Thanks (as always!) for the post, John. Being in the same business as Generation, I'd generally agree with many of the points. But holding those companies having the largest datacenter buildouts 'accountable' to their climate commitments is impossible in practice. Firms like Amazon (which Generation owns and does engage with) will build out as much new power capacity as needed to meet their growth ambitions and will be comfortable with reducing carbon intensity even if absolute emissions go up (ironically not much different than an oil company). Unlike heavy industry, they do have the margins to do much of the buildout with renewables or eventually nuclear, but you will see deals for dedicated NG plants. Unsuprisingly, the public grid will backfill with cheaper NG baseload when existing nuclear or renewables (+storage) is recontracted to datacenters at higher rates. From a systems perspective, the accountability just isn't there, while the need to own the names is usually paramount.
Thanks John. Good to have another piece of credible research on this. Perhaps a counter-view is Hannah Ritchie's recent analysis of AI's energy / emissions profile (https://substack.com/home/post/p-151770312), albeit she concludes that uncertainty remains high. The key conclusion, for me, is not to panic, or try to put AI back in the genie lamp, but rather to be vigilant and to hold data companies accountable. Another entangled problem to solve
John, I've been a fan of yours since 1995 and have practiced People, Profit and Planet ever since. However, can I suggest that articles like this one are very one-sided. That's because the LLM providers don't want big energy bills either which put them at a significant disadvantage against their more waste reducing rivals. Plus, nowhere in any analysis do you ever see the dramatic positives generated by LLMs.
That's very different from the 3 Ps which clearly showed, and I proved in spades at my business, that increased profit but not at the expense of people and planet.
We might have a different view if we could draw a proper balance sheet.
Thanks for the insightful as ever article John. I just wanted to share a hot take...
The original paper and hence your article, and much of the media discourse around this topic, seems to conflate AI with Generative AI, particularly the current implementation of Large Language Models (LLMs). While everything stated is accurate for LLMs, it risks overshadowing other promising and valuable fields of AI that do not share the same energy and resource demands. This conflation may inadvertently slow progress in areas of AI that could deliver significant benefits with lower environmental costs - leaving us with higher emissions for longer.
Gary Marcus has discussed the limitations and overhype of Generative AI at length. He argues that the bubble is likely to burst soon due to diminishing returns and its structural inability to achieve the goals set for it. Meanwhile, other forms of AI and machine learning - like the efficient, problem-specific solutions developed by organizations such as OceanMind - operate with far less storage and compute power, often on modest budgets. Similarly, complementary fields like Neurosymbolic AI might offer a more sustainable path toward AGI, but they risk being sidelined unless some of the investment currently funnelled into Generative AI is redirected.
The key issue is that as long as we propagate the narrative that Generative AI is AI, we suppress investment in alternatives that could deliver better outcomes while being more sustainable. I don’t think this dynamic is unique to AI. Many industries see incumbents dominate the conversation, with viable alternatives suffocated by a lack of attention and resources - an effect often driven by large-scale corporate manipulation.
Another concern is what happens when the Generative AI bubble bursts. Who stands to lose money? And where will those losses be recouped? Will the billions be invested in renewable infrastructure—or will they flow into new fossil fuel facilities? Insights from our work tracking LNG trade suggest that the latter is a real risk. The LNG sector, for instance, is poised to grow by 300% in coming years, as countries pivot from coal to LNG for power generation. However, a new study highlights a troubling reality: methane leakage from transport ships and the energy demands of refrigeration make ocean-transported LNG more polluting than coal in terms of total CO2e emissions.
Finally, the question of investment also applies to infrastructure. Redirecting some of this capital to developing nations offers an opportunity to cut emissions more effectively than investments at home. Did you see the Climate TRACE event at COP29? Apple’s initiatives in this are interesting. These efforts offer some hope, but it’s hard to ignore the sleight of hand: focus on the good things there to distract from the bad things here. Cynically though, this might be the fastest way to drive meaningful progress.
(I asked ChatGPT to fix my grammar for this post, and it estimated I used 150 grams of CO2e in doing so, apparently the equivalent of using a 60-watt lightbulb for 5 hours. Sorry.)
