AI data centres are less thirsty now, tech giants say

Read the original at The Straits Times ↗
Cite

The Straits Times. (2026, September 7). AI data centres are less thirsty now, tech giants say. The Straits Times. https://www.straitstimes.com/world/ai-data-centres-are-less-thirsty-now-tech-giants-say

Dated when The Aporia collected the article, usually within hours of publication. Check the original for its exact date and byline.

The Straits Times · collected 2026-09-07 · by The Straits Times

Quick Summary

Tech giants such as Microsoft, Amazon Web Services (AWS), and Meta claim to have solved the water consumption issue at their AI data centres by implementing closed-loop cooling systems, which use no net water loss. According to Nvidia's June report, its DSX system can eliminate almost all water consumption at some facilities. Industry estimates suggest that data centres consumed 222 billion litres of water worldwide for cooling in 2025 and could nearly triple to 644 billion litres by 2030 if adaptive measures are not taken. The companies argue that using closed-loop systems allows them to reduce water use without increasing energy consumption, although some experts warn that there is still a trade-off between water and power usage.
Written locally by qwen2.5:14b on 2026-09-07, using this article's own text rather than the other coverage of the same event (that is the story summary below).

AI analysis runs on qwen2.5:14b, locally

Story summary

OpenAI CEO Sam Altman has made a statement downplaying concerns about the water usage of data centers like ChatGPT, saying that 38,000 queries use only as much water as it takes to grow a single almond in California. This comparison was made during an appearance on the "Sources" podcast, where Altman acknowledged that the industry's water usage is a concern, but argued that consumers' own actions have a greater impact on water consumption. He noted that modern data centers no longer use evaporative cooling methods, which were once criticized for their high water usage. In contrast, data centers globally consumed 222 billion liters of water in 2025, according to consultancy Rystad Energy, and this number is expected to triple to 644 billion liters by 2030 without adaptive measures.

Written for “Environmental Impact of AI” on 2026-10-22, grounded in this article and the 1 other(s) covering the same event.

Story

📰 Environmental Impact of AI
Environment · 2 article(s) covering the same event. See how they differ ↓

Compared with similar articles

This article reads unscored and hedges 27% of its claims. Each row says how that neighbour differs.
New York Post
⚖️ Leans right 🔴 14% hedged 4 of 28 📰 publisher trust 66
“Both articles mention OpenAI CEO Sam Altman addressing concerns over ChatGPT's water usage, implying a recent and likely identical public appearance or statement.”
NPR
⚖️ Leans left 🔴 9% hedged 4 of 45 📰 publisher trust 60
“Article A discusses political advertising and campaign spending on data centers in various states, while Article B reports on tech giants' efforts to reduce water consumption in data centers, indicating separate topics”
Evening Standard
⚖️ Leans strongly left 🔴 0% hedged 0 of 27 📰 publisher trust 65
“Article A reports on protesters calling for a moratorium on AI data centres in Larbert, Scotland, while Article B discusses tech giants' claims that their AI data centre water consumption is solvable, and mentions a specific report from Nvidia but does not relate to the same event or location”
Los Angeles Times
⚖️ Leans left 🔴 12% hedged 8 of 64 📰 publisher trust 95
“Article A discusses public backlash against data centers on fairgrounds in California, while Article B mentions tech giants' claims that new technology can reduce water consumption at data centers”
The Independent
⚖️ Leans left 🔴 15% hedged 12 of 78 📰 publisher trust 59
“Article A discusses President Trump's tweet on data centers and its impact on midterms, while Article B talks about tech giants' efforts to reduce water consumption in data centers”
Washington Examiner
⚖️ Leans strongly right 🔴 3% hedged 1 of 31 📰 publisher trust 66
“Article A discusses a Gallup survey and general public aversion to data centers, while Article B reports on a specific solution offered by Nvidia to reduce water consumption in data centers”
Fox News
⚖️ leaning not scored 🔴 14% hedged 3 of 21 📰 publisher trust 67
“Article A mentions protests against AI data centers and Flock cameras, while Article B focuses on tech giants' efforts to reduce water consumption at data centers”
NBC News
⚖️ Leans left 🔴 8% hedged 3 of 39 📰 publisher trust 95
“The articles discuss different aspects of data centers and their impact on the environment, but there is no mention of a shared specific incident or event.”
The Globe and Mail
⚖️ Leans left 🔴 14% hedged 5 of 35 📰 publisher trust 53
“Article A reports on a federal framework announcement by Minister Evan Solomon in Markham, Ontario, while Article B discusses data centres' water consumption and mentions Nvidia's report from June without referencing the specific event or location mentioned in Article A”
Semafor
⚖️ leaning not scored 🔴 8% hedged 2 of 24 📰 publisher trust 60
“Article A discusses the competitive talent market for law firms related to data center deals, while Article B reports on tech giants addressing water consumption concerns at their facilities, which are a type of data center”
How this article was measured signals, leaning charts, publisher record, byline, topics and claims

