Backblaze, Inc. Class A Common Stock Status update

Hi, everyone. Welcome back to our quarterly Drive Stats report. My name is Stephanie Doyle, and I am the Senior Manager of Market Intelligence and keeper of stats. Joining me today is David Johnson, who is going to walk through this quarter's Drive Stats report, as well as we are going to get into a discussion of some of the drive technologies on the rise in existence that are bringing the industry to what it is today and will be tomorrow. David, you want to introduce yourself?

Yeah. Thanks, Stephanie. It is a pleasure to be here today. I have followed Drive Stats for a very long time. I have been in tech for a while now, so it has just been like a backbone to me. I think we are in our 13th year of Drive Stats, I believe, so we are still kicking strong.

Yeah, it is 13 plus. We started tracking the data, and then we did our first report with a little bit of a gap. If you look at the full data set, I think we are not quite at 14. But if you look at where the report started, we are at 13 solidly. So it is a data set that only keeps on growing, and definitely keeps on giving. Always fun to do these reports. A little housekeeping for folks who have joined us before or for joining us for the first time today, webinar is being recorded. Of course, please ask questions, and there will be attachments that we add at the end, including the slide deck from today. You can always catch the blog and all that other good stuff on our website. We will have links to those at the end of this slide deck as well.

This is the project. Of course, since 2013, David, you and I just talked about how long it has been around. We have collected, curated, and published- Yeah raw device metrics, and then transformed that into annualized failure rates and related statistics from hard disk drives and solid-state drives in our data centers.

Each quarter, we publish this open source data set, which can be downloaded on the Drive Stats page. Check that out at backblaze.com/drivestats. We do try to keep up an Iceberg version of this data set on our GitHub, so you can find that as well linked. This is a top-line snapshot. I always find interesting our drive count up and only going up, as they say. Drive days, of course, is how many days all of our drives had to date. Drive failures is just the raw count of numbers there. Drive population by manufacturer, of course, we had WDC purchase HGST.

From that perspective, we're at about a third in drive manufacturers, which is always helpful when we're looking at numbers. Certainly not ones to play favorites. There's a lot of great drives out there, and so we like to have everything nice and balanced.

Yeah, and that's the benefit of our platform. It allows us to be able to use a good amount of these drives from different manufacturers and be able to have very efficient software and infrastructure that can help run it all.

Yeah. It's that software layer on top of commodity hard drive, driving all the value there, of course. This is the Q2 2026 data. As always, incredibly difficult to read the full table. We always encourage you to dig in on the blog itself. You can see here at the bottom that our annualized failure rate for this quarter was 1.73, and this is a marked rise in quarterly AFRs. Why? First thing that I always do when I see a quarterly rise is to run an outlier test. Before we get there, let's talk about drives that had zero failures. Clean sweep from the Seagates this month. We had zero failures on the 8 terabyte, 12 terabyte, and 14 terabyte Seagates, these models identified here, and one on this 16 terabyte Seagate.

We also saw the retirement of some very old drives in our system, so that would be the 4 terabyte HGST, 8 terabyte HGST as well, clocking in at 9 and 8 years old. Then a little bit of curiosity here. No new drive models. We certainly had new deployments, but they were all of existing drive models. I flagged last quarter that it is actually quite rare for us to not have a new drive model, I think in the last 3 years or something like that. It has only happened once. So to get 2 quarters in a row feels a bit odd. But it is nothing to say there. It is really just that we are getting the same models deployed in our system at about the same volume.

Do you think that speaks to just efficiency of where we are at nowadays? It could be a number of things, but it seems like it is a good thing that we are sticking with the same models.

Yeah, I think the biggest thing is that it is not that our growth of platform has not slowed down. I think it is the biggest reason I really wanted to talk about it. It is just that when you are deploying vaults, as much as possible, you want them to be really identical. So if we can get our hands on the same models and those models are functioning well within our system, why wouldn't we continue with the drives that are already qualified? Of course, there is work on multiple ends of the spectrum there. Qualifying drives also means you have more capacity to swap in and out if you have failures, all that good stuff. So, in terms of creating a happy system, it certainly helps to have similar models, especially if they are functioning well.

I think we did not pull our review of the outliers, but you can certainly go check out the blog for that. But let us go back to the table for just a minute, and you can see here that there is a couple of models. This one up here is 7.63%, and then you have got 8.26% here. A third with, excuse me, I am losing it here, 9.33% on the Seagate 10 terabyte. So those 3 are, from a statistical standpoint, outliers. They did have very normal reasons for potentially reaching this high a failure rate. So, in one case, or actually in 2 cases, you had drives that were quite old, so more than 7 years. Then in 2 other cases, so that would include one of those quite old drives, you had drives with a very low drive count.

That means that if you get a raw number of drive failures, like looking at the Seagate 14 terabyte with the 8.26% failure rate, it's really only 25 failures. But when you only have 1,200 drives, that makes more of a difference in your AFR at the end of the day. Two of them that were quite old, two of them that had very low drive counts, and for that, those are all pretty normal indicators that they're actually already being cycled out of our system. Not a huge amount of worry even though the AFR did seem high on the surface. Let's talk lifetime data. Again, you've got the full table here. The AFR down here at 1.41%. That's very consistent actually. The AFR for the lifetime data, for understandable reasons, tends to not fluctuate as much as the quarterly.

I think it's usually 1.39, 1.40, 1.41. I feel a little bit like a broken record when we review the data on this one. Which once again speaks to the fact that we are a managed system. Somebody's trying to make sure we're within good and healthy ranges there.