Cloudflare TV

🌐 Cloudflare Emissions Update

Presented by Michael Aylward, Patrick Day
Originally aired on 

Welcome to Cloudflare Impact Week 2022!

Cloudflare's mission is to help build a better Internet. We believe a better Internet can be not only a force for good, but an engine of global sustainability. This week we'll be highlighting an array of initiatives inspired by these optimistic ideals, as well as stories from partners who share them.

In this episode, tune in for a conversation with Cloudflare's Patrick Day and Michael Aylward.

Tune in all week for more news, announcements, and thought-provoking discussions!

Read the blog posts:

For more, don't miss the Cloudflare Impact Week Hub

Impact Week

Transcript (Beta)

Welcome back to Cloudflare TV. As you can tell, we are already in progress.

My name is Patrick Day.

I'm here.

I'm from the public policy team here at Cloudflare. I'm here with Michael Aylward from our infrastructure team.

We are in the middle of this is day three of impact week.

So all things Cloudflare impact if you're not familiar.

Cloudflare impact is our company wide Sustainability and Human Rights and Corporate Responsibility initiative.

So Impact Week is sort of a week that we set aside every year where the whole company is focused on all kinds of impact issues.

We launch a ton of new announcements, some new products, some contributions to our corporate responsibility programs like Project Galileo and the Athenian project.

This is day three. So if you haven't seen all the content from Impact Week so far is on the Impact Week Hub page, so I encourage you to check that out.

There's some really cool announcements on Monday about making our Zero Trust available for Project Galileo, which is really exciting.

Tuesday was all about protecting critical infrastructure, which is a really neat new program, and we're expanding that to a couple of new markets outside of the United States.

So really cool stuff. I encourage you to check it out if you haven't seen it so far.

But today, Wednesday, day three is all about sustainability, which is my favorite day.

I know it's one of Michael's favorite days, given how long he's worked on sustainability issues here in Cloudflare.

So really excited to have him here with you.

We'll get into sort of the history of emissions and how we've set up our sustainability program at Cloudflare in just a second.

But just as a preview of the types of things we have coming out today.

It's a pretty long and extensive lineup that I'm pretty excited about.

We have we have the first findings from an independent study we commissioned that is looking at the carbon efficiency of Cloudflare cloud based services versus sort of equivalent on prem providers, which will give us a better understanding of sort of the sustainability impact of moving things into the cloud, particularly sort of down the stack where Cloudflare has all of its network solutions.

We're talking about sort of our next generation of hardware, which is near and dear to Michael's heart.

Rebecca Weekley and her team and how they're sort of incorporating modular design and open firmware standards.

All has the effect of driving down embodied carbon within our hardware.

That is really exciting.

I actually learned a lot reading that blog. If you haven't seen it on, encourage you to check it out.

We have a new partnership with Iron Mountain to help our customers recycle their hardware boxes.

We have an update on our bots and trees commitment, if you're not familiar with that.

Pretty cool stuff. And then I think actually my favorite blog of the week, which is our facilities team or Places team, has updated our offices in San Francisco after the pandemic and some of the sustainable materials that they've incorporated into the design.

And we have a hive of entropy that they're planning for next year. It's very cool.

I won't give it away. You should definitely check out that blog.

That's all the things we have sort of going out the door today.

I think the point of this segment, the reason I wanted to have Michael here is to talk all things carbon emissions.


I think starting at the beginning is probably the easiest way to do it. And Michael's been working on these issues long before we had an impact report, before there was any Cloudflare impact collection of materials sort of at the beginning stages of Cloudflare thinking about renewable energy and hardware, etc..

So, Michael, my first question is sort of take us back to the beginning of these 28 team, probably when Jeff Bailey sort of started the conversation about renewable energy and the infrastructure team and sort of how all that conversation.

Yeah, yeah.

Before I go there, it's just really cool to hear you sort of tie up all of these themes together.

I've been sort of neck deep in just a couple of blogs this week, but it's really cool to zoom out.

I know that, you know, your job is impact and this is your week, so it's cool to hear the sort of 10,000 foot perspective on all these things.

The funny thing is.

There are so many teams involved and yes, we have sort of a central organizing place, how we do these innovation weeks where all the announcements are piled up.

But until I read the list out a few minutes ago, I was like, Wow, this is a lot of stuff.