Signals How these are calculated →

Claims extracted
26
sentences that state something as happening or true
Uncertain
27%
7 of 26 hedged ("reportedly", "may", "according to")
Leaning
not political
takes no side on a contested political question
Publisher signals
68/100
The Straits Times's corrections and hedging, across everything collected from it. Not a measure of accuracy.
Outlets on this story
2
Environment
Same framing elsewhere
1
articles that frame the event the way this one does
Analyzed 2026-09-07 · how these are computed

How this is being covered How these are calculated →

Article leaning vs. publisher reliability
Source leaning vs. consistency

Publisher

The Straits Times · 3359 article(s) · 4 correction(s) detected
Running correction rate · 4 correction(s)

Who wrote this

The Straits Times
2145 article(s) here · 1 carrying a prediction
🔮 French, Belgian and Polish diplomats and a French mayor were attacked by masked settlers in another part of the West Bank on Saturday.
🔮 MOUNTING THREAT Merz said on Thursday the arrest did not surprise him and warned that there would be more such cases in the future.
2026-10-22 · assertive framing · Suspected spy for Russia arrested in Germany
🔮 Uganda's health ministry said the woman had travelled to Congo, where the outbreak was first declared in mid-May.
🔮 The epidemic, first detected 3,000 km away in Democratic Republic of Congo in May, is the second-largest on record and had already hit Uganda and Kenya before the Nigerian case was confirmed on Tuesday — the widest known spread on the continent.
🔮 France’s Macron tightens curbs on protests, blaming hard left for unrest PARIS, Oct 22 - French President Emmanuel Macron said on Thursday that "illegitimate" protests would be banned in the coming days and blamed "seditious" ultra-left groups and vandals for last weekend's riots and looting.
🔮 At least nine publicly identified Americans are believed to remain imprisoned in Russia, although the full number may be higher.
2026-10-22 · assertive framing · Who are the Americans detained in Russia?
🔮 The police and other officials would “carry out all the necessary procedures to identify the repatriated deceased”, it said.
2026-10-22 · assertive framing · Ukraine receives 252 bodies in swop with Russia
🔮 The island's top transportation official estimates passenger trips on public transportation would plunge 75% this year versus last.
🔮 But Reuters/Ipsos polling and interviews with white evangelical Christians show they remain his most reliable constituency and could prove crucial in the Nov 3 midterm elections.
🔮 Melenchon, who launched his fourth presidential bid in May and remains the hard left's strongest hope of reaching next year's election run-off, has faced mounting criticism since unrest engulfed several French cities, with opponents saying years of confrontational rhetoric helped fuel tensions now spilling onto the streets.
Also by The Straits Times
Suspected spy for Russia arrested in Germany
2026-10-22 · The Straits Times
Nothing else under this byline is closely related to this article, so these are simply their most recent.
All 2145 articles by The Straits Times →

Topics

AWS Microsoft NEW YORK Nvidia SpaceX

Subjects

Nvidia ORG · 5× AFP ORG · 2× AWS ORG · 2× Microsoft ORG · 2× Rystad ORG · 2× SpaceX ORG · 2× American NORP · 1× NEW YORK GPE · 1× Rystad Energy ORG · 1× the United States GPE · 1×

