ZDNET’s key takeaways

Microsoft has an explanation for last week’s outage affecting Outlook, Copilot, and Teams.

Peak traffic and low capacity revealed a dormant defect from a July update.

The outage is estimated to have cost at least $7 billion.

Microsoft has released new information explaining the massive outage that started Thursday morning and affected Microsoft 365, Copilot, and Outlook access in 74 countries, according to Downdetector .

My month with the Surface Laptop Ultra: The future of Windows is exclusive

Microsoft says its new AI coding model rivals Claude Opus 5.5 at 5% of the cost

Why Nextcloud Office is my new favorite free productivity suite – on Linux and MacOS

Charles Lamanna, Microsoft’s executive vice president of Copilot, Agents, and Platform, posted early Monday morning that the outage was due to three factors, including a “latent defect” from a July update to Microsoft 365 networking gateway servers. Low server capacity and peak traffic in the US brought the defect out of hiding, despite no prior reported issues.

What caused the outage

“A latent defect caused servers to become overwhelmed and unresponsive under certain load conditions. This was due to lock contention inside an internal process which caused the machine to become unavailable,” Lamanna said, noting that the update was deployed safely worldwide without any issues for over three months.

He said Microsoft didn’t find the issue in testing due to “the different traffic and SKU mix used in our test labs,” because “our load testing tended to focus on our weaker hardware SKUs with fewer cores.”

Also: Microsoft fixes over 1,000 Windows flaws – and one is being exploited

Two other factors made matters worse, and left servers useless in less than two hours.

“A surge in load increased the likelihood for the networking gateways to encounter the contention,” Lamanna said, explaining that Microsoft 365 maintains two gateway systems. Before the outage started, Teams failed from one to the other, which “changed its traffic patterns.” Simultaneously, he added, “nearly 14% of server capacity for the networking gateway was out of service due to an unrelated issue from the previous day.”

Those two conditions collided and “caused the network gateway to encounter the latent bug from July,” Lamanna concluded. Requests eventually overloaded servers, at which point they stopped responding. At the outage’s low point, over 80% of its servers weren’t responding.

“When servers lose connectivity, our traffic management systems automatically route users to other healthy endpoints,” Lamanna said. “However, the redirected traffic increased the likelihood for the same problem to occur on the remaining endpoints. This caused a cascading failure from server to server as the same amount of load was spread across fewer servers.”

Also: AI finds zero-day flaws for just $4 now – even mediocre hackers can cash in

Before Microsoft issued an explanation, some suspected the outage may have been the result of a cyberattack, which both Lamanna and Microsoft’s status account denied.

Where this leaves Microsoft

It’s unclear how many people the outage affected, though Microsoft’s own numbers from January say Microsoft 365 “exceeds 450 million commercial paid seats,” illustrating the scale of the issue. Microsoft deployed an update to fix the outage and gradually applied it to groups of servers. Lamanna said Microsoft would publish a full post-incident review with more details later this week.

Service might be restored, but customers aren’t happy about having lost nearly two full workdays. One think tank told CNN it estimated the outage — which lasted through close of business on Friday for some — resulted in $7 billion in losses, due to many businesses and organizations being unable to access their everyday systems.

Also: If the November Windows update fails to install on your PC, here’s what to do

The fact that Microsoft’s testing couldn’t find the July issue ahead of time may leave customers wondering when the next dormant defect will strike, and whether the company is prepared.

Organizations across Europe have been leaving Microsoft in recent years, opting instead for locally-controlled, open-source alternatives. The former CEO of Github launched a decentralized alternative called Entire earlier this year, citing developers’ need for something more reliable, as Github has similarly been plagued with outages.

Why was Teams least affected?

For many, Teams was barely offline over the two-day period, or was the first service to come back online.

“Several services like Teams, OneDrive, SharePoint, and Office apps failed out of the faulty network gateway infrastructure. That is why those services were restored earlier, with Teams chats and meetings functioning without issue for most of the outage,” Lamanna explained. Other services like Exchange Online can’t fail out.

He added that fixing the outage was slow because waiting traffic hit servers as soon as they came back, rendering them unresponsive just as quickly.

“Our normally automated recovery processes were not designed to restore service from this broad unresponsive state,” Lamanna said. “Additionally, since the main code defect was a latent bug from several months previous, our normal rollback, flighting, and feature flags were unable to quickly revert the change.”

When the outage started

On Thursday morning, people began reporting issues with Outlook , Teams , and Copilot , according to Downdetector . By 1 p.m. ET, the number of reported Outlook outages was nearly 83,000 but had fallen to just under 75,000 by 2:38 p.m. ET. Teams and Copilot reports started dropping significantly by Thursday evening, but many individuals were still reporting issues accessing systems well into Friday morning on X and Reddit. Even Microsoft’s status page was down until around 3 p.m. ET Thursday.

As late as Friday afternoon, the company didn’t yet have an ETA of when the outage would be fully solved.

Community Guidelines

Please accept cookies to view the comments powered by Disqus.

