If these data centers are going to be so profitable, then it should be simple to make guarantees about clean self-power, closed loop cooling, and noise and light pollution mitigation. There definitely shouldn't be deals to avoid taxes while lying about job creation.
Maybe if they did those things, there would be fewer permitting fights.
I agree with this in part. Funneling some of these massive investments into clean power, reduced water use, and noise mitigation, should certainly be a big policy priority, but I think actually achieving those things requires a lot of other policy flexibility. For example, if we want data center builders to use clean power, we also need to clear the way for them to actually build nuclear power plants. I suspect the water problem will solve it self as closed loop cooling systems come down in cost (and besides, 80% of data centers' net water use happens at the power generation stage).
Also, I am a big fan of data centers profit-sharing directly with communities.
Nuclear plants will not be built in the timescales these plants demand. No data centre is going to wait 5-10 years for a nuclear plant to be built in the USA, and the modular passively safe designs aren't ready.
I also seriously doubt the ability of the Trump admin to co-ordinate the required reforms and investments in nuclear energy.
Renewables, batteries and gas are probably going to be the only viable option.
There are few large scale US projects that don’t fall behind schedule so that’s not notable. It is notable that the demand-responsive pricing is so useful you can get 100 MW out of it by paying people for power in critical periods. And it is notable that a wind and solar company is Google’s other way out of this. How interesting.
Sooner or later the need for energy will become crucial. If it comes to war, then the US can build but if it comes to anything short of that, environmental concerns will dominate and probably make things unlikely.
It’s no wonder they want to put datacenters in space, calculating that it’s easier to solve the problems there than to solve the problems here. Though I’m surprised: are there no allied nations where we could put these things? The Trump admin has only been here since 2025 but opposition to these projects has existed locally for years. But universally? I imagine Japan or Korea, countries with less superstition around nuclear reactors, would have been happy to have done this.
People aren't stupid. They have very valid concerns about what AI will mean for their jobs over the next 5-10 years. Since AI companies (and companies in general) refuse to offer any real solution to being permanently unable to use your skills to create value that enables your survival, people are shoving back.
"Falling behind schedule" doesn't really seem like the right term, for a sector of the economy that has been accelerating for the past few years.
You could easily describe this trend positively rather than negatively, like:
"Google has built an incredible amount of datacenters in the past few years, which makes sense since Google Cloud revenue has tripled since 2021. But they are trying to grow even faster and add more revenue."
Disagree. The idea of rapid data center buildout, to the tune of gigawatts, is a major factor in the valuation of a bunch of companies. Overpromising and underdelivering, when such huge stock valuations are built on those promises, is a genuine problem. The hyperscalars can be building, but if they're building slower than their investors expect, it's still a real problem which can be fairly described as "falling behind schedule"
I think delaying the build-out of data centers is a good thing. If you've got a scarce resource, you either need to find a way of getting more of that resource or change how you use that resource to meet your end goal. The scarcity of data centers, compute, or storage should create the pressure to drive increased engineering to reduce demand for those resources.
We have seen some changes that improve LLM efficiency, but not enough yet. We need to put more pressure on the AI industry to reduce its resource use through engineering.
10 comments
[ 3.7 ms ] story [ 25.4 ms ] threadMaybe if they did those things, there would be fewer permitting fights.
Also, I am a big fan of data centers profit-sharing directly with communities.
I also seriously doubt the ability of the Trump admin to co-ordinate the required reforms and investments in nuclear energy.
Renewables, batteries and gas are probably going to be the only viable option.
Sooner or later the need for energy will become crucial. If it comes to war, then the US can build but if it comes to anything short of that, environmental concerns will dominate and probably make things unlikely.
It’s no wonder they want to put datacenters in space, calculating that it’s easier to solve the problems there than to solve the problems here. Though I’m surprised: are there no allied nations where we could put these things? The Trump admin has only been here since 2025 but opposition to these projects has existed locally for years. But universally? I imagine Japan or Korea, countries with less superstition around nuclear reactors, would have been happy to have done this.
You could easily describe this trend positively rather than negatively, like:
"Google has built an incredible amount of datacenters in the past few years, which makes sense since Google Cloud revenue has tripled since 2021. But they are trying to grow even faster and add more revenue."
We have seen some changes that improve LLM efficiency, but not enough yet. We need to put more pressure on the AI industry to reduce its resource use through engineering.