Rising energy costs and AI workloads are forcing the internet’s biggest players to rebuild their data centers around efficiency-first chips.
Every day, billions of requests race through data centers to make every technology experience better, faster and more intuitive. Yet most people never think about the machines at work behind the scenes. With rising energy demands and costs, thanks in a large part to AI’s deep appetite for compute, the pressure is now on cloud and data centers providers to rethink their computing foundations.
A report by the International Energy Agency (IEA) warns that global electricity use for data centers could more than double by 2030, rising from around 415 terawatt-hours in 2024 to nearly 945 terawatt-hours. That’s the equivalent to powering every home in the United Kingdom for nearly a decade. Therefore, scaling computing infrastructure on the old model is becoming challenging without blowing cost and energy budgets.
Why Energy Is Forcing a Reset in The Cloud
For decades, data centers around the world were predominately based on the x86 architecture, given its reliability and long-standing prominence in enterprise computing. Today, that dominance is being challenged with the world’s leading hyperscalers, including Amazon Web Services (AWS), Google Cloud, and Microsoft Azure, creating their own custom chips built on power-efficient designs, such as the Arm architecture, to deliver the high performance their AI-scale workloads demand.
At AWS for instance, more than half of the CPU capacity that it brings online now comes from its Arm-based Graviton line of chips, designed in-house for scale. The payoff from this move is significant, up to 40% better price to performance and 60% lower energy usage compared to the traditional x86 servers. This means platforms running on AWS, such as shopping, streaming, or banking, can scale large volumes of traffic and meet service-level demands with consistent, high performance and efficiency.
Google Cloud followed with Axion, it’s first custom data center CPU. The hyperscaler initially put it to work on heavy internal services like Gmail and Google Workspace, proving it could handle workloads with billions of daily users, without compromising efficiency. Since Google Cloud has made these chips container ready, its customers can rely on the same chips without rewriting their code. Meanwhile, for the end-user, it means Google’s productivity apps load instantly and more efficiently.
Microsoft Azure took a similar route with its Cobalt 100 processors, already powering parts of Teams and Azure SQL and ensuring the platforms remain responsive even under heavy workloads. Collaboration tools and databases are still among the most power-hungry workloads online and moving them to more efficient hardware makes the entire platform more sustainable.
As early adopters of new computing architectures, these hyperscalers are setting the pace for the rest of the industry and proving what’s possible when efficiency and performance are designed to scale together.
The Ripple Effect on Everyday Apps
This transition in the cloud and data center is rippling into the services that people and companies rely on every day.
Take Cloudflare, which secures and accelerates millions of websites. Cloudflare rebuilt its global edge servers around energy-efficient chips – a shift that lets it process ten timexs as many requests per watt compared with to its 2013 fleet. For users, that means faster load times and more resilient sites during surges.
Pinterest, home to billions of images, has migrated more than a quarter of its compute to these processors. This helps to keep feeds scrolling smoothly and image boards responsive, even when the traffic spikes during seasonal peaks. Meanwhile, Spotify’s move to Google Cloud’s Axion chips had led to 250% performance improvements on key workflows used by the app.
And Datadog, the monitoring platform used by thousands of companies, has shifted much of its Kubernetes fleet to the same efficient compute platform. The result is a dashboard that loads faster, has alerts that trigger on time, and has a more sustainable way to keep mission-critical systems observable.
For most users, the takeaway is simple. While the apps they use continue to feel as fast as ever, the infrastructure behind them is being redesigned to deliver more performance and work per watt, which further ensures that the experience stays seamless and efficient.
A New Foundation for the Internet
This cloud hardware shift is a structural reset for how the internet is growing. For decades, the focus was on throwing more raw power into the equation; however, that approach no longer works, with hyperscalers moving to processors built on efficiency-first designs.
Architectures once known for powering billions of mobile devices are now running the world’s largest data centers, doing more work at higher performance with less energy. While most users will never realize this fundamental shift, it’s making a difference – not just to global energy demand but to the app experiences and services that people use every day.
This is a radical rebuild of the internet will run in the years to come.
Discover how Arm’s efficiency-first architectures are transforming the cloud and powering the applications billions rely on every day.