It took me a while to go through the list. anyway.

It's funny, it's very Cloudflare.

Yeah, it's I'm continually blown away about when we sort of announce that the Impact Innovation Week is coming, how many teams and people across the company sort of raise their hand and say, This is what I've been working on?

We think it's a cool story.

So yeah, it's it's turned into quite, quite a bit.

I actually it reminds me that I'd like there's a like a bunch of the things that you mentioned.

I realized that I want to just kind of extracurricular dig into. So I'm going to have to follow up with each of the teams that worked on them and and just learn about them and.

Tomorrow's human rights.

And we have folks from the UN who are going to come and do a fireside chat, and it's pretty cool.

So pretty.


Yeah. Yeah. There's, there's the, the volume and depth is yeah, it's, it's striking, but okay, so let me bring up.

In the beginning, take us back to the early days when it was like what the conversation was like and sort of how challenging it was to get started.

It's really interesting.

So I had just joined the infrastructure team like only I think a couple of months after Jeff Bailey had completed the first exercise that we had done.

That was that was actually prompted by Matthew, our CEO, having sort of vaguely had that in the back of his mind, I think for maybe even a couple of years.

I remember at the time I didn't know and.

He was he was sort of it doesn't surprise me, but I just didn't do it.


I think like among he just had so many things that he wanted to prioritize and I think he got nudged a couple of times and he was like, you know what?

Maybe this is a good opportunity for us to dig in and actually just see what it would take for us to do this since our network was growing.

And I think his intuition, although his intuition was that sort of our footprint was going and one of the things we needed to do is was actually put some flesh on the bones there and just see what it was.

We hadn't done that exercise before.

It turned out in, in that when I so our first our first effort actually that I wasn't involved in just did right before I joined the team as I was sort of talking with her about joining the team was to do basically just an accounting of our carbon footprint, our energy use and the embedded carbon within that energy just for North America alone.

And we started there because it was tractable and easy.

I think although we always had the ambition to sort of to do the whole network, we just wanted to start with the first problem first and not let sort of accounting issues that we hadn't solved yet.

Stop us from doing what we could do at the time.

So when I came on board, we, we sort of started thinking about how to expand that effort to the whole world.

And I sort of quickly realized that one of the sort of first basic foundational parts of a company's sustainability efforts is really just a carbon accounting exercise.

And I hadn't really appreciated that before because I'd worked on, you know, sort of climate policy and energy policy for states and country level.

And so I'd always thought about just sort of finding the sources and but I hadn't thought about just the actual exercise of doing the basic accounting.

And it turned out that was kind of non-trivial.

I mean, you've worked with and some other folks about like how we do the details.

And I think we've actually gotten a lot better at how we do it. We've gotten more streamlined.

And yeah, I think the well, I should say when we had our first conversation about what Cloudflare had done in the past and what we needed to do to sort of make sure we were aligned with some of the accounting standards moving forward.

You told me that you thought someone on the team had sort of built a tool to measure the energy of all of our metals all over the world and sort of have those tallies.

So we could actually do sort of real time calculations. And by the time we reached out to Liam Reese on the infrastructure team, he had in fact had that tool ready.

So really my job was made much easier because instead of sort of doing rougher calculations that had gone on previously, we now sort of had this tool that was giving us real time information about how energy was being used across our network.

And I think if you talk to other folks on the infrastructure team, this is one of the things I wanted to ask you later, but sort of relevant now is we pay for power, right?

So it's not just that we were going through this carbon accounting exercise.

It was also that one of our cost drivers for the infrastructure team is how much energy we're using.

So to the extent we drive that down, it's in our business interest and it sort of benefits the environment totally.

It was also I mean, I think as we collected this information initially for the purpose of carbon accounting for this the sustainability exercise, a whole bunch of other teams were really interested in the results, right.

They were curious what the overall level was, what the what the total energy footprint of our of our global network was, and then where it was distributed and also maybe how it was changing over time in each region.

So those things were really neat lessons and insights that came out of this.

I mean, I think we would have wanted to do that anyways, but this is just a great accelerator.

Yeah, I think you're right.

It's still one of my favorite things. Like what we do.

We have the end of year calculations and then you actually see in real time, sort of not only how energy is used across our network, but you translate that by jurisdiction using emissions factors, which we can talk about into the rate of carbon emissions all around the world.

