With environmental responsibility and sustainability becoming increasingly part of regulation and customer requirements, Moments Lab has embarked on a journey to understand the impact of our work and technology on the world. In this second article of a series, we describe the methodology behind our process, and why we’re not taking any tool for granted.

The digital world creates a surprising amount of impact on the environment, when you really dig into it. The digital sector represents around 2-4% of global greenhouse gas emissions, with energy consumption growing by 9% each year—that’s doubling every eight years. Perhaps more worrisome for an AI and video processing company such as Moments Lab, video represents 82% of internet traffic, and the volume of data stored in data centers is experiencing hyper-growth of 40% per year. 

So when your company is built in the digital world, and relies on data centers to host and maintain clients’ ever-growing digital databanks, can you really get a handle on the company’s environmental impact and work towards a more sustainable way of doing things?

Getting to grips with sustainability in the cloud

That’s exactly what Moments Lab has been working on for the last few months. Back in March, we introduced you to our work on getting to grips with sustainability in the cloud, and promised we’d be back with the first real data from our experiments. We believe sharing our methodology and results with the community will help us all work together towards a more eco-conscious digital world. 

We wanted to estimate greenhouse gas emissions as accurately as possible for all of our activity and its value chain. That means going beyond scope 1 and 2 emissions to get a real sense of our scope 3 emissions and our total carbon footprint. To recap:

• Scope 1 emissions are fairly straightforward; these are the direct emissions from owned or controlled sources.
• Scope 2 emissions are indirect emissions from the generation of purchased energy—nearly 40% of global greenhouse gas emissions can be traced to energy generation, and half of that energy is used by industrial or commercial entities. These are the emissions that can be reduced by switching to low-carbon electricity. Both Scope 1 and Scope 2 emissions are mandatory to report on.
• Scope 3 emissions are the trickiest to track, and so you’ll find a lot of hidden values and pure greenwashing. These emissions are the indirect ones that occur in the value chain of the reporting company, including both upstream and downstream emissions—and they are voluntarily reported. 

And, our first tangible results confirm our suspicion that our emissions under scopes 1 and 2 are negligible. We’ll detail our methodology in the next section, but first let’s show our results. The main sources of carbon emissions in our work are:

47% digital, of which: 

• 18% are from data centers
• 10% are from network traffic
• 1% from Moments Lab's own IT equipment
• 1% from customer user devices, and,
• 17% are from outsourced digital services.

22% are from the purchase of services, and 17% are from business trips. The impact of activities such as remote working, premises, and equipment, each make up less than 5% of the total.

So with our scope 3 emissions comprising 99% of our total footprint, but being the trickiest to track, our subject matter expert and external consultant Eric Ferrachat clearly had his work cut out for him. We wanted to assess our scope 3 emissions as exhaustively as possible, paying particular attention to the emissions related to our cloud infrastructure.

What needs to be considered when assessing digital sustainability?

Any carbon footprint requires listing all of the company’s emission sources. For this project, we looked at categories such as:

• Premises
• Telework
• Trips
• Accommodation and food
• Equipment
• Service purchases
• Purchases of supplies
• Digital.

All non-digital categories have been assessed in accordance with the usual carbon accounting methodologies—such as Bilan Carbone, the GHG Protocol, and ISO 14067— favoring the use of physical rather than monetary data as much as possible for greater accuracy.

When it came to digital—home to most of those trickier scope 3 emissions—we looked at subcategories including:

• Company computer equipment, such as laptops and monitors.
• Outsourced digital services, such as the purchase of other SaaS services (chat and marketing systems etc).
• Internal web use, such as the use of data centers, networks and terminals related to team work.
• Social networks, including the use of data centers, networks and terminals linked to them.
• And, importantly, the use of Moments Lab’s own services by customers—that is, the use of data centers, networks and terminals linked to the use of our platform by customers.

Given the work we do and the service we provide is generative AI-powered indexing and storage of media in the cloud, this final point was the big one. And it was also the assessment that required fresh thinking.

