Making the Internet Sustainable— Starting from Its Infrastructure

Simon Wistow

VP Strategic Initiatives, Fastly

An illustration of a keyboard with a lever in the middle and a hand pushing the lever forward

For those who are not network engineers, the internet might feel like a weightless, invisible force—something we interact with every day but never physically see. But for us, network engineers, the internet is inherently physical. It consists of ports and cables and is made up of data centres filled with machines that sit in racks and cages.

And here’s the kicker: all that physicality comes at a cost. The internet is one of the world’s largest consumers of energy. But because its physical nature is largely hidden, it’s managed to dodge the same sustainability scrutiny that other industries have had to face. That’s finally starting to change.

The Shift Towards More Sustainable Internet Infrastructure

More and more organisations are waking up to the fact that digital services have a hefty energy footprint. Standards like SustainableIT, backed by Perforce, MasterCard, Harvard, Siemens AG, and Morgan Stanley, aim to "provide a cross-industry, consistent set of topics and metrics for measuring, tracking and reporting technology’s ESG impact". The IEEE Standards Association has also released a report, Emerging Technology, Standards, and Sustainability, highlighting best practices for sustainable digital infrastructure.

Additionally, The Interactive Advertising Bureau (IAB) has been developing frameworks to optimise digital media delivery. One of their key areas of focus is reducing unnecessary computational work, like balancing compression against CPU use to deliver an optimal experience while cutting down on wasted energy. In plain terms, the idea is to do as little unnecessary work as possible.

How Smarter Infrastructure Saves Energy

One obvious way to reduce waste is by optimising how we deliver content. Instead of handling every request from scratch, we can:

  • Pre-create mezzanine files of different sizes rather than converting them on the fly

  • Aggressively cache frequently used data to prevent redundant computations

  • Cache common fragments in personalised content instead of generating everything dynamically

This is where edge computing and serverless architectures come into play. By processing data closer to the user, edge computing reduces the distance data must travel, minimising energy consumption. We’ve also moved past outdated cache-busting techniques. Now, efficiency is key.

And let’s be honest: not every web experience needs to be a Single Page Application (SPA) loaded with megabytes of JavaScript. Not everything needs to be a React app.

The Decline of Crypto and the Rise of AI

The cryptocurrency hype has died down, but not before making a significant environmental impact. As of February 3, 2025, Bitcoin’s energy consumption was estimated to be at least 141.2 terawatt-hours annually—more than the entire country of Sweden consumes in a year. To lower the environmental impact of blockchain, Ethereum transitioned from a power-hungry Proof of Work system to Proof of Stake in 2022, drastically cutting its energy consumption.

But even as crypto's dominance fades, AI is stepping in as the next energy-intensive industry. Training large AI models like GPT-4o, LLaMA 3, and Google’s Gemini requires enormous amounts of power. The energy costs don’t stop at training. Researchers estimate that each AI query uses 5 times more energy than a standard web search. The Washington Post reports that a single 100-word email generated by GPT-4 can use up to a bottle of water for cooling. If just one in 10 working Americans used AI to generate an email once a week, it would consume 435 million litres of water annually.

Data centre water consumption:

Rethinking Internet Infrastructure: A Smarter Path to Sustainability

The key lies in doing less work whenever possible and ensuring that the work that must be done, is done in the right place and powered by renewable energy:

  1. Optimize AI Model Training – Instead of relying on massive models for everything, use smaller, more targeted models that consume less power.

  2. Cache AI Responses – Even with natural language queries, caching is possible. Since Large Language Models (LLMs) convert words into vectors, we can cache and reuse outputs for semantically identical queries, significantly cutting computational costs.

  3. Rethink Data Centre Locations – Moving computing workloads to locations with abundant natural cooling (like Iceland) or renewable energy sources (like solar-powered underground facilities in deserts) can drastically reduce environmental impact. Google and Microsoft have even experimented with underwater data centres to leverage ocean cooling, while start-ups such as Deep Green are disrupting the traditional colocated data centre model by building data centres where the heat that compute generates is contributed to public projects, for free.

The Role of Tech Companies

Tech companies have a responsibility to be part of the solution. At Fastly, we recognize that we are part of the problem. That’s why:

  • We prioritise green colocation providers: about 70% of the data centres in which we operate our infrastructure incorporate either 100% or partial renewable electricity 

  • We work with partners to ensure 100% of end-of-life equipment in PoPs is repurposed or recycled

  • We’ve invested in high-integrity carbon credit projects to compensate for the estimated emissions associated with non-renewable energy consumption in our Points of Presence, while not a perfect solution, they contribute to broader sustainability efforts.

We continue to prioritise sustainability with launches like:

  • AI Accelerator: a semantic caching solution for large language model APIs which can reduce the volume of origin queries (and therefore energy consumption) associated with generative artificial intelligence applications.

  • Sustainability Dashboard: which allows our customers to know how much energy we are using for their workloads - allowing them to fine-tune their delivery chain to reduce energy consumption (currently in beta - and we’re working to get this to GA ASAP, with valuable feedback from the community getting us there even faster!). 

Moving ahead, we’re investing in a sustainability program that will help us to more accurately measure our environmental impact and find ways to raise our climate ambition.   

Building a Truly Sustainable Digital Future

We all need to start thinking about the internet as something physical — because it is. It uses real resources, generates emissions, and has real-world consequences.

It’s time for companies, developers and policymakers to take responsibility. That means optimising infrastructure, pushing for stronger regulation, and holding ourselves accountable.

Because let’s face it: keeping this planet healthy is a lot easier than trying to terraform a new one.

If you want to learn more, watch this video on how individuals and organizations can reduce their carbon impact and contribute to a more sustainable digital future:

Speed and sustainability can go hand in hand. Discover 5 simple ways to make your website faster and more eco-friendly with Fastly