Microsoft's 100% Service Credit Cap Leaves Business Losses With Customers
A cloud company sells deeper dependence. Its standard remedy still prices failure as a fraction of one monthly service bill.
Some Microsoft servers failed automated health checks on 12 November 2026. An alternative check later brought a subset back. The servers got a second opinion; the standard contract stayed put. An outage interrupts a working day. The contract measures the subscription.
Microsoft's standard service-level agreement, dated 1 October 2026, calls service credits the "sole and exclusive remedy" for availability problems. It caps those credits at 100% of affected monthly fees and excludes lost revenue and operating costs. The Fee Firewall is the contractual boundary between what the supplier credits and what the customer loses when work stops. Microsoft sells integrated work while leaving its interruption costs with the buyer. You pay for access. You carry the business risk.
In the Form 10-Q for the quarter ended 30 September 2026, Microsoft 365 cloud revenue rose 17%, to $27,778 million from $23,694 million a year earlier. Commercial seats rose 5%. Premium offerings included Copilot, E5 and E7. They helped push revenue per user higher. You see the pricing power. Every new function sits under the same roof. It puts more of the working day behind one supplier.
The 12 November 2026 status updates began with Exchange Online mailbox access problems in the Americas. Microsoft then warned of possible connectivity trouble in any region. Elevated load across multiple regions reflected unavailable servers, rather than rising traffic. It paused deployments, moved traffic and validated a different health check. Its cash additions to property and equipment were $39,367 million in the quarter ended 30 September 2026, according to the 10-Q. That covered Microsoft broadly, rather than a dedicated budget for this service. The machines still depend on judgments by other machines.
For Exchange Online, the October 2026 SLA defines downtime around the inability to send or receive email with Outlook Web Access, then measures downtime in user-minutes across affected users. A failure confined to another mail client falls outside that definition. A public incident notice and contractual downtime are different animals. You might lose access before the SLA starts counting. The distinction decides when the fee protection starts.
November 2026 has 30 days, or 30 × 24 × 60 = 43,200 minutes. The 1 October 2026 Exchange Online SLA awards 25% below 99.9% availability, meaning more than 0.1% × 43,200 = 43.2 average downtime minutes per user. Below 99% earns 50%; below 95% earns 100%. In a scenario with eight qualifying hours lost by every covered user, 8 × 60 ÷ 43,200 = 1.11% downtime and a 50% monthly credit. Fair money, for the fee. The workday has another price.
Microsoft has an enormous denominator for these refunds. Its 10-Q filed on 28 October 2026 lists $27,778 million in Microsoft 365 cloud revenue for the quarter ended 30 September 2026. That quarter's operating income was $45,915 million. In a scenario granting a 25% credit on 1% of implied monthly revenue, $27,778 million ÷ 3 × 1% × 25% = $23.1 million. That is $23.1 million ÷ $45,915 million = 0.05% of quarterly operating income. The revenue base includes products outside the covered services, overstating eligible fees. This is a scenario, never a record of credits paid.
The credit has a deadline. The 1 October 2026 SLA requires claims for a November incident to arrive by 31 December 2026 and says Microsoft typically processes them within 45 days after receipt. An outage notice confers no eligibility by itself; customers must document the missed service level. Some licensing programs receive service days instead of monetary credits. You might call that paperwork. It also allocates the burden of proof to the interrupted business.
Evercore ISI's Kirk Materne, as reported by MarketWatch on 13 November 2026, called his calculation a "rough estimate" and the impact "benign." His assumptions were perhaps 100 million affected paid seats, a 50% credit and $4 to $8 in monthly per-user fees. The scenario arithmetic is 100 million × $4 × 50% = $200 million; 100 million × $8 × 50% = $400 million. Against Microsoft's $92,279 million in revenue for the quarter ended 30 September 2026, either scenario looks manageable. The seats are an analyst assumption, not a verified affected count.
A Forrester study commissioned by Microsoft in January 2025 quantifies the appeal of integration. The composite organization had 30,000 employees. For that $10 billion revenue business, the study modeled 197% return on investment and $101.6 million of net present value from Microsoft 365 E3. It included "Vendor consolidation savings of $60 per user per month." You would count those savings before arguing an integrated suite is a bad purchase. This is a pre-incident commissioned model, rather than a measure of November 2026 losses.
Some customers demand more than service credits. In its Form 8-K dated 8 August 2024, Delta Air Lines attributed roughly 7,000 cancellations over five days to the CrowdStrike-caused outage of 19 July 2024. It estimated a $380 million direct revenue impact for the September 2024 quarter and said it was pursuing at least $500 million in damages from CrowdStrike and Microsoft. Delta pursued compensation measured by business damage, not subscription fees. The outage was CrowdStrike-caused; its filing gives neither a governing SLA nor the claim's outcome. Demand and recovery are different things.
A standard SLA prices availability rather than business-interruption insurance. Negotiated agreements, insurance and litigation sit outside it. Microsoft's actual November 2026 credits, customer losses and final root cause remain undisclosed. Still, Microsoft's 1 October 2026 terms settle the standard contractual arithmetic: at most 100% of monthly affected-service fees, with operational costs and lost revenue excluded. The eight-hour scenario returned 50% of a fee. The Fee Firewall governs the routine refund, regardless of the value of stranded work.
The 28 October 2026 earnings call illustrates the buyer's dilemma. Microsoft said Volkswagen planned to roll out Copilot to more than 330,000 employees. You get common tools and workflows; you also concentrate failure exposure. An alternative relying on the same unavailable identity or service is redundancy in name only. Real backup needs independent communications, authentication or recovery. Microsoft keeps the integration revenue. The customer buys the escape route.
I would hold Microsoft shares while its standard 100% credit ceiling holds, and buy an independent recovery path before adding another enterprise workload. The Fee Firewall protects the vendor's subscription economics while leaving the interrupted business with the buyer. A Microsoft disclosure by the end of March 2027 of one approved claim for the November outage under that standard SLA exceeding 100% of the affected service's monthly fees would prove me wrong. Price the two risks separately.