Responsible infrastructure at hyperscale: Managing the full lifecycle of Azure hardware

Every time we introduce a new generation of cloud infrastructure, we create two essential roles. We need to quickly deploy more advanced systems, and we also have to manage the old hardware responsibly.

Although these tasks are often unnoticed, they significantly impact Microsoft Azure users. The latest systems provide enhanced computing power while using space and electricity in datacentres more efficiently. Meanwhile, reusing, refurbishing, and recycling outdated equipment can prolong the lifespan of valuable components and reintroduce materials into the supply chain, promoting both cost savings and reliability.

This management process begins early, right from the design phase of our silicon, servers, networks, and datacentres. It continues throughout several years of operation. Once equipment is retired, Microsoft Circular Centers assist in determining its next useful purpose.

With new Circular Centers recently established in Newport, Wales, and Sydney, Australia, our operations now cover North America, Europe, and Asia-Pacific. These facilities are crucial for recapturing components, preparing spaces for new infrastructure, and maximizing the value we gain from our existing hardware and materials.

Maximising Every Rack and Watt

Cloud infrastructure must evolve to support the workload that customers require. This means boosting computing power while optimally utilising the available space and energy in each datacentre.

Today’s servers offer significantly more processor cores, memory capacity, and network bandwidth compared to earlier general-purpose systems. As illustrated in the chart below, Azure’s infrastructure has become both denser and more efficient over the years. Since 2014, cores per rack have soared by around 13 times, whilst the energy required for the same tasks has plummeted by nearly 90%. For instance, a workload like code compilation that previously needed about 100 watts can now be managed with less than 10 watts.

This increase in computing density enables us to provide greater server capacity to accommodate customer workloads within the same datacentre footprint. By extracting more significant performance from each rack, the latest infrastructure enhances resource efficiency, supporting ongoing growth and scalability to meet evolving demands for cloud and artificial intelligence.

Azure Cobalt exemplifies this comprehensive approach to systems. Our custom-designed cloud processor allows Microsoft to optimise silicon, servers, networking, storage, and software together. Cobalt 200 offers up to 50% better performance than Cobalt 100, integrating the latest in Microsoft’s security, networking, and storage innovations.

This ongoing innovation cycle is a pivotal aspect of cloud infrastructure. As we deploy newer cloud and AI technologies, older hardware is systematically retired. But what happens to the replaced equipment?

Extending the Life of Hardware

Last year, Microsoft achieved a commendable 92% rate of reuse and recycling for retired servers and components. However, given the scale at which we operate, retiring a server involves more than just sorting parts into recycling bins. Every server must be securely decommissioned, and all data-bearing components need to be properly wiped. After that, we evaluate how to maximise the value of the equipment. Can the entire system be repurposed? Can processors or memory modules be reused in different applications? Which materials can be recycled or returned to the supply chain?

Microsoft Circular Centers oversee these critical decisions as part of our datacentre operations. At these facilities, our teams evaluate the potential next uses for decommissioned hardware. Complete systems can still be valuable, supporting Microsoft labs and training environments, assisting educational institutions in developing technical skills, or being sold to eligible buyers for continued use. Components can potentially act as spare parts for existing systems. When reuse is impossible, materials are recycled responsibly.

Our Circular Centers operate as a connected global network. Following our recent openings in Wales and Australia, there are now eight centres across North America, Europe, and Asia-Pacific. We continue to expand, with another centre planned for San Antonio, Texas.

We’re also investing in technologies that enhance the efficiency and scope of our circular operations. With initiatives incorporating AI-driven solutions, such as robotic disassembly and automated material handling, we are speeding up the recovery of components and materials.

Optimising Every Resource

As the demand for cloud and AI services continues to escalate, each aspect of the infrastructure lifecycle is vital. We aim to extract maximum computing power from every rack and every watt while also getting the best value from every server and component. Our Circular Centers play a crucial role in achieving these goals. They integrate responsible recovery into our Azure expansion strategy, refresh our fundamental infrastructure, and enhance our capacity to meet future customer needs.

Explore Further

FAQs

• What is a Circular Center?
  A Circular Center is a facility where decommissioned hardware is evaluated and repurposed, refurbished, or recycled to ensure maximal value recovery.
• How does Microsoft ensure data security when retiring servers?
  Microsoft securely decommissions every server, ensuring that all data-bearing components are wiped clean of any information before reusing or recycling the hardware.
• Why is sustainable hardware management important?
  Sustainable hardware management helps reduce waste, optimises resource use, and supports a circular economy, promoting overall environmental responsibility.