A 99.9% uptime commitment leaves 43.2 minutes of downtime in a 30-day month, or 8.76 hours a year, by our arithmetic. Each extra nine cuts that by 90%: 99.99% leaves 4.32 minutes a month, and 99.999% about 26 seconds.
What the percentage pays depends on how the contract counts. Microsoft’s October 2026 Service Level Agreement pays a credit of 25%, 50% or 100% of the monthly fee as uptime falls below 99.9%, 99% and 95%. The two outages of 12-14 November show how far real incidents run past those lines.
Business Pill · Embedded Dependency
A short explainer on embedded dependency: what sits inside the systems you buy. It teaches the general idea only and does not discuss Workday or the outage.
The key insight: As we read it, an uptime percentage is only half a promise: the other half is the counting rule, and Microsoft’s user-minutes, Workday’s maintenance carve-out and Copilot’s prompt count can give three different answers for the same hour offline.
The arithmetic is (1 − uptime) times the minutes in the period. On a 30-day month of 43,200 minutes: 99% leaves 432 minutes, 99.5% leaves 216, 99.9% leaves 43.2, 99.95% leaves 21.6, 99.99% leaves 4.32 and 99.999% leaves 0.432. Over a 365-day year, 99% leaves 87.6 hours and 99.9% leaves 8.76 hours, all by our arithmetic.
Minutes of downtime each uptime level leaves in a 30-day month of 43,200 minutes. All figures are our arithmetic; the note gives this week’s Microsoft 365 and Workday outages for reference.
How Vendors Count Downtime
Microsoft counts in user-minutes. Its Service Level Agreement for Microsoft Online Services says downtime is “the sum of the length (in minutes) of each Incident that occurs during that Applicable Period multiplied by the number of users impacted by that Incident”. A tenant’s uptime therefore depends on how many of its users were affected, not only on how long the incident ran.
For Exchange Online, downtime is “Any period of time when users are unable to send or receive email with Outlook Web Access”, and “There is no Scheduled Downtime for this service”. Microsoft 365 Copilot is counted differently: by the share of valid prompts that get an unsuccessful response, with a first credit line at 99.5%.
Workday counts minutes for the service as a whole and sets planned maintenance aside. Its policy, in a copy filed in a City of Palm Desert agenda packet, takes total minutes less unplanned outage and planned maintenance, divided by total minutes less planned maintenance, and commits to 99.9%. It states: “Currently, Planned Maintenance is three (3) hours for weekly maintenance, plus four (4) hours for monthly maintenance, plus four (4) hours for quarterly maintenance”. Maintenance that overruns “is considered an Unplanned Outage”.
What Credits Pay, and What They Do Not
Microsoft’s table pays 25% below 99.9% monthly uptime, 50% below 99% and 100% below 95%, on “the total fees actually paid by you for a Service”. Credits “will not, under any circumstance, exceed your monthly service fees” for that service, and the SLA says: “Service Credits are your sole and exclusive remedy for any performance or availability issues for any Service under the Agreement and this SLA”.
Lost revenue is excluded: credits “will not be awarded to compensate for any other forms of loss, including but not limited to lost revenue, operational costs, or any indirect losses experienced by you or the end-users”. Credits also need a claim with the time and duration of downtime and “the number and location(s) of affected users”; Microsoft says it processes claims “typically within forty-five (45) days of receipt of the claim”.
This Week as a Worked Example
Microsoft’s record puts its 12-13 November incident at 1,430 minutes, about 33 times the 99.9% allowance, by our arithmetic. A user out the whole time ends November at about 96.7%, in the 50% band, as we set out in our Microsoft SLA piece. Workday’s 13-14 November outage ran about 1,065 minutes against roughly 42 minutes its policy leaves in November, as counted in our Workday piece.
The Structural Read
As we read the arithmetic, the gap between levels is steep: by our arithmetic, 99.9% allows 43.2 minutes a month and 99.99% allows 4.32, so one nine separates a long meeting from a coffee break.
As we read Microsoft’s terms, the credit is capped at one month’s fee for the affected service and excludes lost revenue, so the published remedy is sized to the subscription, not to the business loss.
Microsoft, Service Level Agreement for Microsoft Online Services, October 2026
“Service Credits are your sole and exclusive remedy for any performance or availability issues for any Service under the Agreement and this SLA”
Three Implications
FOR BUYERS Ask how downtime is counted (users, tenant or prompts) and what is excluded, as well as the percentage.
FOR FINANCE TEAMSMicrosoft’s credits apply to the fees actually paid for the affected service and need a claim.
FOR OPERATIONS TEAMS Workday’s policy excludes planned maintenance from the count; overruns count as unplanned outage.
The framework says: “The value proposition shifts from features to integration depth.”
As we read it, the SLA is where integration depth gets a price: the deeper a service sits in daily work, the more an outage costs relative to a credit capped at one month’s fee. The framework does not discuss SLAs.
This week’s two vendor outages against the downtime a 99.9% month leaves: 43.2 minutes in a 30-day month under Microsoft’s SLA and about 42 minutes in November under Workday’s policy text. All figures are our arithmetic.
What Is Not Established
Published terms are not every customer’s terms; enterprise agreements can differ. Microsoft’s SLA says its outage communications “are not a confirmation of missed Service Levels or Service Credits eligibility”, so an incident window does not by itself establish any customer’s uptime or credit. The Workday policy copy we read is one filed by a single city.
This Week In The AI Supercycle
AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source
This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.
By our arithmetic, 99.9% allows 43.2 minutes of downtime in a 30-day month and 8.76 hours a year, and each extra nine cuts that by 90%. Microsoft’s published credits run from 25% to 100% of the monthly fee and are its sole remedy. The week’s outages are counted in 3 Outages, 46 Hours in Two Days: Microsoft, Workday, the FAA.
A note on sourcing. Microsoft’s definitions, credit table, claim rules and remedy limits come from its Service Level Agreement for Microsoft Online Services, October 2026 edition. Workday’s formula and maintenance terms come from its service level policy as filed in a City of Palm Desert agenda packet; customer contracts can differ. The minute and hour allowances and the outage multiples are our arithmetic. This explainer will be updated if the published terms change.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
