Lowering streaming video quality: what impact does it have on CO₂ emissions?

First, where does the digital sector stand in global emissions?

Where do these figures come from?

Understanding what happens when you watch a video on YouTube or Netflix, for example

Before you click

A video is stored in several versions (high quality, medium quality, lower quality, etc.) on YouTube or Netflix servers. These servers are housed in data centers (large buildings filled with computing equipment).

When you click Play

Your device tells YouTube or Netflix: “Send me this video.”

That is when the delivery begins:

• 1 user watches the video = 1 delivery to make
• 1 million users watch it = 1 million deliveries

1. The data center: the warehouse

If you choose the 1080p version rather than the 480p version, you simply access one version rather than the other. Choosing the 1080p stream rather than the 480p stream makes relatively little difference to the data center's electricity use.

→ Lowering video quality makes little difference to the data center's electricity use

2. Networks: the delivery route

The video now has to travel from the data center to your device.

That is the role of networks, and this has nothing to do with social networks. Here, we mean fiber-optic lines, cables, routers, home Internet gateways, 4G and 5G antennas… Together, they form something like a road network that allows data to move around.

For networks, the difference is very tangible.

For 1 hour of video:

• 1080p → 2.25 GB
• 720p → 1.13 GB
• 480p → 0.50 GB
• 360p → 0.32 GB

Going from 1080p to 480p therefore means around 78% less data to transport.
How did we estimate these data volumes?

→ Lowering video quality = far less data to transport across networks.

In 2024, ≈ 65% of the data volume carried over networks was video streaming. How do we arrive at 65%?

3. Your device

The data finally reaches your phone, computer or television.

Your screen stays on throughout the video, whether it is playing in 1080p or 480p.

According to Fraunhofer, a device's electricity use depends mainly on factors such as screen size, display technology and brightness.

Displaying a video in 1080p rather than 480p makes little or virtually no difference to the device's own electricity use.

→ Lowering video quality makes little or virtually no difference to the device's electricity use.

In summary

The main benefit of lowering a video's quality is that it reduces the amount of data that networks have to transport.

We saw that the digital sector emitted 0.75 billion tonnes of CO₂e in 2024. According to Ericsson, networks account for 27% of those emissions:

78% less data does not mean 78% less energy

Reducing data volume does not reduce electricity use by the same proportion.

Yet one common method is to allocate all the electricity used by networks across the data they carry. Under this approach, the more data-intensive a video is, the more energy is attributed to it. For example, using the energy-intensity formulas adopted by Fraunhofer, if you watch the movie Inception the movie Inception season 1 of Friends all 10 seasons of Friends over Wi-Fi:

• 1080p → ≈ 525 Wh → ≈ 233 gCO₂e, equivalent to about 1.4 km by car (about 0.9 miles)
• 480p → ≈ 136 Wh → ≈ 60 gCO₂e, equivalent to about 0.4 km by car (about 0.2 miles)

But what these figures mainly answer is: “what share of networks' overall electricity consumption can be allocated to this video?”

They do not tell us how much electricity is actually saved at the moment you lower the quality.

In reality, the network is already consuming electricity before your video starts

Networks are not “virtual”: they are physical infrastructure that exists and consumes electricity continuously.

Of their total emissions, approximately:

• 15% → manufacturing and maintaining equipment (laying kilometers of fiber-optic cable on land and under the oceans, building and maintaining 4G/5G antennas, installing and replacing routers, manufacturing and launching telecommunications satellites…).
• 85% → the electricity needed to operate it (much of that electricity is consumed regardless of what happens, even when very little data is being transferred. Worldwide, around one third of electricity is still generated from coal, the fossil fuel that emits the most CO₂).

Networks are sized to carry the large volumes of data that flow during peak periods. But even when they are lightly used, their capacity remains in place, so they continue to consume a great deal of electricity. A large part of that consumption is therefore independent of traffic: carrying more data adds only a relatively small increment.

You can picture it as a train that is already ready to leave. One hour of video in 1080p represents 100 boxes. A 480p video represents only 22. Removing 78 boxes lightens the load, but it does not make the train, tracks, stations or signaling any smaller.

