A repeatable way for showing how hosting providers are making progress decarbonising the energy they use

Possibly the longest blog post title evvvveeeer!

One of the main things I do for work is manage the Green Web Dataset, a public dataset of website hosting and infrastructure providers who have shared information showing that they're using green energy to power their compute. It powers things like the Green Web Check, and the Green Hosting Directory.

Since its inception, the dataset has captured information at the company level. But that doesn't really represent how hosting companies operate in practice. Many hosting companies operate on infrastructure in multiple regions, and subsequently make it possible for their users to choose which region/s they want to use. A multi-region provider may run on hourly matched green tariffs in one place and the national electricity grid in another. Capturing disclosures at the company level doesn't allow this nuance to be reflected in a useful way.

We're currently working on technical changes in our dataset which allow this reality to be captured. But capturing the information is one thing. A question my colleague Chris and I have been thinking about recently is how can this kind of information be useful in a way that:

1. Allows the progress a hosting provider is making towards decarbonising the energy they use to be shown.
2. Allows sensible, repeatable comparisons to be made across providers, and also across regions within a single provider.
3. Makes it clear where a provider is just using their local energy grid, and where they are actively taking steps to decarbonise their energy use through additional, matched tariffs and power purchases.

In short, we want a way of talking about decarbonising the energy used by hosting infrastructure which shows progress, is repeatable, and incentivises meaningful decarbonisation.

The rest of this post will show one of the ideas we've thought about and run through an example of how it might be applied, using my recent Eurostack excursion as a testbed.

Off the bat, this idea mostly comes from Chris' wonderful mind. The first time he presented it to me, I think my brain glitched. Since then, we've bounced some refinements around to hopefully make it clearer.

A big part of me writing this blog post is so that I can make sure I'm fully understanding these ideas as well. It's also very much a way for me to try and do what I do best - take complex ideas and explain them in a way that most people can hopefully follow along with. It's handy to work in a team where folks have complementary super powers 🤗.

Okay, so here's the idea.

Hosting providers, and the individual regions they operate, can be viewed as falling into a three tier system based on the steps they are taking to decarbonise the energy they are using.

• Tier 0: A supplier is not making any disclosures about using carbon-free energy. They are not actively buying clean energy to decarbonise their operations.
• Tier 1: A supplier makes annually matched disclosures about buying carbon-free energy. They may be 100% annually matched, but generally, there are still hours in the year where the infrastructure ends up running on electricity from fossil fuels, without matching clean generation. We cannot achieve 100% carbon-free energy use in this way.
• Tier 2: A supplier makes hourly matched disclosures about buying carbon-free energy, usually using a combination of different clean energy technologies, to avoid relying on fossil generation. We can reach 100% carbon-free energy use in this way.

Using these tiers, you can compare across providers, or even across regions within a single provider, to quickly assess who is actively engaged in decarbonising their energy use and who is waiting for the grid to catch up.

Another thing that a tiered system like this can offer is the ability to generate scores for a provider, a region, or an organisation. Coupling the tiers with grid data and other open data sources (as we'll do later in this article), we can begin to develop a "scoring system" which can be used to represent a hosting provider, or can even be used to assess the tech stack used by any organisation. To do this, we will assign each tier a scoring factor:

• Tier 0: 0
• Tier 1: 1
• Tier 2: 1

I'll show how we can use these with grid data later in this post, but first ...

• Keep in mind that the aim of this tier system is to deliver a simpler, repeatable, comparable means of assessing the progress hosting providers are making to decarbonise the energy they use.
• This is not a perfect abstraction. It aims to differentiate between the three main kinds of ways hosting providers go about decarbonising their energy use.
• Finding out information about a provider's energy disclosure can take a good amount of desk research, especially when information is not shared publicly. As a default, providers/regions who do not make public disclosures about their energy procurement should fall into Tier 0 until such time as they can share information either publicly or privately.
• There will no doubt be scenarios thrown up where annually matched energy purchases (Tier 1) by a company actually do achieve 100% 24/7 carbon-free energy. These tiers introduce a general rule about how to view annually matched claims, rather than getting stuck in the weeds of edge cases.
• Finally, the current system of annually matched claims was introduced back in 2015. It was designed to encourage companies to do something about their carbon emissions without being explicitly designed to decarbonise energy grids. This suits most companies, who can do the bare minimum to tick a checkbox, rather than doing what science shows is necessary. In 2026, we know how important a decarbonised grid is, and we have the technology and mechanisms to report on and achieve true decarbonisation. The tiered approach presents a way to recognise those providers who are taking these actions.

