Somewhere in your estate there may be a small, unglamorous app that sends a few hundred thousand emails a day, password resets, order receipts, appointment reminders, the digital plumbing nobody thinks about until it stops. If it was built in the last few years by a developer who read the Azure docs, there is a decent chance it sends that mail through Azure Communication Services Email, not Exchange Online. And on September 30, 2028, that pipe gets capped.
Microsoft has announced it will retire most of Azure Communication Services on that date, Email Communication Services included. Two years feels like a lot. It isn’t, and the reason it isn’t is the claim I want to pressure-test, because the reassuring version of this story is going around already.
The claim: “Just move it to Exchange Online High Volume Email”
Here is the tidy narrative. ACS Email is going away, Microsoft has a high-volume email feature in Exchange Online, therefore you point the app at HVE and carry on. Clean. Same vendor, same portal, one less bill.
It’s the kind of claim that sounds right because both things have “email” and “high volume” in the name. So did margarine and butter. Let’s see if they spread the same.
What was ACS Email actually for?
Think of Exchange Online as a corporate mailroom. It’s superb at what it does, mailboxes, calendars, compliance holds, a real human’s identity attached to every message. But it was built for people writing to people. It throttles hard, it cares deeply about who you are, and it gets nervous when one sender suddenly emits a hundred thousand identical messages, because from a mailroom’s point of view that is exactly what a compromised account looks like.
ACS Email was the opposite animal: a loading dock. No mailboxes, no calendars, no identity, just an API endpoint, a verified sending domain, a connection string, and the ability to fling large volumes of transactional and application-generated mail at the internet. Developers reached for it precisely because it lived outside Exchange Online’s assumptions. That’s the whole point, and it’s the part the tidy narrative skips.
So is HVE the same loading dock?
Not really. Exchange Online High Volume Email is a genuinely useful thing, and if your ACS use was modest internal-leaning bulk mail, it may well be enough. But it was designed to solve the “line-of-business app needs to send internal notifications without burning a licensed mailbox” problem. Its comfort zone is mail to recipients inside your own organisation. Push it hard toward large-scale external sending, the newsletters, the customer receipts, the millions-of-recipients workloads ACS was happy to carry, and you’re back inside Exchange Online’s throttling, reputation, and volume ceilings. The same ceilings ACS existed to escape.
Margarine, not butter. Fine on some toast. Don’t build a croissant on it.
What isn’t retiring, and why that matters less than it sounds
“Most” of ACS is a deliberate word. Microsoft has signalled that some communication capabilities continue, particularly the ones that fold into the Teams and Teams Phone story, the calling and interoperability plumbing that has a clear home elsewhere in the platform. I’d treat the exact surviving-services list as something to confirm against Microsoft’s own retirement notice rather than take from any single blog, because these lists get revised and I’d rather you check than trust my memory of an announcement.
But here’s the thing to sit with: even in the best reading of what survives, email is not on the survivors’ list. Microsoft is exiting the dedicated, high-volume, external email business. After September 2028 there is no ACS-shaped replacement inside the Microsoft cloud. That’s not a migration between two Microsoft products. It’s an exit.
Where do you actually go?
For real high-volume transactional or marketing email, the honest answer is a third-party email service provider, the SendGrids, Amazon SESes, Postmarks, Mailguns of the world. That is not a failure of planning on your part; it’s the market Microsoft is stepping out of. Budget for it, because “we’ll absorb this into our existing M365 licences” is the assumption that will quietly blow up in a change-review meeting eighteen months from now.
The migration itself is rarely a one-line config change either. You’ll be re-verifying sending domains, re-doing SPF, DKIM, and DMARC records for a new provider, rewriting the app’s send code against a different SDK, and re-warming sender reputation from scratch. None of that is hard. All of it takes calendar time and DNS propagation and at least one round of “why did that batch land in spam.” Two years is not the runway. Two years minus the six months it takes to get budget, pick a vendor, and schedule DNS changes is the runway.
How do you find every place ACS is hiding?
ACS Email lives as Azure resources, Microsoft.Communication/emailServices, their child domains, and the parent communicationServices resources, so Azure Resource Graph is the fastest way to sweep every subscription at once. This is read-only; nothing here changes state.
That inventory is the easy half. The hard half is the part Resource Graph can’t see: which applications hold the connection strings. An ACS resource is worthless without an app calling its API, and those apps carry the endpoint and access key in app settings, Key Vault, or, the ones that will hurt, hardcoded in a repo. Grep your Key Vaults and App Service configurations for the ACS endpoint hostnames the query returns above, and talk to the dev teams. The resource is the smoke; the app is the fire.
The verdict, more precise than the headline
The headline is true: most of ACS shuts down on September 30, 2028, email included. The comforting follow-on, that Exchange Online quietly absorbs the load, is not. Exchange Online HVE is a real option for a narrow slice of what ACS did, and a trap for the rest. For genuine high-volume external mail, Microsoft is handing you a taxi voucher to a third-party provider, not a replacement service of its own.
Run the inventory this quarter. Not because 2028 is close, but because the first thing you’ll learn is how many teams didn’t know they were using ACS at all, and that number is always higher than the room expects.
