To determine the environmental footprint of software, transparency is needed throughout the supply chain. If digital infrastructure constitutes the supply chain of software providing digital products and services, then the first supplier in this chain is the server manufacturer — because the server converts electrical energy into digital resources, which fuel software.

Therefore, each server needs a transparent life cycle assessment (LCA), whose data must be available as open data in a machine-readable structure — queryable and usable by software calculating its own footprint. This assessment should include the estimated power consumption, but above all, the embodied resources and emissions of each device.

IDED (formerly known as SDIA) has proposed a data structure within its Open Data Hub Steering Group and is seeking feedback and support for building a unified database for hardware life cycle assessments. There are hurdles to overcome — inconsistent LCA methods, different configurations, non-machine-readable publications — but all of this can be resolved through industry-wide collaboration.

Different life cycle methods yield different results

A member of our community, involved in the preparation of IT hardware, pointed out that there is already a wealth of information available — for instance, from Dell (example) and HPE (example).

Interestingly, both companies in this example use the same LCA software — the Product Attribute to Impact Algorithm (PAIA) from MIT's Materials & Systems Laboratory. Once other LCA tools like OpenLCA come into play, the results deviate even more.

To tackle this challenge, the Öko-Institut in Germany has developed a simplified life cycle model that translates the configuration of a server into a comparable, emissions-focused life cycle assessment. IDED's work on the digital CO2 footprint and the SoftAWEre project is based on this simplified approach — due to the lack of robust life cycle data from manufacturers.

Manufacturers need a uniform configuration for comparability

Anyone who has ever used a configurator for planning a new vehicle will be amazed at the number of configuration options available per server — for example, in the configuration tool from HPE. The list of parameters is long: various processors, memory, chassis, storage configurations, riser cards, network components, power and cooling, advanced security features — and so on.

How can it be determined for which configuration a life cycle assessment should be conducted? Minimal and maximal configurations vary significantly per server. However, there are some approaches to make this manageable — two examples:

• Divide servers into categories (low-grade compute, mid-grade, storage, blade systems, etc.) and define a standardized maximal and minimal configuration per server category that applies across manufacturers.

• Define the life cycle costs per component and transform the configuration tool into a component-based LCA tool — each component in the server has its own environmental footprint. Thus, the LCA becomes configuration-specific rather than product line-specific.

In both cases, manufacturer agreement on the approach is a prerequisite — otherwise, the result is unusable for calculating the environmental footprint of software. When the PAIA method was developed, this was done through a pre-competitive consortium. This approach could be repeated to create an open data platform for life cycle data of server hardware — with IDED as the platform for the consortium and the infrastructure.

Manufacturers need machine-readable LCA data for practical use

An aspect often forgotten in assessments and sustainability reports: for data to be truly useful, they must be machine-readable.

An example: Anyone operating a fleet of 1,000 different servers, all recorded in the CMDB or IT inventory system, and wanting to determine the footprint per device or of the entire portfolio, cannot manually extract CO2 equivalents from PDF documents. What is needed: a database that, upon request, returns the LCA data including CO2 equivalents with the server name and configuration.

Boavizta, a group of developers and advisors advocating for sustainable IT in France, has converted PDFs from Apple, HPE, Dell, and many other manufacturers into such a machine-readable database.

The project uses an Open Database License to share benefits as broadly as possible. However, extracting data from PDFs is labor-intensive and not in the manufacturers' interest — the PDF is merely a snapshot of life cycle data, not reflecting the latest improvements.

The logical way forward is an open database where manufacturers can feed, regularly update, and provide configurable life cycle data. Such a database would enable comparability and research — but above all, it would allow software applications to consider the environmental impact of their own infrastructure for the first time.

It's time for server manufacturers to come together and create a common database

The competition for lower environmental impact will be a significant part of the future market. Transparency enables this competition and opens a new differentiation field for an otherwise largely standardized hardware market.

Furthermore, every IT department, procurement, and sustainability officer must include the embedded CO2 footprint of IT infrastructure — including servers — in their reports. Where are they supposed to obtain reliable data for this?

Now is the right moment to move forward together and create an open database for the life cycle data of server hardware. The process has begun — will you lead it or wait?