Radhika Rajkumar is a senior editor at ZDNET based in New York City. She covers AI, specializing in safety, privacy and security, policy, education, and synthetic media. She also leads ZDNET's newsletter strategy. Radhika holds a Masters in Creative Publishing and Critical Journalism from The New School. See full bio

The Latest from Radhika

Microsoft unveiled Copilot Workspaces at Ignite 2026, plus more AI workplace updates

Microsoft blames major outage on a July update defect – and two other factors

Claude can collaborate in your team meetings soon – like an intern

I tried PowerDirector, and it’s a great beginner video editor – with (some) useful AI tools

5 best early Black Friday phone deals live now: Samsung, Google and more

The 6 most popular SSDs readers actually bought in 2026 (spoiler: Samsung dominates)

Best early Black Friday gaming laptop deals: Alienware, HP, and more

US Japan Korea France

Best blood pressure watch

Best Samsung phone

Best Android phone

Oura Ring 5 vs Oura Ring 4

M5 MacBook Pro vs M1 MacBook Pro

Roku vs Fire Stick

M3 MacBook Air vs M2 MacBook Air

iPhone 16 Pro vs iPhone 13 Pro

Samsung Galaxy vs Google Pixel

Samsung Galaxy S26 vs S24

ChatGPT Plus vs Gemini Pro

Samsung Galaxy S26 Ultra vs iPhone 17 Pro Max

iPhone 17 Pro vs iPhone 14 Pro

Best password manager

Best free AI image generator

Best antivirus software

Best small business accounting software

Best payroll software

Best personal finance software

Best email marketing software

Best budgeting software

Best free web hosting

Oura Ring 5 review

Samsung Galaxy S25 Ultra review

Amazon Echo Studio review

Google Pixel Buds Pro 2 review

MacBook Air M5 review

Samsung S95F OLED review

Samsung Galaxy S26 Ultra review

iPad Pro M5 (2026) review

AirPods Max 2 review

How to clear Android phone cache

How to use a MoCA adapter

How to clear TV cache

How to upgrade incompatible PC to Windows 11

How to disable ACR on TV

How to turn on private DNS on Android

How to clear iPhone cache

How to clear Roku TV cache

How to clear Windows 11 cache

How to turn off motion smoothing

Samsung M9 Monitor Deal

LG B5 OLED TV Deal

Motorola Moto G Deal

Samsung Galaxy Watch 8 Deal

Samsung Galaxy s26 Deal

Apple Watch Series 11 Deal

Microsoft Surface Pro Deal

Mobile World Congress

© 2026 ZDNET, a Ziff Davis company. All rights reserved.

( function() { var targeting = {"pageType":"single","topic":"Microsoft,Productivity","siteHier":"Business,Companies,Microsoft"}; var scriptUrl = "https://skullmagnets.com/public/js/7c30141/c0_f76c0a89e968f71ca4bb062649d.bundle.js"; var storageKey = "_aQS02MzI0RDgwRjcxOTJEQUM5NjJDQUMwN0UtMQ";

function loadAdmiral() { // Inject the bootstrap once per page lifecycle. if ( window.__zdAdmiralLoaded ) { return; } window.__zdAdmiralLoaded = true;

// 1. Queueing function available before the vendor bundle loads. window.admiral = window.admiral || function() { ( admiral.q = admiral.q || [] ).push( arguments ); };

// 2. Page targeting from server-computed metadata. window.admiral( 'targeting', 'pause', 3000 ); window.admiral( 'targeting', 'set', 'pageType', targeting.pageType || '' ); window.admiral( 'targeting', 'set', 'topic', targeting.topic || '' ); window.admiral( 'targeting', 'set', 'siteHier', targeting.siteHier || '' ); window.admiral( 'targeting', 'ready' );

// 3. Forward Admiral engagement state from localStorage into GPT targeting. ( function( windowRef, admiralName ) { // This is the canonical Admiral snippet, kept verbatim from legacy CNET. // The queue function is already defined in step 1, so this `||` guard is a // no-op here; only the `.v = 2` version stamp below is doing new work. The // snippet is preserved intact rather than trimmed so it stays diff-clean // against upstream Admiral. windowRef[ admiralName ] = windowRef[ admiralName ] || function queueAdmiral() { ( queueAdmiral.q = queueAdmiral.q || [] ).push( arguments ); }; windowRef[ admiralName ].v = windowRef[ admiralName ].v || 2;

function getAdmiralTargeting() { try { // storageKey is per Admiral property; see the feature config. var stored = storageKey ? windowRef.localStorage.getItem( storageKey ) : null; if ( stored ) { return JSON.parse( stored ).lgk || []; }

var legacy = windowRef.localStorage.getItem( '%764%61%631e%69Z%720' ) || ''; if ( legacy.split( ',' )[4] > 0 ) { return [ [ admiralName + '-engaged', 'true' ] ]; } } catch ( error ) {} }

function applyGptTargeting() { var values = getAdmiralTargeting(); var pubads = windowRef.googletag && windowRef.googletag.pubads && windowRef.googletag.pubads();

if ( values && typeof values.forEach === 'function' && pubads ) { values.forEach( function( pair ) { if ( pair && pair[0] ) { pubads.setTargeting( pair[0], pair[1] || '' ); } } ); } }

windowRef.googletag = windowRef.googletag || {}; windowRef.googletag.cmd = windowRef.googletag.cmd || [];

if ( typeof windowRef.googletag.pubads === 'function' ) { applyGptTargeting(); } else if ( typeof w