To show these tiers in action, I want to approach this from the perspective of a company or individual that wants to assess their current digital suppliers. With this information, they can have a quick overview of how much carbon-free energy is being used in their stack and where they might start to look in order to increase that number. To do that, I will do a comparison of the two tech stacks my website runs on - the Cloudflare-based stack for fershad.com and the recent EU-based stack I spun up for fershad.me.

The stack that fershad.com runs on is almost entirely on Cloudflare. I'm starting with this one partly for that reason, there hopefully won't be a lot of data thrown around and so it can be a gentle on-ramp to the ideas presented above.

The below table shows the hosting related services that I use for fershad.com, including hosting for third-party services, and the regions that are used for each.

So in total, I'm using three providers across two regions for that site. Looking at the table above, the math might not be mathing so let me explain.

• Due to the globally distributed nature of Cloudflare's services, it's not possible for me to pin down where exactly my stuff is running. For that reason, we set the region as "Global" and will use global averages where appropriate.
• For Fathom's analytic service, I have Extreme EU Isolation turned on which means my data is processed on Fathom's EU servers in the three locations listed. Because I don't know which exact location I'm using, I'll group those together as one region and use averaged data for them where appropriate.
• Fathom also uses a distributed network to host the script file, which I load onto my site. I'm considering this to be a separate provider. For that I will use a global average where appropriate.

In their latest 2025 Impact Report (Jan 28, 2026), Cloudflare on page 50 that:

Cloudflare consumed 177.89 gigawatt hours (GWh) total energy in CY2024. All consumed energy was obtained through grid electricity. Cloudflare matched its grid consumed electricity with renewable energy purchases as part of its commitment to 100% renewable energy.

Cloudflare also publishes an Emissions Inventory annually. Unfortunately the one for this year isn't ready yet, but the last available one from 2024 points to these purchases being renewable energy attribution certificates. Aside: that statement also mentions offset purchases, but then further down states that it's based on the RE100's technical guidance which doesn't allow for offsets to be claimed. 🤔

With this information, we can list Cloudflare as a Tier 1 - Annually matched provider.

I'll be honest with you, I had no idea where Falkenstein was until researching this. Turns out that it's in Germany, and there's a Hetzner data center there. Hetzner's other EU locations happen to be Nuremberg and Helsinki. So although they don't state it explicitly on their website, I'm pretty confident that Fathom's EU Isolation is using these three Hetzner locations.

As a result, the information we would use for Fathom (as a service provider I use) would be that of Hetzner. Hetzner publishes annually matched hydropower certificates for their German and Finnish locations.

With this information, we can list Fathom's data processing services as a Tier 1 - Annually matched provider based on their use of Hetzner's locations.

Fathom uses a global CDN to serve their analytics script. I've included this as a separate category for the purposes of demonstration more than anything else.

From my testing of a few locations, it seems like Fathom's CDN provider is most likely Bunny CDN. Bunny does not publish any discoverable sustainability information online, and for the purposes of this blog post I'm not going to chase them up for private disclosures.

As a result, Fathom's script CDN services would be classed as a Tier 0 - No disclosures based on their use of Bunny CDN.

Straight away I hope you can see how this approach can present an easily scannable snapshot of a tech stack's sustainability profile. To make this even more useful, we can combine information we have about the energy grid and/or the energy being used by each provider to paint a more data-rich picture.

To start with, we can get regional grid data from open data sources like the information Electricity Maps has made available through the CO2.js library. We can find the latest available data for 2025 in the CO2.js repository on GitHub. For the global average and Hetzner EU average I wrote some crude code to calculated those in this Codepen. The results are:

Using these figures, we can give the tech stack an overall score. To do this, we'll take the carbon-free energy value for each provider-region pairing, multiply them by their respective scoring factor, and then get the average of that amount.

Doing that gives us an overall score for the fershad.com tech stack of 42% through a combination of Tier 0 and Tier 1 providers.

As a further example, say I want to compare analytics providers? I could take the information about Fathom and generate a score for it and then run the same analysis of other analytics providers I was looking at.

Now let's take a look at the Eurostack domain I recently spun up. That domain lives at fershad.me with the backstory available for your enjoyment in this blog post.

