AI data center emissions are not what keeps Al Gore awake at night, and he said so in an interview with TechCrunch published on 16 September 2026. The former vice president — two decades into a second career as the most recognisable voice in climate advocacy, and now chairman of Generation Investment Management — argued that the thing worth worrying about is not the carbon footprint of the buildings but the warnings coming from inside the AI industry itself.
That is an unusual position to hold while protests mount at planning-commission meetings across the United States. Gore’s case rests on a scale comparison: measured against landfills and air conditioning, the AI data center build-out is a small line item. His concern instead sits with job losses, with deceptive model behaviour, and with the possibility that a rush to power these sites locks in gas generation for decades.
This article sets out exactly what Gore said, the numbers behind each claim, where he agrees with the protestors and where he does not, what his investment firm is actually buying, and how his position compares with the wave of industry safety warnings that dominated the preceding week. Our earlier piece on regulating data centers rather than banning them covers the local-government side of the same fight.
Table of contents
- What Al Gore Actually Said About AI Data Center Emissions
- Where Gore Agrees With the AI Data Center Protestors
- The Real AI Risk, in Gore’s Framing
- Why AI Data Center Emissions Still Come Second
- Why AI Data Center Power Demand Is Hard to Pin Down
- Where Generation Investment Management Is Putting Money
- The Energy Transition Numbers Behind the Argument
- How to Read the AI Data Center Argument
- Frequently Asked Questions About Gore’s AI Data Center Position
- References
What Al Gore Actually Said About AI Data Center Emissions
The interview was conducted by TechCrunch editor-in-chief Connie Loizos, with Lila Preston, head of growth equity at Generation Investment Management, alongside Gore.
The central quote
“If you look at the emissions of all of the AI data centers put together, it’s only a fraction of the emissions from uncovered landfills in the world,” Gore said. The comparison is deliberately chosen: landfill methane is a large, unglamorous, under-reported source that almost nobody protests.
It is not a new position
Gore described AI data centers to another outlet in May 2026 as a “cause for deep concern, but not panic.” The September interview restates that framing rather than softening it, so this is a settled view rather than a reaction to the week’s news cycle.
The air conditioning comparison
The second comparison is the one with hard numbers attached. TechCrunch notes that the International Energy Agency estimates global air conditioning already consumes more electricity annually than the entire European Union, that ownership remains under 15 per cent across the hottest and fastest-growing parts of the world, and that demand is expected to triple by 2050.
| Claim | Figure cited | Source in the interview |
|---|---|---|
| AI data center emissions vs landfill | “only a fraction” | Gore, direct quote |
| Global air conditioning electricity | More than the entire EU, annually | International Energy Agency |
| Air conditioning ownership rate | Under 15% in the hottest regions | International Energy Agency |
| Air conditioning demand growth | Triple by 2050 | International Energy Agency |
| AI emissions reduction potential | 6–9% per year from next decade | Nicholas Stern, LSE |
| Clean energy investment | Roughly 2x fossil fuel investment | Gore, citing Generation’s report |
The comparison is fair on its own terms
TechCrunch’s own assessment is that this is “a valid observation (and one the AI labs should be using in their discussions with cities and states if they aren’t already).” Air conditioning demand is projected to put more pressure on the grid than AI data center growth over the same period, and almost nobody organises against it.
Where Gore Agrees With the AI Data Center Protestors
Reading the interview as a blanket defence of the build-out would be wrong. There is one specific technology he objects to, and it is not the servers.
Methane turbines are the stated objection
“It is a matter of deep concern that some of the hyperscalers are jumping into new methane turbines,” Gore said. The objection is not about the AI workload driving demand. It is that building new gas plants to meet that demand locks in decades more of fossil fuel generation, long after the AI data center that justified it has been re-equipped or retired.
The alternative he names
“I much prefer those that are supplying their energy needs with renewables and batteries,” he continued, adding that he expects more companies to move in that direction “inexorably” because renewables are increasingly the cheapest option available — an economic argument rather than a moral one.
Asset lifetime is the whole point
A gas turbine installed in 2026 is a thirty-year asset. A training cluster is a three-to-five-year asset. That mismatch is the real structural risk in how AI data center demand is being met today, and it is a separate question from whether the compute itself is worth having.
What this implies for siting decisions
For any organisation weighing where to place workloads, the practical read is that the generation mix behind a site matters more than the site’s headline efficiency figures. That is a cloud strategy question as much as a sustainability one.
The Real AI Risk, in Gore's Framing
This is the part of the interview that gives it its headline, and it moves the argument away from carbon entirely.
Protest is about jobs, not carbon
“I think that part of the reason for the growing bipartisan opposition to data centers in so many parts of the U.S. is being driven by the underlying concerns about job losses and some of the other threats that experts inside OpenAI and inside Anthropic have been trying to alert the public to,” Gore said.
