggest Azure machines yet arrive without a temp drive

Azure made 248 and 372 vCPU sizes of the D and E v7 series generally available on 25 August 2026, reaching 372 vCPUs and 2.8 TiB of memory on Intel Xeon 6. The size documentation carries three constraints the announcement does not: the whole v7 series has no temporary storage, Generation 1 VMs are not supported, and the series only runs on operating system images that support NVMe. None of the eight launch regions is in Africa.

What changed

Microsoft announced on 25 August 2026 that 248 and 372 vCPU sizes of the Dlsv7, Dsv7 and Esv7 families are generally available, powered by Intel Xeon 6 processors and delivering, in Microsoft’s own words, up to 20 percent better compute performance than prior-generation Intel-based v6 VMs. The scale figures are substantial: the new sizes reach 372 vCPUs and 2.8 TiB of memory, up to 400 Gbps of networking bandwidth on the 372 vCPU Esv7 and Edsv7 sizes, up to 800,000 IOPS and 20 GBps to Premium SSD v2 and Ultra Disk remote storage, and up to 9.6 million IOPS to the local NVMe temporary disk on the Ddsv7 and Edsv7 sizes. The announcement names its launch regions: Central US, East US, East US 2, Germany West Central, South Central US, Sweden Central, West US 2 and West US 3. Eight regions, six of them in the United States and two in Europe. South Africa North is not among them and neither is any other African region, which for a South African tenant makes this a planning input rather than a deployment option. The Dsv7 size series documentation, updated on 24 August 2026, supplies what the announcement does not. The series host specification records local storage as None, and the page states directly that Dsv7-series virtual machines do not have any temporary storage, framing the absence as something that lowers the price of entry. The feature support table marks Generation 1 VMs as not supported and Generation 2 as supported. And a note on the same page states that the series will only work on operating system images that support NVMe, and that an image without NVMe support produces an error message.

The operational risk is easy to miss because the service can continue to look healthy. The control becomes visible only when a capacity request fails, an unsupported runtime is removed, or an extension blocks an enforced ERP update. Waiting for that moment transfers a planned decision into an incident.

What the term means in plain language

A vCPU is a virtual processor presented to a virtual machine. A VM series is a family of sizes sharing a hardware platform and a feature set, so Dsv7 is the seventh-generation general-purpose series with premium storage support. The temporary disk, historically the D drive on Windows and /mnt on Linux, is local storage attached to the host, fast and explicitly not durable, which is why it has traditionally been used for paging and scratch data. Generation 2 refers to the VM’s firmware model, using UEFI rather than the older BIOS model of Generation 1. NVMe is the interface the newer Azure platforms use to present disks; an operating system image must carry NVMe drivers to boot on hardware that offers only that interface.

This distinction matters because product status is not the same as business readiness. Availability, support and compatibility are separate questions. A service can be available but unsupported, supported but capacity-constrained, or technically updated while a customer-specific process has stopped working.

Why this matters to a South African organisation

South African teams often operate with tight specialist capacity, rand-sensitive budgets and business processes that cannot be paused while a replacement is sourced. Localisation, regional cloud capacity and long procurement lead times can narrow the recovery options. The practical response is to use the available test window before it becomes an emergency window.

The consequence belongs to the business process, not only the technology team. Finance month-end, customer transactions, data pipelines and ERP extensions all cross technical and operational ownership. A change should therefore be accepted only when the service owner and the business owner can see the same evidence.

The hidden exposure

Azure made 248 and 372 vCPU sizes of the D and E v7 series generally available on 25 August 2026, reaching 372 vCPUs and 2.8 TiB of memory on Intel Xeon 6. The size documentation carries three constraints the announcement does not: the whole v7 series has no temporary storage, Generation 1 VMs are not supported, and the series only runs on operating system images that support NVMe. None of the eight launch regions is in Africa.

Normal operation is weak evidence. It proves only that yesterday’s combination of platform, configuration and workload completed. It does not prove that the next capacity allocation, lifecycle enforcement or major release will preserve the same result. An owner needs an inventory, a representative test and a dated decision.

Decision path

The decision this changes is not whether the new sizes are impressive. It is whether the workload someone has in mind can actually move to them, and that has to be settled before a number goes into a budget. Take the temporary drive first, because it is the constraint most likely to be discovered late. A Windows server that has run in Azure for several years is very likely to be using the temporary drive for something, whether that is the paging file, a SQL Server tempdb placed there deliberately for speed, or an application writing scratch files to a path an administrator chose once and never revisited. On a v7 size that storage does not exist, so the work is not a resize, it is a configuration change to every consumer of that path, tested. Second, the generation. A Generation 1 virtual machine cannot run on this series at all, and a machine that was created years ago in an earlier Azure estate is a plausible candidate for being Generation 1 without anyone having checked. Third, the image. NVMe support is present in the most popular current images and Microsoft says image compatibility is improving, but an older or custom-built image may not carry it, and the failure appears at deployment as an error rather than as a warning during planning. Braintree’s view, not Microsoft’s, is that these three questions are also a free audit. Whatever the outcome on v7, a workload that cannot answer them is a workload nobody currently understands well enough to move anywhere. For a South African organisation there is a fourth question underneath the other three. If the workload genuinely needs this class of machine, and none of the launch regions is local, then the choice becomes a latency and data-residency decision rather than a compute one, and it should be made by whoever owns that risk rather than settled inside a sizing exercise.

