.NET 8 and .NET 9 both reached end of support on November 10, 2026, according to the official .NET support policy. They no longer receive security fixes from Microsoft. The recommended target is .NET 10 LTS, supported until November 14, 2028, reached by retargeting each project to net10.0, clearing the .NET 10 breaking changes behind a test suite, and rolling out service by service with the old build kept ready for rollback.
Why do an LTS and an STS release end on the same day? The policy now gives STS releases two years: "LTS releases get free support and patches for three years. STS releases get free support and patches for two years." .NET 9 shipped a year after .NET 8, so both landed on November 10, 2026. The same arithmetic puts .NET 10 LTS and .NET 11 STS on one shared end date too.
Dates are from the official .NET support policy and release metadata, checked November 21, 2026.
We recommend .NET 10 over .NET 11 for most production estates. Both end on the same date, but .NET 10 has a year of patch history behind it, its breaking-change list is settled, and most third-party libraries you depend on have had time to ship net10.0 builds. .NET 11 makes sense only when you need a specific feature it adds and you accept being early.
Microsoft defines end of support as the date it "no longer provides fixes, updates, or online technical assistance." The policy page goes further and says out-of-support versions "may put your applications, application data, and computing environment at risk." Final patches 8.0.34 and 9.0.23 are the last builds either line will get. Any vulnerability disclosed after November 10 in the runtime, ASP.NET Core or the base class libraries will be fixed in supported versions only.
The same page sets a condition that applies even before you upgrade: "Within a release's support lifecycle, systems must remain current on released patch updates." If you were still on an older 8.0.x patch when the line ended, you were already outside support. Getting onto 8.0.34 or 9.0.23 now is a sensible holding step while the real move happens, but it does not change the status.
Auditors and buyers will notice through your vulnerability process. SOC 2's CC7.1 asks companies to detect and manage vulnerabilities; with .NET 8 out of support, a CVE in the shared framework has no Microsoft patch to apply, so your evidence becomes a compensating control or an upgrade ticket. See SOC 2 for AI-assisted codebases for how we read that criterion against a codebase. Enterprise security questionnaires usually ask whether every runtime is vendor supported, and for net8.0 the honest answer is now no.
Library support will drift as well. Package authors tend to follow the official lifecycle, so expect new releases of the libraries you rely on to target net10.0 and later. That is a pattern to watch, not a date anyone has published, so check each critical dependency yourself.
Retargeting the obvious solution is the easy part; the stragglers live in build infrastructure. Go through this list and put every hit in a single spreadsheet.
Shortcut: the free EOL deadline radar takes a pasted .csproj, global.json, Dockerfile or lockfile and returns a dated report, flagging net8.0 and net9.0 targets against Microsoft's published end dates.
Microsoft keeps a breaking changes in .NET 10 list, sorted into binary incompatible, source incompatible and behavioral changes, and it notes the list is not complete. These are the items we would check first on a .NET 8 or 9 codebase. Teams on .NET 9 skip one release of changes; teams on .NET 8 should also read the .NET 9 list.
The default container tags now use Ubuntu: sdk:10.0 refers to Ubuntu 24.04 "Noble Numbat". .NET 9 and earlier defaults were Debian, and Debian images are not shipped for .NET 10 at all. Any Dockerfile step that installs packages with distro-specific names, or assumes Debian paths, needs a test. If you must stay on Debian, Microsoft's guidance is to build and maintain your own images.
On Unix, the runtime no longer installs a SIGTERM handler, so the OS default terminates the process immediately and ProcessExit is not raised. Typical ASP.NET Core apps and anything on the generic host with UseConsoleLifetime are unaffected. Hand-rolled console workers and queue consumers in containers are the risk: if they relied on ProcessExit to drain work during a rolling deploy, register a handler with PosixSignalRegistration.Create.
For projects targeting .NET 10, NuGetAuditMode defaults to all, so restore warns about vulnerable transitive packages, not just direct ones. With TreatWarningsAsErrors on, restore can fail the build on the first day of the upgrade. Microsoft's options: exempt NU1901 through NU1904 from warnings-as-errors, use dotnet nuget why to find the parent package and upgrade it, promote the transitive package to a direct reference, or suppress a specific advisory. We prefer fixing over suppressing, because these warnings are the exposure you are upgrading to remove.
In ASP.NET Core 10, unauthenticated requests to known API endpoints get 401 and 403 instead of a redirect to the login page. Known endpoints include [ApiController] controllers, minimal APIs that read or write JSON, TypedResults endpoints and SignalR. This is usually what you wanted, but front ends and integration tests that expect a 302 will break. The switch Microsoft.AspNetCore.Authentication.Cookies.IgnoreRedirectMetadata restores the old behavior app-wide, and AllowCookieRedirect restores it per endpoint.
WebHostBuilder, IWebHost and WebHost are obsolete and raise ASPDEPR004 and ASPDEPR008, which fail builds that treat warnings as errors. The fix is HostBuilder or WebApplicationBuilder, and test code that builds a TestServer from a WebHostBuilder needs the same change. The ASP.NET Core 10 list also marks Razor runtime compilation, IPNetwork with ForwardedHeadersOptions.KnownNetworks, and WithOpenApi as obsolete or deprecated.