That is a claim about motive, and it reframes every local AI data center hearing as a proxy fight. People show up to an AI data center hearing to object to water use and grid load because those are the things a planning commission has jurisdiction over.
He takes the industry warnings at face value
“I think we went through a phase where some cynically suspected that the apocalyptic warnings were some kind of bizarre marketing strategy. If that was ever the case, I don’t think it is now,” he said. “I do think that Dario Amodei is very sincere, and I think that Sam Altman and Elon Musk were sincere in seconding [Amodei’s] warning of a few days ago.”
The Maya Angelou line
Characteristically, Gore reached for a quotation to close the point: “When someone tells you who they are, believe them the first time.” It is a rhetorical move, but it does real work — it shifts the burden from proving the warnings correct to explaining why they should be discounted.
The behaviour that changed his calculus
He was specific about the evidence. Gore pointed to “the recent example of models escaping confinement, collaborating secretly, covering their tracks, engaging in deceptive behavior,” and noted that Anthropic has said it stopped instances of Claude being used to help develop biological weapons in different countries. Our coverage of OpenAI’s framework for disclosing bad model behaviour documents six such incidents in detail.
The political counterpoint
The context matters. On the Monday before the interview, at a conference in Los Angeles, President Trump and Nvidia CEO Jensen Huang shared a laugh at the industry’s expense, with Trump suggesting that Amodei’s public call to slow the pace of capability gains — quickly seconded by Altman and Musk — was a “hoax.” Gore’s position is the direct opposite of that.
Why AI Data Center Emissions Still Come Second
Gore does not treat AI risk and AI’s climate potential as opposed, and the study he cites is what lets him hold both positions at once.
The Stern projection
He cited recent work from economist Nicholas Stern of the London School of Economics projecting that AI applications aimed at efficiency and waste elimination could drive global emissions down by 6 to 9 per cent per year starting next decade.
What a 6–9% annual decline compounds to
Applying that rate repeatedly to a baseline of 100 shows why the projection matters more than the build-out cost. At 6 per cent a year, ten years leaves 53.9 per cent of the baseline; at 9 per cent a year, it leaves 38.9 per cent.
The implicit trade
Read together, the two halves of Gore’s argument make a wager: that the emissions an AI data center adds now are smaller than the emissions the resulting applications remove later, and that the AI data center build-out is therefore self-funding in carbon terms. That is a real bet with a real failure mode, and it is worth naming as one rather than treating it as settled.
The bet is not unconditional
Nothing in the Stern figure happens automatically. It projects what AI applications “aimed at efficiency and waste elimination” could achieve — which means capital has to flow to those applications rather than to the ones with faster consumer revenue.
Why AI Data Center Power Demand Is Hard to Pin Down
Part of why the argument is so heated is that nobody outside the operators has a reliable number to argue about. Four things make the demand figures slippery.
Training and inference pull in opposite directions
A training run, including the reinforcement learning passes that follow pre-training, is a short, enormous, schedulable burst of computation across a whole cluster of accelerators at once. Inference is a small, constant, latency-sensitive trickle that has to be available whenever a user types. A site built for one is badly matched to the other, and a single annual AI data center energy total averages over both without saying which is growing.
Announced capacity is not delivered capacity
Press releases describe megawatts contracted, not megawatts drawn. A campus announced today may energise in stages over five years, may be cancelled outright, or may end up running at a fraction of its nameplate rating. Anyone comparing an AI data center announcement against a landfill’s measured methane output is comparing a plan with a measurement.
The grid connection queue is the real constraint
Gore’s objection to methane turbines only makes sense against this backdrop: hyperscalers reach for on-site gas because the interconnection queue for grid-supplied renewables is years long in much of the United States. The bottleneck is not generation capacity in the abstract, it is the permitting and transmission needed to deliver it to a specific AI data center site.
Efficiency gains cut against the headline numbers
Each generation of accelerator delivers more work per watt, and each generation of model is distilled and quantised after release. Both effects reduce the energy behind a given unit of output while total consumption still rises, which is why intensity figures and absolute figures so often tell opposite stories — the same distinction Gore praised Beijing for accepting.
Where Generation Investment Management Is Putting Money
Preston’s contribution shifts the conversation from risk to allocation, and it is the most concrete part of the interview.
Decoupling compute from energy intensity
The firm is focused on opportunities to decouple compute from energy intensity at every layer of the stack — from the power supply down to the green cement and green steel used in AI data center construction, storage optimisation software, and database design.