Digging under the greenwashing for reliable measurement

Estimating the use of Moments Lab services by customers amounted to estimating three separate emission sources:

• Emissions linked to the data center (computing, storage, other services).
• Emissions linked to the network (data transfers on the web), and
• Emissions related to user terminals.

We started with our own cloud provider, AWS. Although AWS provides a Customer Carbon Footprint Tool, we didn’t find its results satisfactory for various reasons:

• It doesn’t include scope 3 emissions, and these are the main contributor to our footprint.
• Its scope 2 emissions reported is rooted in a market-based method rather than a location-based one, which results in hidden emissions. For more information on these methods, this guide is helpful.
• It allows the use of carbon offsetting to virtually eliminate emissions.

Given location-based methods are mandatory under the Bilan Carbone methodology, and the GHG protocol states that “best practice is to report using both location-based and market-based methods”, we wanted to go beyond their in-house carbon footprint tool to see things more clearly. Unfortunately AWS doesn’t provide any physical data—such as the energy consumption of various services—that would allow us to assess our emissions with precision. We needed to use other means.

Instead of the AWS tool, we used several complementary methodologies. 

Data center emissions 

We considered the computing, storage, and other services of the data center emissions block as separate. This helped us to benefit from the best available current accounting methodologies for each block.

• For the computing:
• We evaluated the energy consumption (kWh) of all the virtual machines instances from detailed billing data (physical data)—that is, we estimated Moments Lab’s share of AWS’s scope 2 emissions based upon our usage of their infrastructure—using the methodology and dataset of Teads Engineering, and taking into account PUE (power usage effectiveness) and local energy mix. We consider the Teads Engineering method to be the best available to date when evaluating emissions from elastic cloud computing (EC2) instances; they did great work in assessing the energy consumption of EC2 virtual machines.

• We assessed the impact of AWS scope 3—an important part of the project, considering the fact that these indirect emissions in the AWS value chain represent the biggest share of our data center-linked emissions—using the monetary ratio of 130 gCO2e/€ from the General Carbon Plan. This is not a precise method, but it does give an order of magnitude sourced from other comparable hyperscale cloud providers—it’s a worthy option considering AWS doesn’t share the data. It was important for us to include this assessment, as most companies choose to exclude these emissions from their reports due to the lack of data, which results in a big blind spot.

• For storage:
• We evaluated the energy consumption (kWh) of the servers from detailed billing data, using the Cloud Carbon Footprint methodology (emission factor of 0.65 kWh/TBh). We recognise this is a lack of granularity, but we don’t (yet) have specific numbers to differentiate the different types of storage (S3 - Simple Storage Service) on AWS. These are S3 Standard, S3 Infrequent Access, and S3 Glacier. We’d love some help in this task - feel free to contact us if you have anything to share!

• Here again, we assessed the impact of AWS scope 3 using the monetary ratio of 130 gCO2e/kWh from the General Carbon Plan.

• For all other services (lambda, rekognition) provided by AWS, we have used as-is, the value calculated by our Cloud Carbon Footprint instance directly connected to our AWS accounts.

Network

For the network between the data centers and the users, we evaluated the energy consumption (kWh) of the web infrastructure from detailed billing data, with the emission factor 1.91*10-4 kWh/MB (source).

Devices

For the terminals, we assessed the energy consumption (kWh) of the terminals and the impact of their manufacture (scope 2-3) based on the usage statistics of our platform (Google Analytics) and the methodology of the General Carbon Plan.

Why use so many different methodologies?

This work is unfortunately far from being perfect. We hope data will be made more accessible in future to allow for better, more precise, and more granular evaluations—and the granularity is especially important to help us take appropriate action to address our footprint. We welcome comments and suggestions on how we can improve our methodologies!

We wanted to be more granular than the general practices found in other carbon accounting—especially as we wanted to ensure we included the scope 3 upstream emissions of our cloud service provider. This is too often forgotten, but it represents a significant part in emissions related to data centers.