The EF Core 10 breaking changes are mostly rated low impact, but several touch production query behavior. Contains over a local collection now translates to multiple scalar parameters instead of a JSON array with OPENJSON; UseParameterizedCollectionMode reverts it if a query plan gets worse. Parameter names change from @__city_0 to @city, and Microsoft warns this can make almost all cached plans recompile, with a temporary compilation spike right after deployment. With UseAzureSql or compatibility level 170, JSON columns map to the new json type and the upgrade generates a migration to alter them. EF tools now require --framework on multi-targeted projects, which breaks migration scripts in CI that never passed it.
.NET 10 requires OpenSSL 1.1.1 or later on Unix, and an app without it fails to start. Official images are fine. Old self-built images and long-lived VMs are where this bites.
For a .NET 8 or 9 codebase this is a retarget, not a rewrite. The hard part is catching behavioral changes that compile cleanly. If you still have .NET Framework 4.x projects in the same estate, those are a different job, covered in our .NET Framework to .NET 10 runbook.
Turn the exposure list into one row per deployable: TFM, SDK pin, base image, host, owner and test coverage. Move everything to the final patches, 8.0.34 or 9.0.23, so the baseline is the newest build of what you run. Capture p95 latency, error rate, memory and SQL query counts per service, because EF Core and GC changes show up as numbers, not exceptions.
Update global.json to a .NET 10 SDK, set TargetFramework to net10.0, and bump Microsoft.AspNetCore.*, Microsoft.EntityFrameworkCore.*, Microsoft.Extensions.* and System.Net.Http.Json to 10.0.0 or later, as the official ASP.NET Core 9 to 10 migration guide lays out. Upgrade shared libraries first. The legacy .NET Upgrade Assistant is officially deprecated; Microsoft now points to GitHub Copilot upgrade, whose .NET version upgrade scenario writes an assessment, plan and task list into the repo and commits each step so you can review or roll back. Whichever tool you use, a human reviews each diff against the breaking-change list above.
Ship the net10.0 build of one low-risk service first, on the new Ubuntu-based image, behind a canary or a small traffic slice. Compare it to the baseline on latency, errors, memory and database CPU. Watch shutdown behavior during deploys, auth responses from API clients, and query plan churn on the database. Then move through services in dependency order.
Keep the last 8.0.34 or 9.0.23 image tagged and deployable until each service has been stable on .NET 10 through a normal release cycle. Rollback is an image swap as long as you have not applied an EF Core migration that the old code cannot read, so ship schema changes separately and make them backward compatible. The json column alteration on Azure SQL is the one to schedule on purpose.
Remove 8.0 and 9.0 SDKs and runtimes from build agents and hosts, delete compatibility switches you added for the cutover once the code is fixed, retire temporary NuGet audit suppressions, and update CI so a new project cannot target net8.0 again.
Staff it around the risks above. One lead keeps the service list and sequencing. C# engineers retarget projects, fix ASPDEPR warnings and update tests. A platform engineer owns Ubuntu images, build agents and the canary. If EF Core fronts a busy SQL Server or Azure SQL database, add someone comfortable reading execution plans before and after. Many teams begin with a pod of 3 to 5 senior engineers (a tech lead plus engineers), and a full pod is about seven. Ryz engineers work on your team, reporting to your leads, alongside the people who know your services. Bring in .NET developers for the retargeting and DevOps engineers for pipelines; DevOps consulting fits when the CI estate itself needs rework.
Ryz engineers on your team start from $7,000 per month (our published range is $7,000–$15,000 per engineer per month, mostly mid-level to senior), quoted per team. Once you say go, engineers can typically start in under two weeks. Plenty of .NET teams handle a one-release retarget on their own, and that is fine. When the backlog will not allow it, scope your team with us.
Apps keep running on the final builds, 8.0.34 and 9.0.23, but Microsoft stopped servicing both lines on November 10, 2026 and its policy strongly recommends against out-of-support software. New runtime CVEs will be patched only in .NET 10 and 11. Get onto those final patches today and schedule the retarget.
.NET 10 LTS, supported until November 14, 2028. .NET 11 STS ends on the same date, but .NET 10 has a longer patch history and wider library support, so it is the safer production target for most teams.
Count of deployables, test coverage, reliance on EF Core query shapes, custom images, TreatWarningsAsErrors, and any dependency lacking a net10.0 build are the drivers. A tested ASP.NET Core service is a retarget plus a canary; untested workers on old self-built images need more care.
Microsoft's .NET support policy page offers nothing beyond the end date for .NET 8 or 9 and says nothing about a paid extension. Linux distributions or cloud platforms that package .NET may have their own terms, so ask your vendor, and plan the upgrade regardless.
Our engineers cost $7,000 to $15,000 each per month depending on seniority, mostly mid-level to senior, and leads are quoted per team. A project's total is engineers times months times rate. You receive the scoped plan, the price and named people up front.
Questions we didn't answer? Email info@ryzlabs.com.
Thanks — your message has been sent. We’ll get back to you soon.
Something went wrong while sending your message. Please try again or email info@ryzlabs.com.