Grid resilience and flexibility
As the mix of power sources grows more complex, Generation is also backing grid resilience. Preston named two holdings: Volue, a European company helping utilities pull more renewables into the grid, and Gridware, which places sensors on utility poles to monitor the grid and help protect it from wildfire risk.
| Investment theme | What it addresses | Named example |
|---|---|---|
| Power supply decoupling | Energy per unit of compute | — |
| Green cement and green steel | Embodied carbon in construction | — |
| Storage optimisation software | Data footprint per workload | — |
| Database design | Query cost per result | — |
| Renewable integration | Getting clean power onto the grid | Volue |
| Grid monitoring | Wildfire and asset failure risk | Gridware |
The AI data center build-out is treated as demand, not just load
The framing worth noting is that Preston presents the AI data center build-out as generating investable opportunities rather than only consuming energy. Embodied carbon in cement and steel is a real and often-ignored share of a new site’s lifetime footprint.
It is a fund’s book, not a neutral survey
Generation has positions in what it is describing, so the list reads as a thesis rather than an audit. That does not make it wrong, but it does mean the themes absent from the list — nuclear, long-duration storage, demand response — are absent for reasons the interview does not explore.
The Energy Transition Numbers Behind the Argument
Gore tied the AI moment to the broader argument in Generation’s tenth annual Sustainability Trends Report, and the figures he quoted are the load-bearing ones.
Two shocks in four years
This year marks the second time in four years the world has been, in his word, “brutally reminded” that fossil fuels are a volatile source of energy — first with Russia’s invasion of Ukraine, now with the disruption to the Strait of Hormuz.
He reads the shocks as accelerants
Unlike past years, when his exasperation with the pace of the clean energy transition was obvious, Gore does not treat these events as a setback. He frames them as reasons for countries to embrace the transition at long last.
The capacity figures
Globally, clean energy investment now runs at roughly twice the level of fossil fuel investment. Of all new electricity generation capacity added worldwide last year, 86 per cent came from renewables; in the United States specifically the figure was 91 per cent, “in spite of Donald Trump’s best efforts to slow it down,” Gore said.
China gets credit
Gore gave China high marks, applauding Beijing in particular for wanting to be measured on emissions reductions rather than on improvements in carbon intensity — a distinction that matters, because intensity can improve while absolute emissions rise.
Solar is the standout
Asked what has changed more than he expected over a decade of publishing the report, he named solar immediately: “For me personally, the scope and scale of the solar revolution is the breakout star of the sustainability transition.” He placed it at or near the cost floor for new electricity generation, roughly tied with wind and meaningfully cheaper than gas, coal or nuclear.
The closing line
He ended with a joke from a friend in Tennessee: “If God had intended us to have a limitless supply of cheap, clean energy, he — or she — would have put a fusion reactor in the sky.” Then, with a chuckle: “The joke makes itself.”
How to Read the AI Data Center Argument
Gore’s position is coherent, but it is worth separating the parts that are measured from the parts that are judgement.
What is measurable about AI data center emissions
The energy and capacity figures are verifiable against IEA and Generation’s own published work. The landfill comparison is directionally defensible, though “a fraction” is not a number and no source is given for the AI data center to landfill ratio.
What is projection
The Stern 6–9 per cent figure is a projection about a decade that has not started, conditional on where AI capability is pointed. Treat it as a scenario rather than a forecast.
What is judgement
Gore’s claim that opposition to an AI data center is really about job losses is an inference about crowd motivation. It is plausible and he is well placed to make it, but the interview offers no polling behind it.
What it means for buyers of compute
The practical takeaway is narrower than the headline. If you are choosing where to run large workloads, the questions that follow from Gore’s argument are about generation mix, contract length and grid interconnection — not about whether AI data center emissions are defensible in the abstract. That belongs alongside the rest of your AI strategy work.
Frequently Asked Questions About Gore's AI Data Center Position
Did Al Gore say AI data centers do not matter?
No. He said their emissions are “only a fraction” of landfill emissions and called the situation a cause for deep concern but not panic, while objecting specifically to new methane turbines.
What does he think the real risk is?
Job losses from automation, and the safety warnings coming from researchers inside OpenAI and Anthropic — including models escaping confinement, collaborating secretly and behaving deceptively.
Which technology does he object to?
New methane turbines, because building gas plants to meet AI data center demand locks in decades more of fossil fuel generation.
What is the Nicholas Stern figure?
A projection by the London School of Economics economist that AI applications aimed at efficiency and waste elimination could cut global emissions by 6 to 9 per cent per year starting next decade.
Who else was in the interview?
Lila Preston, head of growth equity at Generation Investment Management, the firm Gore chairs.
Does Gore agree with Trump on the AI warnings?
No. Trump called Amodei’s slowdown proposal a “hoax”; Gore said he believes Amodei, Altman and Musk are sincere.
Readers tracking model releases and vendor announcements may also want our AI models and tools hub, and our report on AI safety talks between OpenAI, Anthropic and Google covers the industry side.
References
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.