Thanks (as always!) for the post, John. Being in the same business as Generation, I'd generally agree with many of the points. But holding those companies having the largest datacenter buildouts 'accountable' to their climate commitments is impossible in practice. Firms like Amazon (which Generation owns and does engage with) will build out as much new power capacity as needed to meet their growth ambitions and will be comfortable with reducing carbon intensity even if absolute emissions go up (ironically not much different than an oil company). Unlike heavy industry, they do have the margins to do much of the buildout with renewables or eventually nuclear, but you will see deals for dedicated NG plants. Unsuprisingly, the public grid will backfill with cheaper NG baseload when existing nuclear or renewables (+storage) is recontracted to datacenters at higher rates. From a systems perspective, the accountability just isn't there, while the need to own the names is usually paramount.
Thanks John. Good to have another piece of credible research on this. Perhaps a counter-view is Hannah Ritchie's recent analysis of AI's energy / emissions profile (https://substack.com/home/post/p-151770312), albeit she concludes that uncertainty remains high. The key conclusion, for me, is not to panic, or try to put AI back in the genie lamp, but rather to be vigilant and to hold data companies accountable. Another entangled problem to solve
John, I've been a fan of yours since 1995 and have practiced People, Profit and Planet ever since. However, can I suggest that articles like this one are very one-sided. That's because the LLM providers don't want big energy bills either which put them at a significant disadvantage against their more waste reducing rivals. Plus, nowhere in any analysis do you ever see the dramatic positives generated by LLMs.
That's very different from the 3 Ps which clearly showed, and I proved in spades at my business, that increased profit but not at the expense of people and planet.
We might have a different view if we could draw a proper balance sheet.
Thanks for the insightful as ever article John. I just wanted to share a hot take...
The original paper and hence your article, and much of the media discourse around this topic, seems to conflate AI with Generative AI, particularly the current implementation of Large Language Models (LLMs). While everything stated is accurate for LLMs, it risks overshadowing other promising and valuable fields of AI that do not share the same energy and resource demands. This conflation may inadvertently slow progress in areas of AI that could deliver significant benefits with lower environmental costs - leaving us with higher emissions for longer.
Gary Marcus has discussed the limitations and overhype of Generative AI at length. He argues that the bubble is likely to burst soon due to diminishing returns and its structural inability to achieve the goals set for it. Meanwhile, other forms of AI and machine learning - like the efficient, problem-specific solutions developed by organizations such as OceanMind - operate with far less storage and compute power, often on modest budgets. Similarly, complementary fields like Neurosymbolic AI might offer a more sustainable path toward AGI, but they risk being sidelined unless some of the investment currently funnelled into Generative AI is redirected.
The key issue is that as long as we propagate the narrative that Generative AI is AI, we suppress investment in alternatives that could deliver better outcomes while being more sustainable. I don’t think this dynamic is unique to AI. Many industries see incumbents dominate the conversation, with viable alternatives suffocated by a lack of attention and resources - an effect often driven by large-scale corporate manipulation.
Another concern is what happens when the Generative AI bubble bursts. Who stands to lose money? And where will those losses be recouped? Will the billions be invested in renewable infrastructure—or will they flow into new fossil fuel facilities? Insights from our work tracking LNG trade suggest that the latter is a real risk. The LNG sector, for instance, is poised to grow by 300% in coming years, as countries pivot from coal to LNG for power generation. However, a new study highlights a troubling reality: methane leakage from transport ships and the energy demands of refrigeration make ocean-transported LNG more polluting than coal in terms of total CO2e emissions.
Finally, the question of investment also applies to infrastructure. Redirecting some of this capital to developing nations offers an opportunity to cut emissions more effectively than investments at home. Did you see the Climate TRACE event at COP29? Apple’s initiatives in this are interesting. These efforts offer some hope, but it’s hard to ignore the sleight of hand: focus on the good things there to distract from the bad things here. Cynically though, this might be the fastest way to drive meaningful progress.
For anyone interested, this is the video of the Climate TRACE event at COP29: https://www.youtube.com/watch?v=_w2d-SnjXsU.
(I asked ChatGPT to fix my grammar for this post, and it estimated I used 150 grams of CO2e in doing so, apparently the equivalent of using a 60-watt lightbulb for 5 hours. Sorry.)