Blog / Tech, New & Old /

The CrowdStrike outage: how one update took down millions of PCs

On Friday 19 July 2024, airport departure boards went blank, hospital systems stopped, banks and broadcasters struggled, and a lot of Windows computers showed the blue screen of death. The cause wasn’t a cyberattack. It was a software update from a security company, and it took down millions of machines in a few hours.

Here’s what happened, in plain words, and what it means for the rest of us.

What happened

CrowdStrike makes a security product called Falcon. It runs on company computers and servers, watches for suspicious behaviour and gets updates from CrowdStrike regularly. On 19 July, one of those updates, a content file for Windows machines, contained a fault. When Windows computers running Falcon received it, they crashed. Many then got stuck rebooting and crashing again.

CrowdStrike said the problem was not a security incident or attack and that it had identified the cause and was working on a fix. Microsoft said roughly 8.5 million Windows devices were affected. That’s a small share of all Windows machines worldwide, but they were concentrated in large organisations, so the effects were very visible: airlines, airports, hospitals, banks, shops and TV channels reported disruption.

Windows machines without Falcon weren’t affected. Macs and Linux machines weren’t hit by this particular fault either. A separate cloud outage at Microsoft hit some services around the same time, which muddied early reports, but that was a different problem.

Why a security tool could crash a computer

To spot threats, security software needs deep access to the operating system. It runs very close to the core of Windows, at a level where a mistake can bring down the whole machine rather than just one program.

Think of a building’s security guard who has a master key and permission to enter every room. That’s useful. But if the guard trips and falls in the main corridor, nobody can get through, however good their intentions.

The faulty update was crashing Windows at start-up, so the machines couldn’t even load far enough to receive a corrected update from the internet.

Why fixing it was so slow

This is the part that surprised many people. A normal software bug is fixed by pushing a corrected update. Here, affected PCs couldn’t boot properly to receive it.

The practical fix was manual. Someone had to start each affected machine in Safe Mode or the Windows recovery environment and delete one faulty file. For a company with thousands of laptops, point-of-sale terminals or servers, that means a person touching each machine. Where disks were encrypted with BitLocker, staff needed recovery keys first, which made it slower still.

It also explains why the damage hit the places it did. Airlines and hospitals run on many connected machines, and some sit in locations where nobody technical happens to be standing.

What’s actually new here

Software updates breaking things isn’t new. What stands out:

  • Scale. One vendor’s update reached a huge number of machines at once.
  • Depth. Security tools sit low in the system, so a bug there is a crash rather than an error message.
  • Automatic delivery. Many organisations let security content flow in without staging it first, because speed matters for security.

That’s the trade-off at the centre of the story. Fast updates protect you from new threats. Slow, tested updates protect you from bad updates. You can’t maximise both.

Who should care

  • Anyone who manages a business’s IT, even informally. Ask how updates are rolled out and whether there is a way to test them on a few machines first.
  • Anyone who runs websites or online services. Your host, DNS provider, payment gateway and plugins all depend on someone else’s updates. This is a reminder to know what you depend on.
  • Regular users. If your work PC was stuck on a blue screen, there was nothing you did wrong, and nothing you could easily fix yourself. Wait for IT to repair it.

If you got caught up in this on a Mac or on your phone, you probably saw something else, like a delayed flight or a card terminal that wouldn’t work, rather than a crash.

My take

This is an opinion, not a verdict. I don’t think the lesson is “stop updating”, and I don’t think it’s only “blame the vendor”. It’s that resilience is boring and necessary: staged rollouts, a manual fallback, and knowing how to run the business without a system for a few hours.

For small teams, the equivalent is cheap. Keep paper or offline copies of anything critical. Know who to call. Keep backups you’ve actually tested. I’d treat monitoring a website with an automated check the same way, as a basic habit that you’ll be glad of the day something odd happens.

The tech is also bigger than any single company. Modern computing depends on a few widely shared tools, and when one fails, a lot fails together.

What to watch next

  • A fuller explanation from CrowdStrike of how the faulty file passed its testing.
  • Whether customers change how they stage security updates.
  • Whether regulators and large buyers ask for more rigorous release practices from critical-software vendors.

What small teams can do this week

You don’t need a security department to learn from this. Three cheap habits go a long way:

  1. Write down your critical systems. What would stop the business if it failed? Many small teams find the list is shorter than they feared, and some items have no fallback at all.
  2. Stagger your own updates. If you manage several computers or sites, update one first, wait a day, then do the rest. It costs very little and catches a surprising number of bad updates.
  3. Keep recovery details somewhere that isn’t the thing that broke. Encryption keys, admin logins and the number for your IT contact are no use if they live on the machine that won’t start.

Key takeaways

  • A faulty content update from CrowdStrike’s Falcon product crashed an estimated 8.5 million Windows devices on 19 July 2024.
  • It wasn’t a cyberattack, according to CrowdStrike.
  • Machines couldn’t boot, so recovery needed hands-on fixes.
  • Speed of updates and safety of updates pull in opposite directions.
  • Know what you depend on, and have a plan for when it fails.

What would your business do if its most important computer wouldn’t start tomorrow morning?