By Mariana Silvera
During Environment Month, an honest look at the environmental impact of technology — and the concrete steps we’re taking at Pyxis.
For a technology company, the environment isn’t something that happens outside. It happens in the systems we operate, in the code we write, in every architectural decision, in every AI model we train or use.
June is Environment Month, and at Pyxis it’s also a moment to take stock: where we stand, what we’re measuring, and what we’re actually doing. This article brings together global context, emerging standards, and the initiatives we’re advancing as an organization.
The growth of artificial intelligence use comes with an environmental cost that’s only beginning to become visible. A few reference figures:
What matters is that not all AI use consumes equally. A typical text query in a chatbot uses between 0.1 and 0.6 Wh — similar to a Google search. But agentic tasks, where a model plans, iterates, and chains multiple steps autonomously, can reach 30–50 Wh per execution: between 50 and 100 times more. As the use of AI agents grows, the impact scales non-linearly.
Three fronts show that the problem has concrete answers:
From academia, a July 2025 report by UNESCO and University College London found that using smaller models fine-tuned to specific tasks, shortening prompts and responses, and applying compression techniques can reduce the energy consumption of language models by up to 90% without affecting accuracy. Simply shortening interactions can cut consumption by more than 50%.
From applied research, the team at Tufts University published results this year on neuro-symbolic AI: an approach that combines neural networks with logical reasoning. Training consumption drops to 1% of a conventional system, and to 5% during use, with better accuracy.
From the market, platforms like GreenPT — hosted in Europe on 100% renewable infrastructure — show that it’s possible to offer AI services with greater transparency about their impact, reporting CO₂ and energy consumed per interaction.
Software has a footprint. And for some time now, there’s been a standard way to calculate it.
The Green Software Foundation developed the SCI (Software Carbon Intensity), a metric that measures carbon emissions per unit of useful work a system performs. Rather than an accumulated total, it’s a rate: how much carbon your software emits per transaction, per request, per model inference.
The SCI became the ISO 21031:2024 standard — the first standardized way to measure the carbon impact of software as a rate. In late 2025, the Green Software Foundation published two relevant extensions:
The GSF also offers free, self-paced online courses for those who want to incorporate these practices: SCI Fundamentals, SCI for AI Fundamentals, and Introducing SOFT. More than 130,000 people have already completed the Green Software Practitioner. Available at movement.greensoftware.foundation.
Our environmental commitment doesn’t stop at the declaration. There are concrete initiatives we’re advancing as an organization:
Zero Waste. We’re working to become a Zero Waste company: ensuring the largest possible share of the waste we generate is recycled, composted, or recovered, and that what ends up in landfill is minimal. Our goal is to achieve an 80% recovery rate. It’s ongoing work that depends on every person on the team, and we measure it continuously.
Carbon footprint measurement. Since 2022, we’ve been measuring our corporate carbon footprint using the GHG Protocol methodology. Four years of continuous measurement give us something concrete: our own data, real trends, and a basis for making decisions that don’t rely on generic industry estimates. In a year when the environmental impact of technology is under more scrutiny than ever, knowing where we stand is not optional.
Green software and artificial intelligence. At Pyxis, technical decisions around AI are not evaluated solely on what they produce, but also on how they produce it: what they consume, what they generate, what standard they follow. Sustainability is a cross-cutting criterion in our work with artificial intelligence.
Protea. We developed Protea, a platform designed to support the management, monitoring, and research of oceans, backing the work of scientists, public agencies, and organizations committed to marine conservation. Because our environmental commitment isn’t only in how we operate internally — it’s also in the solutions we build.
Two resources we recommend for those who want to go deeper:
Free courses from the Green Software Foundation: SCI Fundamentals, SCI for AI Fundamentals, and Introducing SOFT. Aimed at technical, architecture, data, and team leadership profiles.
movement.greensoftware.foundation
Agilists4Planet 2026: a free virtual conference on July 9–10, organized by the Agilists4Sustainability community. Speakers include Asim Hussain, founder of the Green Software Foundation. Topics range from digital sustainability to product and development decisions.
The environmental impact of software is here to stay on the agenda of any technology company that takes its role seriously. At Pyxis, we address it from multiple angles: measuring, improving internal processes, embedding sustainability criteria into our technical decisions, and building solutions that support environmental conservation.
We don’t have everything figured out. But we do have data, commitments we’ve made, and teams working on them. And that, we believe, is the difference between a declaration and a real commitment.
With a 360° potential, our solutions matrix accompanies the lifecycle of any project, with skills and experience in Development, Design, Q&A, Devops, Operation & Deploy, and Architecture
We are here to help you!
You can leave us your query or recommendation through this form.
I accept the terms & conditions and I understand that my data will be hold securely in accordance with the privacy policy.