It actually tells a really interesting story about our network and how electrical grids work and how different countries where it's just a really interesting way to look at sort of the business.

But I don't think I certainly would have.

I mean, I know it's sort of part and parcel of what the infrastructure team does, but I wouldn't have been exposed to it otherwise on the policy team.

It was just really cool to see and I feel like I have a much better understanding of sort of the hardware side of things that I would have otherwise for sure.

Yeah, yeah.


I also, yeah, I think this also accelerated my appreciation and insight into those things.

I sometimes work a little bit with our our what we call hyperscale data center strategy team.

Yeah, got it. Michael King runs it.

Who I know that you probably also work with a little too.

And Michael's done this for a little while and it was really interesting when he came on board and talked exclusively about co-location and physical footprints in terms of megawatts.

He didn't ever talk about anything else.

He didn't talk about any of the other sort of metrics that you might use to measure what you would want as a as a consumer, as a as a user of that space.


And I realized later that's just industry standard. And then and sort of over time learned why.

But it was really interesting that that's kind of just like you say, it sort of gives you an interesting insight into it's a suggestion, the fact that you that you actually measure and plan and implement data center sort of functionality almost exclusively by its power footprint.

And what that suggests is exactly what you said it when you dig into it, it gives you the lesson that you mentioned about how power and sort of the distributed city, the sort of distribution of these things sort of works across the system.


Super interesting.

So back to the sort of our story narrative. I think we're going to get off track several times.


No, it's fine.

You're just bringing me back to this.

I'm just realizing we never sort of fully answer the question for the viewer.

Anyway, so 2019, we sort of expanded it to our global footprint.

Is that right?

And then yeah, yeah, we basically did this same exercise and we actually deepened.

It was really fun.

That was the first one that I had been involved with, and I sort of was given the responsibility to do it.

Volun told basically in the true fashion.

And it was, it was neat to start with where we begun the previous year with the American accounting exercise and the choice of what to buy.

And that was the first time I also was faced with the set of options that you have as a company, whether it's if you're a giant energy user like, say, Google or Microsoft or Facebook or an Apple and you use enough power to actually be able to justify buying and building your own projects that are large enough that that they would actually go through the engineering, procurement and construction of, say, a solar farm or a wind farm that might be dedicated to one large facility, for instance, like Google's Council Bluffs data center in Iowa, one of the largest data centers in the world.

And it's supported by, I think, one or two large utility scale wind farms that they're either part or co owner of in the Midwest.

So whether you're at that scale or which we weren't at the time or we're sort of learning about other more sort of sliced trenched options, it was just really interesting to learn about what was available to us and what other companies were doing at our scale and sort of growing.

Yeah, sure.

I think one of the really interesting and also candidly challenging things about Cloudflare in terms of renewable energy.

So, as you know, one of the principles for if you look at our 100 or some of the other standards about how companies or organizations are supposed to purchase renewable energy, you generally have to purchase in the same countries in which you're using the energy.

So that's easy.

If you have one data center in one location and you can sort of sort of structure your asset to make sure that you have it available.

But for Cloudflare, we're so geographically distributed, we're everywhere.

We're all over the world. Where in those 270 locations, right in over 107 countries, is that right?

Yeah, actually, yeah, exactly.

More than 275 cities.

Many of those cities have multiple data centers, so.

More than 450 locations.

It's amazing.

So part of the challenge for the sustainability team is not only do you have to sort of when you do your carbon calculations, you have to pay for the energy you're using in a particular location with the relevant emissions factor, and that's how much carbon is produced by that electrical grid in whatever location you're talking about.

So if you live in California, there's a California state average emissions factor.

If you live in Texas, if you live in Japan, if you live in South Korea, all of those there are companies that sort of go around to all the energy sources that populate an electrical grid, like the one that connects to your house.

And then to calculate your carbon emissions, you take the amount of energy you use.

So whatever's on your utility bill megawatts per month, and then you multiply it by the emissions factor for your local electrical grid.

So the challenge for Cloudflare is because we're in so many locations, in so many countries, we have to pair sort of the actual energy use in each of those locations.

We talked about how many countries that is with the rate of emissions for that local electric grid, and then that's how we get our carbon emissions totals, which we add up and then publish.

That's our carbon emissions inventory that you can see on our website.

