Google’s exponential path to climate-wrecking digital bloat
136 points by sjamaan
136 points by sjamaan
This highlights what I dislike about the "nooo AI is actually not that bad for the environment" crowd: they wring their hands over specific quantities of water or watt-hours per query when it's indisputable that "AI", as an industry, is driving very bad environmental outcomes. We see it all the time, and here is another example. In their own words (as quoted by TFA):
[R]eaching our climate moonshot is getting harder. Growing our data center footprint to build out the infrastructure needed to make AI as helpful as possible to everyone requires energy and resources. [... O]ur AI infrastructure buildout is currently accelerating faster than the grid is decarbonizing.
We did not have to do this. We are breaking the world more and more for tools we did not need to build.
We did not have to do this. We are breaking the world more and more for tools we did not need to build.
This is the insight I'm missing so much from public discourse. It's always presented as inevitable technological "progress". As if progress is a straight line that we have no influence over!
And of course, if you object on these grounds, you're just a stupid Luddite. I'd like to see some more mindfulness in this space about what we're building, and why.
At least in tech spheres I don't like the focus on the median or average query or whatever, because they can also play this game where they say "well gee it's not any worse than a Google search." Well yeah, but if you leave your frontier LLM running overnight what exactly is the cost incurred?
But the people I know seem to only get use from running these things for periods much longer than what, say, Google's median LLM user would run them for. It's reasonable to suspect that the non-technical median user would make a few queries a day for simple questions and would generate an image or two. I could be convinced that such a thing might be sustainable (especially if we can continue to improve on inference and training costs). But that seems far from the kind of use of the CEO who claims to be 10x'ing --- or the kind of uses that might give Google any return on their massive investment.
If we're literally running out of gas turbines to build these data centers, maybe semantic games about power use are transparent misdirection. You can parcel up the power use into whatever small units you want, we're still burning a massive amount of oil to run this trash.
Just a thought.
If we're literally running out of gas turbines to build these data centers
This has been a problem for a lot longer as our industrial base collapsed. Lead times stretched out multiple years a while ago and limited stockpiles were drawn down when Russia started targeting Ukrainian power plants etc. (the same issue applies to transformers and a lot of industrial equipment. Even just auto motors have exploded in price in the past years.) All before the LLM and boom.
You may also ask why we need gas turbines to build data centers. The answer is that we're building gas pipelines to them directly.
I'd like to just offer one correction to your terminology while agreeing with you overall. Diesel-powered engines run emergency generators. They can be converted to natural gas rather easily, and are still called engines. Some generators run on kerosene, which is identical to jet fuel. These are called turbines. Turbines are basically jet engines.
Yes. This is not about emergency power generation, and not about covering gaps in power. This is about avoiding connections to the grid entirely. And, while we're also using generators, base generation for many data centers is being done with turbines.
https://www.gevernova.com/gas-power/resources/case-studies/crusoe-ai-data-centers-lm2500xpress
This is also not powered by green energy. There are gas pipelines being built directly to data centers for on site generation. Wait times for turbines have gone up years.
We are setting our future on fire so that we can accelerate enshittification.
I am very glad I don't have children.
This is also not powered by green energy
If they were, it would be an irrelevant distraction. Electricity is fungible. 1 MW of consumption means 1 MW of consumption. If your 1 MW is produced by green energy, that means 1 MW of green energy that isn't available for other uses. Unless it's powered by green energy from production projects built specifically to power that thing, then it's an accounting trick not actually a green shift. And, even then, the huge demand spike may be making green energy less affordable for other uses (up to a certain point, more demand means more economies of scale and pushes prices down. Beyond that, shortages in the supply chain mean that the price goes up).
I was depressed when I was at MS that there were people on the Sustainability Committee, who seemed to genuinely care about not making the planet uninhabitable, who would nevertheless justify running enormous machine-learning jobs because power to Azure could be met with green energy production.
I am very glad I don't have children.
I think this mindset is flawed. You can put yourself in any point in history and you will still find reasons to not have hope for the future. I mean I definitely won't argue the world is perfect, but it's a whole lot better than 100 years ago. or 200 years ago. or 300... Each generation fights its own fight, that's the way it goes.
To be fair, no generations other than those currently alive had to contend with accelerating anthropogenic climate change at this scale.
I mean I definitely won't argue the world is perfect, but it's a whole lot better than 100 years ago. or 200 years ago. or 300...
"The Dawn of Everything" by David Graeber and David Wengrow in part argues, I think quite effectively, that these kinds of statements are based on rather arbitrary criteria of "better" biased on the status quo. There is no linear progress of human cultures that can be assessed with a moral value system, the way we like to see it; it's more complicated.
no generations other than those currently alive had to contend with accelerating anthropogenic climate change at this scale.