The below table shows the hosting related services that I use for fershad.me, including hosting for third-party services, and the regions that are used for each.

For this stack, I'm using four providers across four regions. Again, I'm considering Fathom's data processing and script CDN as two separate providers.

See above for the research on both Fathom services.

Based on the Static Host status page, they use Hetzner's location in Finland and Nuremberg. However, when I test the domain from multiple locations they all return the same Finland-based Hetzner IP address. For that reason, I have set Finland as the region used for this provider.

With this information, we can list Static Host's services as a Tier 1 - Annually matched provider based on their use of Hetzner's locations.

In researching my Eurostack project, I used https://european-alternatives.eu/ to sort through providers. The European Alternatives page lists OVHcloud's nameservers as being hosted in France.

OVHcloud publishes very detailed sustainability reporting. On Page 171 of their latest report, they report 305 gigawatt-hours (GWh) of energy consumption in France (304 from EACs & PPA, 1 from fossil fuel). That the 1 GWh of fossil fuel consumption appears to get rounded away when calculating the Renewable Energy Factor (REF) of 100%. On Page 75 of the same report, we get a sense of the split between PPAs and EACs used in France - 40 GWh/year of PPAs in production (about 30% of total consumption). I couldn't find any information as to whether the EACs purchased by OVHcloud were annually or hourly matched, so I will default to annual.

With this information, we can list OVHcloud's services as a Tier 1 - Annually matched provider based on their French locations.

Again, we'll use data from Electricity Maps to work out the carbon-free energy percentage for each region, and then put all that together to get an overall score.

Doing that gives us an overall score for the fershad.me tech stack of 68% through a combination of Tier 0 and Tier 1 providers.

Through looking at my tech stacks in this way, I can see the Eurostack (fershad.me) is closer to being carbon-free than the more generic stack (fershad.com). I'm also able to quickly compare across service categories and providers should I wish to.

I can use this to assess how other decisions I might make about my tech stack may impact the overall score. For example, if I was to eventually use a global CDN for the fershad.me domain, I would see my overall score fall by 14 points if I were to use a Tier 0 provider (or 5 points if I used a Tier 1 provider). In that case, I have more information to make a decision about how important the performance gains from a CDN are compared to the sustainability goals I want to achieve.

Unfortunately, none of the providers I use have hourly matched energy purchases that I could find. So let me give you an example of one who does.

34SP is the hosting provider we use over at the Green Web Foundation. They operate in the UK where the grid runs on an annual average of 70% carbon-free energy. However, through their verification in the Green Web Dataset (remember that thing I mentioned I do for work about 2,500 words ago 😅) we know their hosting infrastructure is powered by Bryte energy. We can then use Matched Energy - an open source initiative to publish hourly information about how UK energy providers are matching customer demand with clean energy - to see how Bryte energy performs.

This research gives us a figure of 78%, above the UK national grid's baseline of 70%! We'd use this figure of 78% in any scoring calculations, rather than using the nation grid average of 70%.

Let's say I wanted to compare three providers who all operate in the UK, with each conveniently being a different tier. One way I might do so is in a chart like this:

Or, let's say that I want to show the results from my analysis of the fershad.me tech stack. I could use a chart like this:

Or, what if we want to present information about the different regions a single provider offers so that leadership can make an informed choice on the location we should use:

You might use this to select Region 5 because robust, hourly matched energy claims are more important to your organisation's sustainability goals than anything else. Or you might decide that for now, annually matched claims are okay but that you'll return to this data in a year to see if there's been any change (either more hourly matched regions, or a higher percentage of carbon-free energy in Region 5).

That was a long one. We're about 3,500 words deep, so thanks for getting this far. I'm hoping that the ideas above make sense, and that they present a way in which we can frame the conversation about decarbonising the energy hosting infrastructure runs on in a way that shows progress, is repeatable, and incentivises meaningful decarbonisation.

If you have feedback, ideas, or questions about this please do drop me a line and start a conversation.

If your organisation would like this kind of discovery and analysis carried out for your own tech stack, project, or digital suppliers then get in touch via the Green Web Foundation for a consultation. We're a small, but talented team and these are the kinds of things we nerd out over!

Comments

@fershad @mrchrisadams this is quite the undertaking. Thanks!

Mike Gifford, CPWA @ UNHQ for #UNOSW - 20 August 2026