Record the alternatives that were rejected and why. That prevents the next reviewer from reopening the entire question without context. Where the preferred path cannot be completed inside seven days, approve a time-bound exception with a responsible owner, expiry date and compensating control.

Technical test plan

The inventory is small and mechanical. For each candidate workload record the VM size and generation, which the Azure portal, Azure CLI and PowerShell all expose, and treat any Generation 1 result as an immediate exclusion pending conversion. Record the operating system image, including whether it is a marketplace image or a custom one, and confirm NVMe support before scheduling any deployment window rather than after. Then find the temporary-drive consumers, which is the part that takes real work. On Windows, check the paging file location, the SQL Server tempdb data and log file paths, any application configuration pointing at the temporary drive letter, and any scheduled task or service writing there. On Linux, check for anything mounted at or writing to the ephemeral mount point. Where a consumer is found, the replacement is a data disk of the appropriate type, and the Dsv7 documentation records support for Standard SSD, Standard HDD, Premium SSD and Premium SSD v2, with Ultra Disk available subject to regional availability. That change alters the cost model, because storage that used to arrive with the VM now appears as a separately billed disk, which is the reason to price it during planning rather than after a migration. The series supports Live Migration, accelerated networking and nested virtualization, and the feature table marks memory-preserving updates as not supported, which belongs in the availability conversation for anything that cannot absorb a host maintenance restart. Finally, confirm quota and region before committing to a date, because a size being generally available does not mean the subscription has the vCPU quota to deploy it.

Use production-representative conditions without exposing production data unnecessarily. Capture the starting configuration, exact version, time of test and expected result. A pass requires evidence from the real workflow, not only a successful login or an unchanged dashboard.

Primary owner

Primary owner: The infrastructure or cloud platform lead, with the application owner for any workload proposed for a v7 move.

The named owner coordinates platform, application, commercial and business-process decisions. Contributors may perform the work, but accountability cannot be distributed across a meeting invite. The owner closes the test, exception and evidence record.

Action within seven days

Action within seven days: Before any v7 sizing appears in a budget or a migration plan, answer three questions per workload. Does anything write to the temporary drive, including SQL Server tempdb, page files and application scratch directories. Is the virtual machine Generation 2. Does the operating system image support NVMe. Any single no turns the move into a rebuild and changes the cost.

Start with the highest-consequence workload. Assign the people, date and pass criteria before the test begins. If the first test fails, record the failure as evidence and open remediation with a deadline rather than hiding it behind a general project status.

Evidence to retain

Evidence to retain: The dated inventory of candidate workloads with, for each one, the VM generation, the operating system image and its NVMe support, and a yes or no on temporary-drive use with the path that uses it. Add the region the workload must run in, and the written note of whether the new sizes are available there.

Store the evidence with the platform or change record. Include source exports and machine-readable results where possible. The next reviewer should be able to reproduce the conclusion without rebuilding it from email, chat or memory.

Frequently asked questions

How big are the new sizes?

Microsoft states the new VMs scale to 372 vCPUs and 2.8 TiB of memory. The Dsv7 size series page lists Standard_D248s_v7 at 248 vCPUs and 992 GiB, and Standard_D372s_v7 at 372 vCPUs and 1,488 GiB; the 2.8 TiB figure belongs to the memory-optimised Esv7 sizes.

Can we deploy them in South Africa North?

Not among the launch regions. Microsoft names Central US, East US, East US 2, Germany West Central, South Central US, Sweden Central, West US 2 and West US 3. No African region is listed. Microsoft has not published a statement about future regional expansion for these sizes, so nothing should be assumed either way.

Is there really no temporary drive?

Correct for the series. The Dsv7 page states that Dsv7-series virtual machines do not have any temporary storage, and the host specification lists local storage as None. Microsoft frames this as lowering the price of entry.

What happens to our paging file and tempdb?

They need somewhere else to live, normally an attached data disk. This is a configuration change to be planned and tested, not a side effect of resizing, and it changes the storage line on the bill.

Will our existing VM just resize onto one of these?

Only if it is Generation 2 and its image supports NVMe. The feature table marks Generation 1 VMs as not supported, and Microsoft states the series will only work on OS images that support NVMe, with an error message where support is missing.

How much faster are they?

Microsoft’s own claim is up to 20 percent better compute performance than prior-generation Intel-based v6 VMs. That is a comparison against the previous Intel generation, not against every alternative, and it is an upper bound rather than a guarantee for a specific workload.

Do they support premium storage and accelerated networking?

Yes. The feature support table lists Premium Storage, Premium Storage caching, Live Migration, Generation 2 VMs, accelerated networking and nested virtualization as supported. Memory-preserving updates are listed as not supported.

The Braintree view

Microsoft’s announcement supplies the platform fact. The customer control begins after that fact: identify the exposed process, name the owner, test the real dependency and retain a decision that can survive audit or staff turnover. Braintree can help structure the inventory, build the representative test and translate the result into a controlled implementation plan.

Use the seven-day action as the entry point. Do not wait for a renewal, support refusal or enforced update to reveal work that can be measured now.

Primary Microsoft sources

AI-use disclosure: AI supported research and drafting. Every factual claim was checked against the Microsoft sources listed here.

Related Posts

Follow one project from meeting preparation to a...
New models arrive almost weekly. Here is how...
Microsoft now prices Copilot in two parts. Learn...