This is what the Power Model captures. It was developed from the work of Malmodin and later used by Fraunhofer.

Returning to our example: suppose you watch the movie Inception over Wi-Fi. Applying Fraunhofer's chosen values from this model, the network's estimated electricity consumption during viewing becomes:

• 1080p → baseline consumption + a small increment ≈ 42.6 Wh
• 480p → the same baseline consumption + a smaller increment ≈ 42.1 Wh

That is a difference of 0.5 Wh, despite 78% less data.

This does not mean the data volume does not matter. It means the networks are already there, powered on and sized to carry a lot of it.

Despite this small immediate increment, overall network electricity consumption is indeed rising: according to Ericsson, it increased by around 8% between 2020 and 2024.

Reducing data therefore changes the network's immediate electricity use only slightly, but the issue is still real, and it is mainly over the longer term that the effect can become significant.

The real effect comes over the long term, through infrastructure

The volume of data carried by networks keeps increasing.

As Mytton, Lundén and Malmodin explain, when existing network capacity approaches its limits during traffic peaks, operators have to reinforce the network: adding or replacing equipment, increasing infrastructure density in some areas, or deploying new technologies, as happened in the transition from 4G to 5G.

That new equipment then adds its own baseline electricity consumption, even when it is not being used at full capacity.

Network operation is therefore not simply proportional to the amount of data being carried at any given moment. Capacity tends to increase in steps instead: when the network approaches its limits, capacity is added; the network then has headroom again until the next upgrade is needed.

As Schien and co-authors show, a lasting reduction in the demand that contributes to traffic peaks can slow network expansion. Existing infrastructure then has more headroom and can absorb traffic growth for longer before it needs to be reinforced.

Over time, where the technology allows it, some equipment can also be put into sleep modes during off-peak periods, further reducing this baseline consumption.

This long-term effect is real, but its magnitude depends on many factors: how traffic evolves, the technologies used, and how operators size and run their networks. It is therefore not quantified in this article.

In summary

Less data does not immediately mean proportionally fewer emissions. The benefits emerge mainly when lower demand helps delay, reduce or avoid the addition of new network capacity.

Reducing video quality collectively and consistently over time → Less data transported → More capacity available during traffic peaks → Less pressure to reinforce the network → Over time, lower baseline electricity use and fewer emissions

ℹ️ If I live in a country with low-carbon electricity, is the impact negligible?

Generating 1 kWh of electricity does not produce the same amount of CO₂ in every country. In France, where nuclear power and renewables play a major role, electricity is far less carbon-intensive than in countries that rely more heavily on coal or gas.

These are the average emissions from generating 1 kWh of electricity in 2025:

FR41gCO₂e/kWh CA191gCO₂e/kWh GB217gCO₂e/kWh IT285gCO₂e/kWh DE330gCO₂e/kWh US384gCO₂e/kWh CN525gCO₂e/kWh IN670gCO₂e/kWh

Where do these figures come from?

Saving 1 kWh therefore avoids less CO₂ in France than in a country with more carbon-intensive electricity.

But watching a video from France does not mean that all the electricity needed to transmit it is generated in France. In exploratory tests carried out by Load Less (learn more), some YouTube videos watched from France were, for example, served from infrastructure located in Germany or Italy. Part of the electricity involved can therefore be more carbon-intensive than the electricity in the viewer's country.

And the estimates presented here concern electricity consumption and CO₂e emissions. They do not cover other environmental impacts on resources, air, water or soil.

Apply these settings automatically with the free Load Less extension

You can lower video quality manually on streaming platforms, but the settings have to be adjusted platform by platform, and they are not always retained from one session to the next.

To make those settings systematic, we created Load Less, a free extension for Chrome and Firefox.

It automatically applies and maintains data-saving video settings on major streaming platforms, including YouTube, Netflix, Prime Video, Disney+, Twitch and Vimeo.

It also reduces unnecessary data use, for example by lowering YouTube quality while a video is playing in a background tab, or by preventing some video players embedded on websites from loading until you actually play them.

The goal is simple: transfer less data in everyday use without having to think about it, and thereby help reduce, over the long term, the pressure placed on networks and the emissions that result.