Last week, the corporate world ground to a halt. If you run a business that relies on Microsoft 365, you probably spent November 12th and 13th staring at error screens, refreshing status pages, and wondering why your entire company’s communication pipeline had suddenly evaporated.
For roughly 48 hours, Exchange Online, Outlook, Teams, and Copilot went dark across more than 70 countries, impacting roughly 430 million users. Government departments stalled, courts suspended deadlines, pharmacies couldn’t process orders, and analysts estimated the lost productivity at somewhere around $7 billion. Even a hotel in Germany had a chemical dosing incident because its automated monitoring system relied on Outlook to send alert emails to staff—and those emails never arrived.
As someone who has spent over twenty years building custom backend systems, managing Linux servers, and running self-hosted mail infrastructure, I watched the whole thing unfold with a mix of sympathy for the IT staff on the ground and a very loud, “I told you so.”
Businesses have spent the last decade blindly migrating every single piece of their operational stack into single-vendor cloud ecosystems. They were sold a dream of “five nines” of uptime, effortless scalability, and zero maintenance. What they actually bought was a massive, single point of failure. When you hand the keys to your email, your internal chat, your file storage, and your identity management over to one tech giant without a backup plan, you aren’t running a resilient business. You’re gambling.
Here is why the “all-in-one cloud” model failed so spectacularly last week, why self-hosting and smaller providers actually beat the giants on resilience, and what you need to do before the next big outage hits.
Anatomy of a Single-Vendor Collapse
To understand why the November 12 outage was so brutal, you have to look at what actually broke. It wasn’t a sophisticated state-sponsored cyberattack. It wasn’t an AI gone rogue. According to Microsoft’s own post-mortem, it was a classic domino effect inside an overly centralized architecture.
A dormant software defect had been sitting in their networking gateway code since July. On November 12, a routine network capacity tweak shifted extra Teams traffic onto the same gateway servers handling Exchange Online. Combine that with a 14% reduction in server capacity from an unrelated issue the day before, and the gateway servers simply choked. Over 80% of them became unresponsive. And because the bug was months old, engineers couldn’t just hit “undo” on a recent deployment. When they finally did bring servers back online, the massive backlog of queued requests immediately crushed them again.
Notice the real problem here: blast radius.
Because Microsoft bundles everything together under one shared routing and authentication fabric, a networking hiccup in one area took down email, chat, calendars, and third-party tools all at once. Businesses didn’t just lose email; they lost their primary fallback for email (Teams). When your primary communication tool goes down, you usually message the team on chat to coordinate a workaround. When the same company owns both, you are reduced to texting people on their personal phones or, as workers in Japan famously had to do last week, dusting off the fax machines.
The Illusion of “The Cloud Has It Covered”
One of the biggest myths in modern business IT is that moving to a Tier-1 cloud provider absolves you of responsibility for backups and redundancy.
Ask ten small-to-medium business owners on Microsoft 365 how they back up their Outlook mailboxes, and nine of them will give you a blank stare. “Isn’t that what we pay Microsoft for?”
No, it isn’t. Microsoft operates on a Shared Responsibility Model. They are responsible for keeping the underlying hardware and software running (most of the time). You are responsible for your data.
When Exchange Online goes down for two days, you can’t even access your historical emails, your contacts, or your calendar invites unless you happen to have a local client cached in offline mode—and even then, you can’t send or receive anything new, nor can you easily spin up a temporary mail server if you don’t have an independent copy of your data and DNS control ready to go.
Compare that to how smaller hosting providers or competent self-hosters operate.
If you host your own mail stack—say, running Postfix and Dovecot on a lean Linux box—or if you use a dedicated, smaller independent provider, backups aren’t an expensive third-party add-on you forget to buy. They are baked into the baseline architecture. On a self-hosted setup, nightly automated snapshots of your Maildir or database to an off-site storage bucket (on a completely different provider, in a different region) take about twenty lines of bash scripting and cost pennies a month.
