How clean does electricity need to get for the European ICT sector to half its emissions by 2030?

This post from John Booth from Carbon 3 about the recent coming European Energy efficiency directive, got me thinking about what kinds of changes are needed sector wise to meet their own targets by 2030.
Here’s the good quote he draws attention to in his post:

‘The ICT sector is another important sector which receives increasing attention. In 2018 the energy consumption of data centres in the Union was 76,8 TWh. This is expected to rise to 98,5 TWh by 2030, a 28 % increase. This increase in absolute terms can also be seen in relative terms: within the Union, data centres accounted for 2,7 % of electricity demand in 2018 and will reach 3,21 % by 2030 if development continues on the current trajectory.

EU Energy Efficiency Directive, published in July 2023

2018 for the EU 76TWh, at a carbon intensity of what – 400 grams per kilowatt hour? Let’s check Ember:

Chart showing European carbon intensity vs the world. It has fallen from the high 300s the low 300s

OK, That’s 320 kilos of CO2 per megawatt hour, and 76 TWh is 76,000,000 megawatt hours.

So that’s about 24.3 million tonnes of CO2. Let’s call it 24 million to make the numbers easy.

For 98.7 TWh, and a carbon budget of half the figure above of 12 million tonnes of CO2 as the target for 2030, you’d need carbon intensity of about 120g of CO2 per kilowatt hour for your electricity.

So, you’d need electricity to be nearly three times cleaner for this to be possible.

I’m now wondering how much additional clean power you’d need to add to the grid, to avoid the residual carbon intensity of everyone else looking horrendous, and how much fossil energy you’d need to displace to reach that figure.

This is the kind of problem I think you can use PyPSA for, and I think it’s possible to repurpose some recent research published by TU Berlin, that was commisioned by Google to help them answer some relevant questions about decarbonising their own infrastructure, and powering it with renewables 24/7.

If you’ve seen any studies along these lines, hit me up.

Note: I don’t think the figure above includes embodied carbon in datacentres – only energy use. Right now, that feels like a wild card, and I haven’t seen any targets or figures from that yet, that I feel I understand well enoufgh to be comfortable talking about.