Internet Carbon vs. GTA VI
Grand Theft Auto VI launches on 19 November 2026. Rockstar has spent 12 years and between $1 billion and $2 billion making the game. It's probably going to be highly successful: 230 million people bought the previous edition, and 10 million already pre-ordered.
Another hefty number: 212 Gigabytes, the size of the game download.
Because this is SustainableUX we can’t just enjoy the biggest event in gaming; instead we get to fret over what the environmental impact will be.
I tried to work this out using different models for estimating internet data carbon and got answers between 487,600 tonnes and 12.5 million tonnes of CO₂. Working out why these numbers vary so wildly forced me to update my understanding of the relationship between data and CO₂.
I'm strictly a dilettante when it comes to this sort of data science, so take all the below with a grain of salt.
Not about gamers
The carbon per player is modest, between 2.1 kg of CO₂ and 15.9 kg, depending on which sum you believe.
Staying indoors playing on a console is far cleaner than driving out to the countryside for a round of golf, let alone robbing banks or blowing up police cars. Gamers have no control over how big a game file is. This only becomes a climate story when you multiply it across 230 million downloads. Rockstar’s design decisions dictate how bloated the game becomes, while Take-Two’s distribution choices decide how that data reaches consoles, so they bear the most responsibility.

NB: exploding a real car emits about 10 tons of CO2, so it's much better for the environment to do this in-game.
Gigabytes to CO₂
How do we turn gigabytes into carbon? Take the total data transfer (file size multiplied by sales) and multiply it by the electricity intensity factor (kilowatt-hours per gigabyte). Multiply that by the carbon intensity of the electricity grid.
Total data (Game size (212GB) x sales (230m)) x kWh-per-GB x grid intensity = tons of CO₂
For the grid intensity, I went with 250g of CO₂ / kWh, the average across higher-income countries. Or you could go with the global average (494g per kWh) and double the numbers.
Data transfer: 46 quintillion bytes
An exabyte (EB) is one billion gigabytes, or one quintillion bytes (quintillion included because of its effervescent mouth-feel).
Global internet traffic is about 16EB a day. If GTA VI sells 230m copies, then the total data transfer is the equivalent of 3 days of global internet traffic, or 48.76EB.
Kilowatt-hours per gigabyte
This is the iffy bit.
My old understanding that energy-per-gigabyte is knowable was wrong. Wholegrain explain why, but for this article I'm just looking at a couple of top-down / bottom-up examples.
Top-down vs bottom-up
A top-down model starts at the macro level, taking the total annual electricity consumed by internet infrastructure across the globe and dividing it by total traffic. This spreads the entire cost of running and maintaining the infrastructure evenly across every gigabyte.
A bottom-up model works in the opposite direction by measuring the actual power drawn by internet equipment to see how much extra electricity a single gigabyte requires. Bottom-up engineering studies typically produce far lower figures per gigabyte than top-down allocations.
The models also vary on which bits of infrastructure count. There’s not space to get into that but here's Wholegrain again to explain.
Show your work
To keep this email within the word count, I’ve stuck the maths and links in a document.
Finding the answer: 4 models
All models are wrong, but some are useful. - George Box
1. Percentage of total internet (4.68 to 12.5 million tonnes of CO₂)
The most wrong method is to treat GTA's footprint as a simple fraction of global internet pollution.
Studies estimate that ICT accounts for between 1.5% and 4% of global anthropogenic greenhouse gas emissions.
Global emissions are 38 gigatons, so one day of internet traffic (16 exabytes) equals between 1.56 million and 4.16 million tonnes of CO₂. Multiply by three days: 4.68 million to 12.5 million tonnes of CO₂. 12 million tons is the annual emissions of Slovenia.
This is the ultimate top-down model. The 1.5% to 4% figure covers the entire global tech sector, not just the internet. So the result is massively inflated.
2. Sustainable Web Design Model v4 (2.36 to 3.66 million tonnes of CO₂)
The Sustainable Web Design Model (SWDM v4), maintained by the Green Web Foundation and partners, takes a more nuanced approach.
SWDM v4 assigns 0.194 kWh per gigabyte for operational electricity. Multiplied across 48.76 exabytes, GTA VI downloads require 9.46 terawatt-hours (TWh) of electricity. At 250g per kWh, that produces 2.36 million tonnes of CO₂, roughly equivalent to the annual emissions of 560,000 petrol cars. If you include SWDM’s figure for embodied hardware emissions, the total rises to 3.66 million tonnes.
Why this is wrong: SWDM is built around web browsing rather than gaming. Games are typically downloaded as compressed packages via edge servers, which is not the same as HTTP requests. For that reason, I switched my attention to data about streaming.
3. Green Streaming model (1.32 million tonnes of CO₂)
Research from Telekom Innovation Labs takes a bottom-up engineering approach, measuring electricity within telecom network segments:
- Core network: 0.020 kWh per GB
- Fixed home broadband: 0.070 kWh per GB
- 5G mobile broadband: 0.200 kWh per GB
If 86% of players download over fixed broadband (198 million copies at 0.09 kWh/GB) and 14% rely on 5G home internet (32 million copies at 0.22 kWh/GB), total electricity consumption hits 5.27 TWh. Total: 1,317,580 tonnes of CO₂.
4. Carbon Trust streaming model (487,600 tonnes of CO₂)
The Carbon Trust and DIMPACT also looked at media streaming.
If you take out the TV and look only at the data transmission, their model puts the footprint at roughly 10g of CO₂ per gigabyte. That comes out to 487,600 tonnes of CO₂.
Why this might not be a good model: it reflects the industry view that treats broadband infrastructure as fixed capacity, so it leaves out the embodied cost.
Bonus: Software Carbon Intensity (SCI) for Web
Measurement standards are catching up. The Green Software Foundation and W3C are currently developing the Software Carbon Intensity (SCI) for Web specification, which aims to measure actual energy draw across servers and networks, and should give far more accurate and nuanced data. However, it isn’t out yet.

So what?
Every model is an approximation. But even if they are very wrong (or more likely, my maths is), it seems probable that there is a big lump of carbon to account for.
Which sum do I believe? None of them, but probably somewhere around the two lower numbers. I haven't fully got my head around whether we should be counting the embodied cost of constructing and maintaining the internet as part of this.
Next issue, I’ll look at Take-Two’s impact report to see how they measure and manage their data carbon (or don’t). If I can tear myself away from the game, that is.