But then it also comes back to the renewable energy question.

You have to find renewable energy in all these locations and some countries don't have renewable energy available.

So you find the next best available product.

But I think that's a very unique Cloudflare story, right?

We're not sort of the same as other organizations like Google that have sort of although Google has lots of locations, sort of they can maximize the utility of one location and sort of drive efficiency and computing power and sort of one location.

We are much more spread out. We're much closer to the user by intention, but that makes it a little bit more challenging of the sustainability and renewable energy side.

Yeah, Yeah.

It's it is really interesting. I don't think there's another company that I'm aware of that sort of purchases in the same way that we do.

There may be, but I just sort of I feel like in every conversation I have about sort of renewable energy discussion group, that we have a unique story that we have so many needs and so many locations.

Yeah, you're right.

It is it is a it is a really unusual combination of being so widely distributed.

Like you said, one of the 275 cities distributed across the globe.

And the interesting thing is the volumes in each place aren't enormous, which sometimes makes it hard for our for our partners, our environmental product partners to find sort of that piece.

The interesting thing about that is that there's a certain efficiency that goes along with that.

We're able to basically meet lots of customer needs in, say, you know, Central America or Brazil or South South Africa or Saudi Arabia or Hong Kong or Singapore or Taiwan without enormous footprints.

We're able to basically deliver a lot of the products that our customers need, including our 27 million plus free customers.


Basically with really small energy footprints. So that part is actually a really interesting part about our network.

I'm curious to see how that grows as our Matthew talks about the acts of our products.

And I'm curious to see if as we go through sort of the second act of our products, which is sort of the more sort of traditional enterprise security distributed security suite, and then move into our third act, which to be honest, I'm still like getting my head around.

I think we're sort of still getting into.

I'm curious if that changes our energy footprint, but I guess you and I will have to figure that out over the next few years.

Thankfully, there are other people who can help with that particular question.

Yeah, we'll react by.

Signing up for.


Actually, no.

What you said raises something which I think is really interesting, which I hope that our team, that our teams are able to study more in the future.

And this is I'm going to transition to something that's happening this week, which I think is worth mentioning.

But we are, I think, globally understanding sort of the carbon footprint of the Internet as a whole.

There's been sort of a lot of research and advancement made on generating efficiency at the individual hardware level.

So every generation of hardware becomes more power, more or more performance per unit of power.

That's obviously one of the major drivers. There's also been sort of a lot of study and important work done and sort of you mentioned hyperscale data centers and how to run them absolutely as efficiently as possible.

So for folks who follow these conversations, power usage, efficiency for a particular data center, so driving that as low as possible in terms of performance.

I think one thing that's relevant to Cloudflare that you sort of touched on that has not been explored to the same extent is the architecture of the Internet itself.

So we you mentioned we are geographically distributed so that we can make content available on behalf of our customers to users much more quickly, reducing latency by being close to the customer, being able to distribute content as quickly and securely as possible.

The flip side of that is there's a qualitative gain in that our customers don't have to have the same sort of internal capacity to serve content all over the world because they're using Cloudflare's architecture.

And we've talked about the way that our this is kind of in the weeds, but the way that our infrastructure is built.

You mentioned sort of how lots of locations, we're only drawing small amounts of power, but we're still able to distribute that content.

To cut to the end, to the point that I'm trying to make is I think there are really interesting efficiency gains to be made in the architecture of the Internet and sort of leveraging these content delivery models like the services offered by Cloudflare so that we have sort of less demand on undersea cables that we have.

We're moving sort of.

More content delivery type services out on prem into more efficient cloud delivery.

I think those are sort of we'll talk about the study that we're announcing this week in just a second.

But I actually think the architecture question is one that would take a lot of actors to get to the bottom of.

But I think that's a conversation we'd really like to be a part of, to really understand that efficiency gain and how we can sort of push it forward.


It's funny. I didn't even I was, I was guilty of, of talking to, of not explaining to our viewers what I was talking about with the acts of our products.

Talking about Act two and Act three probably begs the question, what's Act one?

And act one is what you just said. Act one is the traditional CDN and that Cloudflare began with as its first sort of core customer facing product suite.

And obviously, we didn't invent that stuff.

We sort of followed in Akamai footsteps and there were other pretty good CDNs out there.

