Our Environmental Commitment
Approximately 667g CO2e per student per year. Our full emissions breakdown, methodology, and path to net zero.
Our Environmental Commitment
18 Nov 2025
Environmental responsibility matters to us. As an AI-powered education platform, we want to be transparent about our environmental impact, what we have done so far, and how we plan to improve.
Where We Stand Today
We estimate that the average Medly student generates approximately 667 grams of CO₂e per year.
We calculated this by taking our total estimated annual emissions across all operations (infrastructure, office, commuting, and cloud computing) and dividing by our active user base. This is a rough figure, not a third-party audited number, and we share it in the spirit of transparency rather than precision. As we grow, we intend to commission independent verification.
To put 667g in context using UK Government emissions factors[1]:
- A return economy flight from London to New York produces approximately 1.5 tonnes (1,500,000g) of CO₂e per passenger
- A return short-haul flight within the UK produces approximately 200kg (200,000g) of CO₂e per passenger
- Driving a petrol car for 4km produces roughly the same amount
In other words, a full year of AI-powered tutoring on Medly produces less CO₂e than a short car journey.
How We Keep Emissions Low
Infrastructure choices: We host on Google Cloud, whose data centres achieve an average Power Usage Effectiveness (PUE) of 1.10, compared to an industry average of around 1.56[7]. Google has also committed to matching 100% of its annual electricity consumption with renewable energy purchases since 2017[7], though this is based on annual energy credit matching rather than 24/7 carbon-free power at every facility.
Technical architecture: We use a multi-model approach that routes requests to smaller, more efficient AI models where possible, rather than defaulting to the largest available model for every task. This reduces compute per request. We cannot quantify the exact savings versus a single-model approach, but the architectural principle is to use the minimum compute needed for each task.
Cloud-native operations: We do not own data centres, vehicles, or manufacturing facilities. Our physical footprint is a small office and a remote team.
Our Emissions Breakdown
Here is our estimated emissions breakdown for the past year. These are internal estimates based on cloud provider usage data, office energy bills, and commuting surveys, not third-party audited figures.
| Category | Share | What it includes |
|---|---|---|
| AI and Cloud Computing | 47.4% | Platform infrastructure, databases, AI tutoring engine |
| Office Equipment and Leased Workspace | 34.8% | Laptops, monitors, desks, office space |
| Team Commuting and Remote Work | 14.1% | Travel to work, home office electricity |
| Direct Company Emissions | 2.5% | Heating, other direct operational emissions |
| Office Waste and Utilities | 1.2% | Waste disposal, office utilities |
Almost half of our emissions come from AI and cloud computing. This is the area where we have the most room to improve and the most direct control.
What We Are Doing Next
Near-term (0-6 months)
- Refining our model routing to further reduce unnecessary large model calls
- Expanding use of smaller, efficient models where they maintain educational quality
- Establishing internal per-user emissions tracking so we can measure progress
Medium-term (6-18 months)
- Beginning carbon offsetting through Gold Standard accredited carbon removal projects
- Involving students in choosing which verified offsetting projects we support
- Setting per-user emissions reduction targets
Long-term (18+ months)
- Working toward net-zero emissions across our operations
- Building emissions efficiency into every new market and product launch
- Exploring emerging green AI techniques as they mature
These are goals, not guarantees. We will report honestly on what we achieve and where we fall short.
A Note on Comparisons
You may see companies compare their per-user emissions to social media platforms or streaming services. We have chosen not to do this because the comparisons are rarely like-for-like. A global social media company's total corporate emissions (offices, hardware manufacturing, employee travel across dozens of countries) divided by billions of users produces a per-user figure that is not meaningfully comparable to a small UK startup's cloud bill divided by its user base.
What we can say is that 667g of CO₂e per user per year is a small number in absolute terms. We want to make it smaller, and we will be transparent about our progress.
Transparency First
We are committed to sharing:
- Our emissions estimates and methodology
- Carbon offset investments and their verified impact
- Honest reporting when we fall short of our goals
We welcome questions about our environmental approach. If you think we can do better, or if you spot something in our numbers that does not add up, we want to hear from you.
To our students: your future matters, and we are committed to protecting it while helping you achieve your potential.
The Medly Team
References
[1] UK Government GHG Conversion Factors for Company Reporting. Flight and vehicle emissions factors used throughout this article. https://www.gov.uk/government/collections/government-conversion-factors-for-company-reporting
[7] Google Data Centers, Operating Sustainably. PUE of 1.10 (fleet-wide trailing twelve-month average) and annual renewable energy matching since 2017. https://datacenters.google/operating-sustainably/
Our 667g per-user figure is an internal estimate calculated by dividing total estimated annual company emissions by active users. It has not been independently audited. Vehicle comparisons use UK average petrol car emissions of approximately 170g CO₂e per km (DEFRA 2023 factors).