Welcome to the week.

When AI datacenters first started to become a thing, some optimists thought that maybe this was an opportunity for the clean energy boom.

And while that opportunity still exists in theory, a lot of the power to run them isn’t coming from the clean grid. For example, Amazon alone is backing 17.5 gigawatts of new (keyword!) gas generation to power its data centers (about 1.5x what New York City uses at peak demand, by the way).

And of course, the way this all gets paid for is by shifting costs onto regular households.

This obviously will not stand. So let’s spend some time understanding the current picture, and of course, what we can do about it.

Let’s go.

— Willow

APL Positions of the week:

⚡ 100% clean energy by 2035 (get the metrics)

🏭 Large energy users pay fair rates (get the metrics)

👉 NEW: Explore all APL Positions, evidence for them, and how we measure their status here.

{{active_subscriber_count}}+ people who give a shit got this post in their email, for free.

Three new gas plants tied to Amazon’s AI buildout are now in motion: a 7.65 GW plant planned in West Texas, a 4.5 GW plant under construction in Pennsylvania, and 2.3 GW across three Entergy-built plants in Mississippi (a $3.4 billion investment). That’s all part of the 17.5 GW of new fossil-fuel generation Amazon is backing overall (Cleanview).

Meanwhile, Texas Governor Greg Abbott just ordered the state to stop issuing new environmental permits for data centers altogether, until audits of their electricity use, water consumption, and community impact are complete (The New York Times).

AI’s power demand is growing faster than the clean grid can keep up, and so someone’s just building gas plants instead to fill the gap, and we still need to figure out who pays for it.

Natural gas is already the single biggest source of electricity for US data centers today, responsible for over 40% of the current supply. That number is expected to go up: through 2030, gas is projected to add 130 TWh of new generation for data centers, more than any other source (renewables are second, adding about 110 TWh over the same stretch).

Globally, data centers used about 1.5% of the world’s electricity in 2025. In the US, it's closer to 5%. Some states go far beyond even that: Virginia is over 25%, and several other states exceed 10%.

The AI boom has created a power problem, and gas is the fast answer because it can be built quickly enough to match the pace of data center construction.

Clean generation takes longer to permit and build, so gas fills up the gap by default, even though it’s not the better long-term choice.

Data centers are large, concentrated electricity consumers, and when new gas plants and grid upgrades get built to serve them, the costs don’t always stay with the company that caused them. Instead they can show up on residential rates.

That’s a big reason why there’s a 71% opposition to local data centers, and why Texas is now trying to audit before approving anything else.

If we can’t put AI back in the box, then we need to figure out how to build clean energy to power it fast enough, while also considering who ends up holding the bill in the meantime.

👉 NEW: Find every action recommended in It’s Called Science right here.

Last week, we asked: Measles vaccine coverage needs to hit what percentage to reliably stop community spread?

You said:

⬜️⬜️⬜️⬜️⬜️⬜️ 70% (6%)
⬜️⬜️⬜️⬜️⬜️⬜️ 80% (3%)
🟨⬜️⬜️⬜️⬜️⬜️ 90% (18%)
🟩🟩🟩🟩🟩🟩 95% (74%)

The correct answer: 95%. Measles is so contagious that even a small gap in coverage is enough for it to find a foothold. Most of you nailed it! Here’s a gold star ⭐

Members get a 30 day free trial to:

— Willow

Check out what else we make (or take it easy on the emails, we get it) right in your Subscriber Preferences.

☎️ Work with Quinn 1:1 (slots are extremely limited):
book time to talk climate strategy, investing, or anything else.

🎯 Sponsor Important, Not Important ":
reach {{active_subscriber_count}} (and counting) sustainably-minded consumers across our newsletters, web, and audio.