When you control the metal—or at least the virtual machine and the DNS records—you have options. If your primary server catches fire or a data center loses power, you can spin up a fresh instance, restore your mail directories from your off-site backup, flip an MX record in Cloudflare, and be back online before Microsoft’s status page has even moved from “Investigating” to “Identified.”
Don’t Put Your Eggs in One Basket
There is a reason old-school sysadmins hate monocultures. Diversity in your tech stack isn’t messy; it’s an insurance policy.
If you run a business, your operational survival shouldn’t depend on the uptime of a single corporation’s networking gateway. Embracing a “don’t put your eggs in one basket” philosophy doesn’t mean you have to build a data center in your basement, though I certainly prefer owning my own stack. It just means decoupling your critical services so a single failure doesn’t paralyze your entire company.
Here is what a sane, decoupled architecture looks like in practice:
- Separate your email from your chat: Never host your email and your instant messaging with the same vendor. If you insist on using Microsoft 365 for email, run Slack, Mattermost, or a self-hosted Matrix server for team chat. If Exchange dies, your team barely misses a beat internally.
- Keep your DNS independent: Never let your email or web host manage your domain’s DNS. Put your DNS with a rock-solid, independent provider like Cloudflare or Route 53. If your primary mail provider goes down for 48 hours, independent DNS lets you immediately reroute incoming mail to a backup MX server or a secondary catch-all inbox so customer orders and critical alerts don’t bounce into the void.
- Decouple your file storage: Don’t bury every single operational SOP, invoice, and contract exclusively inside SharePoint or OneDrive. Keep local network-attached storage (NAS) synced on-site, or use an independent cloud storage provider for critical disaster-recovery documents. If the cloud goes dark, your staff should still be able to open a local folder and find your emergency contact list and client records.
- Implement real, independent backups: If you stay on M365 or Google Workspace, set up an automated, off-platform backup pipeline. Even a simple Linux box running an IMAP sync tool (
getmailorofflineimap) pulling down copies of critical company mailboxes onto local storage every hour gives you an immediate, searchable, vendor-independent archive.
Why Self-Hosting Still Wins on Control
Every time I advocate for self-hosting or keeping infrastructure lean and independent, someone inevitably brings up maintenance overhead. “We don’t have time to manage servers; we just want to focus on our business.”
I get that argument. Not every company has a Linux admin on staff. But there is a massive middle ground between hiring a full-time data center team and surrendering your entire company’s nervous system to a black-box SaaS monopoly.
When you self-host—or work with a boutique engineering partner who builds lean, framework-free, self-hosted infrastructure for you—you trade a tiny bit of upfront setup for total operational visibility and control.
When something breaks on a self-hosted Linux stack running Nginx, MariaDB, and Postfix, you don’t have to sit helplessly refreshing a corporate Twitter feed hoping a billionaire’s company fixes a bug they introduced four months ago. You SSH into the box, read /var/log/mail.log, see the exact line causing the bottleneck, and fix it yourself in ten minutes.
Even better, self-hosted systems don’t suffer from the bloat and cross-service entanglement that plague enterprise cloud suites. A dedicated mail server only does mail. An update to your company’s video-calling software will never accidentally overload a networking gateway and knock your email offline for two days.
Take Back Your Infrastructure
The November 2026 Microsoft outage was a wake-up call, though history suggests most companies will hit the snooze button as soon as their inboxes start working again. It is easy to forget the panic of a two-day blackout once the backlog clears and life goes back to normal.
Don’t make that mistake. Look at your current setup this week and ask yourself one simple question: If our primary software vendor disappeared tomorrow morning, how long would it take us to send an email, access our files, and talk to our customers?
If the answer is “we couldn’t,” you don’t have a cloud strategy. You have a dependency. Start pulling your eggs out of that single basket. Back up your data to hardware you actually control, split your critical tools across independent providers, and stop trusting the cloud to save you from itself.
![]()