And the neat thing is that architecture has been sort of not established fully.

Obviously, it's still it's still evolving and it's not fixed.

It's not really static because people are still finding improvements.

But the sort of general architecture that you described is has been sort of in place and in active evolution.

In real, real.

It's it's it's not just experiments, it's not in beta.

It's in like real big scale production for more than a decade.

So that part has actually it's really interesting it that probably has caused maybe a slight uptick in, say, the footprint of undersea cables, because undersea cables are really useful to distribute those things.

On the margin, I actually don't know if I'll stop my head if you actually do need more undersea connectivity or just general global connectivity between data centers in order to have content delivery closer to end users.

It strikes me that you might, but definitely what you what you suggested before is true that if you don't have to go back and every user in every country, South Africa, Malaysia, Hong Kong, Taiwan, Brazil, Texas, California needs to go back to wherever that content sort of came from.

Obviously, there's going to be savings in the energy footprint of the actual servers, the computers that actually process that request from the user.

So for sure, there's a saving there.


The question is and you also mentioned like the, or at least suggested, the sort of big time savings and the savings are amazing.

Actually, I got to show you this, this blog post and maybe we should put it out there at some point.

It's it's not Cloudflare specific, so it hasn't sort of made it yet.

But I did some analysis, I think last year or the year before, you and I were working on a different blog and sort of message that we wanted to convey.

And one of the conclusions there was that even as there are some really interesting research, that even as the overall computing like functionality and hardware and the actual work, I forget what they call it, I think it's like the work product or the amount of work that's done by global computing has expanded by, I think ten X over between 2010 and 2018, which is the latest data that I had at the time.

The increase in energy was 6%, so the increase in use was 1,000% and the increase in energy was 6%.

So the reason that essentially didn't increase at all is exactly what you said.

Every single increase in sort of a functionality or a process that had previously used more energy was offset by by some efficiencies in that process itself.

So that's why people have been able to basically build crazy, enormous data centers, finding incredible economies of scale and sort of processes of scale that allow us to it's basically like the industrial equivalent of Moore's Law.

The bad news is that Moore's Law is going to run out at some point.

And it already is for processing and it almost certainly will it's obviously related to this energy use question.

But at some point, we're we're going to run out of doublings.

We're going to be able to squeeze maybe one more order of magnitude of increase in productivity without an increase in energy use.

But at after one or two more doublings, we won't have them available anymore.

And so then after, if we want to use more energy than that, after that next one or two doublings, we're going to have to figure out like it's inevitably going to use more energy.

And so the question is, will that energy have a carbon footprint?

And my hope and I think a lot of optimists hope, is that by that time it might take five years or whatever.

Maybe we'll even get lucky and it'll only take ten years before we do that.

Next, doubling. At that time, energy use will inevitably increase, but will we be able to power it with carbon free sources?

And I think the good news is we're seeing we think that we will.

But it's basically a race.

Can we get only carbon free sources adding by the time we start adding more energy use and sort of that's the sort of order of operations question.

Very cool.

Okay. Well, so we have just 2 minutes left, so.

Oh, man, I'm sorry.

I went off saying no. I was just sitting along listening.

So it's the report that we are announcing today, or I should say the initial findings of a report we've hired analysis Macy's, a UK based consulting firm.

They are doing an independent analysis of our network as compared to a sort of equivalent on premises providers.

So trying to understand the environmental consequence, the carbon emissions consequence of shifting sort of traditional on premise services like web application firewall, any type of firewall SD, wind up load balancing optimizer, sort of what's the carbon ramifications of moving that to the cloud So through Cloudflare services and analysis of our network, So we don't have the full report yet.

It's coming out first quarter of next year.

I'm anxiously looking forward to it.

But the really cool thing and I was we've worked on Sustainability Cloudflare for some time, but I was sort of surprised by the initial number they gave us that Cloudflare is WAAF alone is about 90% more carbon efficient than equivalent on prem services.


So that means if you if you're a customer using on prem services, you can cut your carbon footprint from those literally in 1/10.

It's all the above.

It's on the blog.

So if you're interested, please check that out.

Thumbnail image for video "Impact Week"

Impact Week
Tune in for all of Cloudflare's Impact Week programming, featuring an array of CFTV episodes spanning environmental, social, and governance issues.
Watch more episodes