For example, working in accordance with the rules of the Bilan Carbone® (the French carbon accounting standards), emissions related to electricity are counted according to the physical location-based and not market-based approach in order to avoid biases and better reflect the physical reality (which amounts to counting more emissions than with a market-based approach allowing the purchase of electricity considered “neutral” in carbon”).

Sure, we could’ve easily used the AWS customer carbon footprint tool. There are some very big companies relying on this single tool, and we are competing against many of them when it comes to pitching our services to potential clients who want to know about our carbon footprint as part of assessment. But we were not satisfied with rough estimates that only focus on a too limited perimeter and leave big blind spots behind. 

This diversion in methodology was quite an emotional decision. AWS’s scope 3 emissions account for a huge chunk of Moments Lab’s own emissions, yet they don’t share that data. Take, for example, AWS’s employees. 

The percentage of vegan vs meat meals served in AWS cafeterias would have an impact on our carbon footprint. If they all had to commute into an office, that would have an impact on our carbon footprint. How much CO2 does their data center warehouse emit? How much concrete did they use in its building? All of these contribute to scope 3 emissions, and yet aren’t shared.

I was shocked to discover that even when complying with official methodologies, a GHG emissions assessment can lead to different results depending on certain choices we make: 

“The application of this method may lead to different methodological choices on the part of its users. As a result, the GHG Reports of different organizations resulting from the use of this method cannot be used for comparison purposes without first verifying that any methodological differences do not cause significant bias in the comparisons. The GHG assessment remains above all an internal management tool for the development and implementation of relevant climate strategies.” (ref “Method for the realization of GHG reports”, p7 (FR)). 

This is a complex topic, and it’s a matter of judging the benefit vs. risk of being fully transparent. For this reason, we have chosen to share our methodology, but not our specific carbon footprint figure.

We’ve found there are two values, essentially: one that includes scope 3, which we’ve estimated for our suppliers, and one without taking into account those supplier scope 3 emissions. The impact and difference between the two is huge. 

It all depends on the perimeter and methodology you use, and so I wanted to make sure that Moments Lab’s carbon accounting was as robust and accurate as possible, even if that meant looking worse than competitors.

Accuracy is more important to us as a company when it comes to sustainability, and we see part of our responsibility in this journey is to explain the impact your supplier’s scope 3 emissions can have on your own carbon footprint.

Even the choice of where you store data can impact things. While AWS data centers have similar hardware and power usage effectiveness (PUE) across the globe, different locations will use different electricity mixes to power the data centers. So, you could choose a data center in Paris (~60gCO2/kWh) rather than Dublin (~350gCO2/kWh) or Cape Town (~700gCO2/kWh) (data from this source). The way a country generates its energy can impact your carbon footprint, and Paris has a more low-carbon energy source. South Africa is still coal-powered!

The way forward for Moments Lab’s sustainability pact

These results show that the emissions related to the digital sector are significant when certain methodological biases are avoided. It’s super important to take into account, in an exhaustive as possible way, the scope 3 emissions both upstream and downstream. Also ensure any scope 2 emissions are calculated with a physical approach and not only with a location-based approach.

But where do we go from here? Well, we’re building out a sustainability pact to drive future business decisions. We’ll detail these in the next article in this series, but here’s a sneak peek.

From now on, Moments Lab will:

• Conform to Net-Zero initiative recommendations to contribute to global carbon neutrality.
• Set ambitious and SBTi (Science Based Targets initiative) compatible objectives to reduce value chain emissions (scope 3).
• Contribute to raising awareness among our value chain and encourage its decarbonation.

In the next article of this series, we’ll set out the objectives we have set ourselves—as well as the detailed action plan written to help us achieve them. 

Have questions or need more information about our generative and multimodal AI-powered cloud solutions? Get in touch with us.

🚨 Our revolutionary AI indexing technology MXT-1 is out of beta and fully released at IBC2023,  September 15-18 at the RAI Amsterdam. Book a demo meeting here or drop by our stand, 7.C30.

Need a free visitor pass code? Be our guest: IBC7813