No, but in the past, generations have had to deal with plagues that killed off 30% of the population (or 90% in the case of the New World contact), regular famines (less these days, which are now more a political problem than a natural problem) and as always, a high rate of infant mortality. It's always getting better (but not equally everywhere), but at the same time, the problems we do face change.
Does it mean Google is currently constantly burning more than 5GW in average? This is insane. It’s like five regular 1GW nuclear power plants!
Google’s power consumption isn’t just growing – the rate at which it is growing is growing. We have a word for this: exponential growth.
No, such a function could just as easily be quadratic, or any other higher order of polynomial. An exponential function is a function whose rate of change is directly proportional to its value.
While the textual description isn't strict enough, the graph (and data) clearly shows that f(x), f'(x) and f''(x) result in the same typical exponential graph.
The "Year-on-Year change" graph looks to me like a fairly linear trend for Google, but with a huge sudden jump at 2024/2025. Eyeballing it, I don't think an exponential curve would fit those data particularly well.
At some level this is the same ancient story of unpriced externalities. Impose a real carbon tax and the market might be able to sort it out.
But in this case, the people participating in the market seem to have lost all pretense of rationality. The investors would just put the tax in a footnote and keep shoveling money. They're investing in PowerPoint slides about sending datacenters into space, for gosh sakes.
In the absence of a carbon tax, I hope local communities everywhere are taking advantage of this massive hype asymmetry to tax the heck out of any datacenters that get built.
The main point made is that the IEA predicts world energy consumption to rise by 6000 TWh by 2030. About 3% of that rise is for data centres, including AI and cryptocurrencies.
https://www.worksinprogress.news/p/ai-is-bottlenecked-by-the-grid also claims that the bottleneck in the US isn't that we're not producing enough electricity, but that the queue to be connected to the grid is years long.
This blog post points to a really important discussion from the Internal Energy Agency (IEA) on the numbers that are given to estimate the cost of a single LLM query; the content is not so surprising, but seeing this in a credible report from an internal agency is important. I had trouble finding from the post (it is a screenshot of the text, and points to a tweet that points to the wrong document). It reasonates with discussions we have been having on Lobsters on whether the existing numbers given by AI companies are misleading, and whether it is reasonable to assume that the cost of model training are amortized by later usage (despite the explosion in number of new models being trained right now by many actors).
The quote is from the IEA's 2026 report, Key Questions on Energy and AI (see the PDF directly):
Box 2.1: How much compute does AI actually need?The first official per-query energy disclosures from major AI companies paint a surprisingly modest picture. Google reports that the median Gemini text prompt consumes 0.24 Wh (Google, 2025b); OpenAI puts the average ChatGPT query at 0.34 Wh (Altman, 2025); and a bottom-up estimation by Microsoft Research arrives at a median of 0.34 Wh for frontier-scale models with over 200 billion parameters (Felipe Oviedo, 2025). These figures date from 2025 and are likely already lower. Google itself reported a thirty-three-fold reduction in energy per prompt over the 12 months to May 2025, driven by advances in model architecture and hardware. As inference optimisation and next-generation accelerators continue to improve, per-query consumption can be expected to fall further
Even taking the 2025 figures at face value and allowing for a generous blended average of 1 Wh to account for heavier workloads such as reasoning and multimodal generation, scaling to 10 billion queries per day would imply annual electricity consumption of roughly 3.6 TWh. To put this in perspective, 10 billion queries per day is comparable to the estimated volume of all daily Internet searches worldwide (Felipe Oviedo, 2025) and roughly four times ChatGPT's reported 2.5 billion daily prompts as of mid-2025 (TechCrunch, 2025). For context, this is less than 1% of the 485 TWh that data centres consume today and would represent only a small draw on the tens of gigawatts of new IT capacity that leading AI companies are individually seeking to secure.
This arithmetic raises a question that the available disclosures do not yet answer. If text- based inference at massive scale would account for such a thin slice of projected electricity demand, the bulk of planned capacity must be destined for other workloads, such as large-scale model training, video and image generation, autonomous agents running multi-step pipelines, or enterprise deployments not yet reflected in public usage figures. Yet none of the companies that have disclosed per-query metrics have offered a comparable breakdown of how their anticipated capacity will be allocated across these categories, nor how per-query consumption is expected to evolve as agentic and reasoning-intensive use cases become the norm. In practice, AI inference is diffusing rapidly across the full spectrum of digital services, from search and recommendation engines to customer support, public administration and autonomous multi-step agent workflows. Understanding the composition of future AI electricity demand, not only the cost of a single prompt today but the full mix of increasingly complex and embedded workloads driving tomorrow’s infrastructure build-out, would give energy planners and policy makers a much stronger basis for anticipating what lies ahead.