Outlook down at least 12 hours: what the SLA credit pays

Exchange Online stopped letting people open their mailboxes at 16:17 UTC on Thursday, under incident EX1550543, which Microsoft first reported as affecting users in the Americas before it spread. Two hours later Microsoft opened a second incident, MO1550821, for everything that reads from Exchange: Teams calendars, Safe Links, Purview labeling, Graph calls for mail and contacts, Copilot, Planner, Intune. At 04:57 UTC on Friday it was still adding virtual machines to carry the load, with some of them stuck in a bad state. By then 12 hours and 40 minutes had gone by, and Microsoft had not called the incident over, so that span is a floor and not a duration.

So the question on a Friday afternoon is no longer what broke. It is what the contract pays, and what you have to do to collect it.

The short answer: a tenant whose users could not get at their mail for that stretch lands in the 50 percent tier of the Exchange Online service level agreement. On Microsoft's own list prices that is worth $2.00 a seat if you buy Exchange Online Plan 1, and nothing at all if you do not file a claim before the end of December.

The three numbers that decide the tier

Exchange Online promises 99.9 percent uptime in a calendar month. The SLA measures downtime in user-minutes, and defines User Minutes as every minute in the month, less any scheduled downtime, multiplied by every user. Exchange Online has no scheduled downtime under the agreement, so for this service it is the full month. November has 30 days, so each seat carries 43,200 minutes.

Work the thresholds from that figure and the tiers stop being abstract:

• More than 43.2 minutes of lost mailbox access in the month takes you under 99.9 percent, which pays 25 percent of the fee. Exactly 43.2 minutes is 99.9 percent, which pays nothing.
• More than 432 minutes, 7 hours and 12 minutes, takes you under 99 percent, which pays 50 percent.
• More than 2,160 minutes, 36 hours, takes you under 95 percent, which pays the whole month back.

Thursday 16:17 UTC to Friday 04:57 UTC is 760 minutes. For a seat that was down the whole way, uptime for November comes out at 98.24 percent, which is the 50 percent tier with room to spare. Rudy Mens at LazyAdmin reaches the same tier. Note what the clock measures, though: the SLA counts the specific failure of being unable to send or receive mail through Outlook Web Access. The longer list of degraded features in MO1550821 is not what the meter reads.

What 50 percent is worth

Microsoft lists Exchange Online Plan 1 at $4.00 per user per month paid yearly, and Plan 2 at $8.00. Half of that is the credit:

Plan 1 pays $2.00 a seat, so $400 for a 200-seat company and $4,000 for a 2,000-seat one. Plan 2 pays $4.00 a seat, so $800 and $8,000. Microsoft 365 E3 lists at $39.00 and E5 at $60.00, and if the whole suite fee counted, E3 would pay $19.50 a seat, which is $39,000 across 2,000 seats, and E5 would pay $30.00 a seat, which is $60,000.

It will not be the whole suite fee. The SLA restricts a credit to the fees paid for the particular service whose service level was missed, and caps it so that it can never exceed your monthly fee for that service. Microsoft publishes no Exchange component price inside E3 or E5. LazyAdmin says the calculation uses the Exchange share of the licence price; Microsoft's document does not say how that share is worked out. For a suite customer the honest range is therefore wide, with the standalone Exchange price at the bottom end, and the person who decides where in that range you land is Microsoft, after you claim.

What the agreement will not cover

Three limits matter more than the tier.

The credit is the only remedy. The SLA calls service credits the sole and exclusive remedy for availability problems and says you may not unilaterally offset your fees against them. It also rules out compensation for lost revenue, operational costs or any indirect loss, which covers almost everything a day of dead email actually costs a business.

Degradation without unavailability pays nothing on the uptime meter. The SLA excludes performance and latency problems where the service was still available. Slow mail has a separate performance level: credits begin when the fastest 95 percent of delivery takes more than a minute, at 25 percent, then 4 minutes and 10 minutes for the higher tiers. If your users had mail arriving late rather than not at all, that is the table to read.

Microsoft's outage posts are not evidence. The agreement states plainly that outage communications are meant to help customers protect their own systems and are not a confirmation that a service level was missed or that credits are owed. The admin center advisory you screenshotted on Thursday is a starting point for a claim, not the claim.

What IT teams should do

• Pull your own evidence now, while the mail logs still cover it. A claim must carry a description of the incident, the start and end times of the downtime, the affected resource names, the number and locations of affected users, and a description of the errors. The SLA says an incomplete claim is rejected, not queried.
• Fix your own end time per tenant rather than copying a news timeline. Recovery came in waves: servers that came back around 21:08 UTC went unresponsive again at 00:00 UTC once they took traffic. Your minutes are the ones you can show.
• Raise the claim in the admin center, as a service request. LazyAdmin puts the deadline at December 31, the end of the month after the outage, and says the credit arrives on a future invoice rather than as a refund. That date is LazyAdmin's, not a term we can quote from the SLA text read here, which sets out what a claim must contain. Confirm the deadline against the edition of the SLA that governs your own term, and work to the end of December rather than the new year.
• If you buy through a CSP partner, the claim goes through the partner, who may file one claim covering every customer. Ask yours today whether it is filing, and for your tenant's downtime figures.
• Ask your account team, in writing, which component price it will apply to a suite seat. That single answer is the difference between $2.00 and $19.50 a seat on E3, and it is better settled before the claim than argued after it.
• Read the version of the SLA that governs your term. The copy read in full for this story is a reseller-hosted archive of the January 1, 2026 edition; Microsoft's own licensing library lists an October 2026 English edition as the current one, and the terms that apply to you are those in force when your subscription renewed.

One thing to leave out of the claim: the exploited Outlook flaw patched on Tuesday, CVE-2026-124194, is unrelated. It was one of 1,094 fixes in the November updates, and Microsoft's service advisories do not connect it to this outage. Microsoft has described a failing mechanism, a process that sets up secure connections becoming overwhelmed on servers that were also failing automated health checks, but it has declared no root cause, reports no indication of malicious activity, and says the post-incident report is where the explanation will appear.