1. Headline + intro
Sam Altman wants us to remember that humans also consume a lot of energy. That’s true – and almost entirely beside the point in the current AI debate. His latest comments in India, downplaying concerns about water use and reframing AI’s power draw as comparable to “training a human,” show where the industry conversation is heading: away from transparency and towards clever narratives. In this piece, we look at what Altman actually said, why the comparison to human brains is misleading, and how energy-hungry AI collides with Europe’s climate targets and regulatory agenda.
2. The news in brief
According to TechCrunch, OpenAI CEO Sam Altman spoke at an event hosted by The Indian Express during a major AI summit in India this week. Asked about AI’s environmental impact, he reportedly dismissed recent viral claims about ChatGPT’s water usage as “totally fake”, saying this was only relevant when data centres relied on evaporative cooling, which OpenAI no longer uses.
Altman acknowledged that AI’s overall energy consumption is a legitimate concern, but argued that discussions are often unfair. He pushed back on estimates that a single ChatGPT query might equal multiple iPhone battery charges, calling that highly exaggerated. Instead, he suggested people should compare the energy used to answer a question via an already‑trained AI model with the lifetime energy needed to “train” a human being to respond to the same query. On that basis, he claimed AI may already be competitive in energy efficiency. He also repeated his view that the world must rapidly expand nuclear, wind and solar power.
3. Why this matters
Altman’s intervention matters less for the specific numbers he disputes and more for the framing he is trying to establish. If the CEO of the most influential AI company can successfully shift the narrative from “AI systems are stressing power grids” to “humans are inefficient too,” the pressure for hard regulation and mandatory transparency weakens.
Winners from this framing are obvious: hyperscalers, model labs and cloud providers who want to grow AI usage as fast as possible without binding sustainability constraints. If AI becomes socially accepted as just another – or even more efficient – form of cognitive labour, scrutiny over data centre locations, grid impacts and water usage risks turning into background noise.
The losers are equally clear: local communities near data centres, utilities managing already‑strained grids, and policymakers trying to reconcile digitalisation with climate targets. Altman’s “humans also use energy” analogy blurs a critical distinction: individuals don’t typically decide where power plants are built or how water‑stressed regions host new server farms. AI rollouts are industrial decisions with concentrated impacts.
There is also a subtle rhetorical move here. By focusing on energy use per answered question, Altman sidesteps the two real issues: absolute consumption and growth rate. AI demand is not replacing some fixed quantity of human knowledge work; it’s creating whole new categories of activity – from always‑on copilots to endless synthetic content generation. Even if each query became marginally more efficient, a massive volume effect could still push total energy use sharply upward.
In other words, Altman is arguing about efficiency, while the public debate is increasingly about scale, transparency and governance.
4. The bigger picture
Altman’s comments land in the middle of a broader industry pivot: AI companies are moving from “wow factor” demos to the messy realities of infrastructure, regulation and externalities.
Over the past two years, independent researchers have published a series of papers estimating the carbon and water footprints of large model training runs and inference workloads. While numbers vary widely, the direction is consistent: frontier models are among the most energy‑intensive digital systems ever deployed. In parallel, several regions hosting dense data centre clusters have reported noticeable effects on local electricity demand and, in some cases, water stress.
This is not unprecedented. We saw a similar arc with crypto mining a few years ago: at first, enthusiasts highlighted how small the sector was relative to global emissions, then critics focused on where exactly the consumption was concentrated and who bore the side‑effects. AI is following a comparable path, but at a different scale and with far stronger political and economic backing from big tech.
Compared to competitors, OpenAI is trying to position itself as both aggressive and responsible: promising faster, more capable models while talking up the need for new clean energy sources, especially nuclear. Other players, such as Google, lean heavily on claims of carbon neutrality and matching energy use with renewables. Cloud providers increasingly market “green regions” and granular energy dashboards as differentiators for enterprise customers.
Yet across the industry, one thing is still missing: compulsory, comparable reporting. Without standardised, auditable disclosures on AI‑related power and water use, each company can pick its own narrative and baseline – exactly the dynamic Altman’s comments exemplify.
5. The European / regional angle
For Europe, this is not a theoretical argument. AI is colliding with three hard constraints: tight climate targets, ageing grids, and some of the world’s most demanding digital regulation.
On paper, the EU wants to be simultaneously the global leader in trustworthy AI and in decarbonisation. That means the energy footprint of data centres – many of which now serve AI workloads – is squarely on Brussels’ radar. Several member states already see local hotspots, from the data‑centre belt around Frankfurt to fast‑growing hubs in Dublin, Amsterdam and the Nordics.
European regulators are also more inclined than their U.S. counterparts to treat environmental transparency as a non‑negotiable. The EU’s new sustainability reporting rules will force large companies to disclose more detailed climate and resource data. It is hard to imagine AI infrastructure escaping that trend, especially as the EU AI Act starts to bite and the Digital Services Act raises expectations around platform accountability.
For European enterprises adopting AI, the risk is reputational as much as operational. Banks in Frankfurt, insurers in Zurich or telcos in Madrid cannot simply say, “Altman told us it’s efficient.” They will be asked by investors, regulators and customers to quantify the real footprint of their digital transformation, and to explain why certain workloads run in specific regions.
Put bluntly: even if Silicon Valley embraces Altman’s narrative, Europe is structurally set up to ask harder questions.
6. Looking ahead
What happens next will depend less on what Sam Altman says on stage and more on how three forces evolve: regulation, infrastructure and customer expectations.
On regulation, expect the European Commission and national regulators to move toward more granular reporting requirements for data centres and large AI providers, beyond generic corporate sustainability reports. The first step is transparency; the second will be incentives or caps in particularly stressed regions.
On infrastructure, the industry will push exactly what Altman advocates: more nuclear and more renewables, often tied to long‑term power purchase agreements. That is positive in climate terms, but it raises a political question: who gets access to new clean capacity first – critical public services and households, or commercial AI clusters optimised for serving chatbots and media generation?
On the customer side, large European corporates and public sector bodies will quietly become the de‑facto regulators of AI energy use. Procurement teams will bake sustainability clauses into cloud contracts, demand clear metrics on AI workloads, and prefer providers that can demonstrate low‑carbon regions and efficient models.
Watch for three signals over the next 12–24 months: major cloud providers publishing verifiable AI‑specific footprint data; local moratoria or constraints on new data centres in high‑stress areas; and investors pricing energy and water risk directly into AI valuations.
Altman is right about one thing: energy will define the AI era. Where he is less convincing is in suggesting that clever comparisons to human brains resolve the hard policy choices ahead.
7. The bottom line
Sam Altman’s attempt to reframe AI’s environmental footprint as “not worse than humans” is a smart soundbite but a poor guide for policy. The real questions for Europe and the wider world are about absolute scale, local impact and who controls the levers of expansion. As AI demand explodes, will we get binding transparency and smarter planning, or will we let narratives substitute for numbers? Readers – especially in Europe – should be pushing their providers, employers and regulators for the former.