Narrative

Nvidia’s new method uses a closed-loop cooling system in which liquid flows directly through the servers, as close as possible to the chips, whose temperature can rise above 80 deg C. The problem is that “there’s a pretty direct trade-off between how much water is used and how much energy is used” for temperature control, said Andy Masley, an independent researcher who covers AI and data centres.
framing: mixed · carried by 1 article(s) · first seen 2026-09-07
🔮 Without adaptive measures, the figure could nearly triple to 644 billion litres by 2030, Rystad estimates.
2026-09-07 · The Straits Times
AI data centres are less thirsty now, tech giants say · mixed framing

Claims (26 extracted, 7 hedged)

AI data centres are less thirsty now, tech giants say NEW YORK – Data centres are running into a wall of public anger in the United States over their thirst for water and power. asserted
centres → say → water
The tech giants spending billions of dollars on them say the water part, at least, is solvable. asserted
part → spend → them
Data centres are the warehouses full of computer servers that run the internet and, increasingly, artificial intelligence. asserted
that → run → internet
The computers get hot and keeping them cool takes water. asserted
keeping → get → water
American chip giant Nvidia said in a June report that it can eliminate water consumption almost entirely at some facilities when deploying DSX, its newest system for designing and managing AI data centres. asserted
it → say → centres
It is a bold claim and one that the industry is under mounting pressure to make good on. uncertain
industry → mount → pressure
In 2025, data centres consumed 222 billion litres of water worldwide for cooling, according to consultancy Rystad Energy. uncertain
centres → consume → Energy
Without adaptive measures, the figure could nearly triple to 644 billion litres by 2030, Rystad estimates. uncertain
Rystad → triple → 2030
Steps to limit the growth could keep it under double. uncertain
Steps → limit → double
Nvidia’s new method uses a closed-loop cooling system in which liquid flows directly through the servers, as close as possible to the chips, whose temperature can rise above 80 deg C. The problem is that “there’s a pretty direct trade-off between how much water is used and how much energy is used” for temperature control, said Andy Masley, an independent researcher who covers AI and data centres. asserted
who → use → AI
Cutting back on water use usually means more power as the liquid in those sealed pipes still has to be cooled down somehow, usually by blowing air over it – and that takes electricity. asserted
that → cut → electricity
Nvidia gets around some of this by letting the liquid enter the servers warmer than usual, at 45 deg C. Most other closed-loop systems ran at about 32 deg C in 2024, according to the Uptime Institute, which certifies data centres. By starting with warmer water, Nvidia does not need to pump in cooled air year-round. uncertain
Nvidia → get → air
Josh Parker, Nvidia’s head of sustainability, said “simple fans circulating the air” are often enough, although sometimes, a mix of methods is needed. asserted
mix → say → methods
Public relations Microsoft, Amazon Web Services (AWS) and Meta told AFP that they also use closed-loop systems, which they said involve no net water loss. asserted
they → tell → loss
The two cloud computing giants, which have been expanding their already huge data centre footprints, used more water overall between 2022 and 2025, but the water-use efficiency improved by 25 per cent at Microsoft and 37 per cent at AWS, according to their most recent sustainability reports. uncertain
efficiency → expand → reports
There is no industrywide consistency, however, in how companies report data on so-called environmental, social and governance (ESG) efforts. asserted
companies → be → efforts
Elon Musk’s SpaceX, now a major player in data centres after it acquired his AI company, xAI, has never published an ESG report. asserted
it → acquire → report
In June, ratings agency MSCI gave SpaceX its lowest ESG score. asserted
MSCI → give → score
Ren Shaolei, an engineering professor at the University of California, Riverside, said that “because water is generally much cheaper than electricity”, companies have less incentive to cut water use on cost grounds alone. asserted
companies → say → grounds
There are incentives, Ren said, but they have more to do with public relations amid the growing backlash to data centres across the US. asserted
they → be → US
Another obstacle is that upgrading an older data centre to newer, less thirsty technology is expensive. asserted
upgrading → upgrade → technology
This may matter less than it sounds. uncertain
it → matter → ?
Older data centres are smaller and less powerful than the enormous new ones being built now, so they need less cooling in the first place, said Minh Khoi Le, who leads data centre and hydrogen research at Rystad. asserted
who → build → Rystad
And the water a data centre uses directly is only part of the story. asserted
centre → use → story
Water is used to generate the electricity that powers the data centre and to manufacture its chips and servers. asserted
that → use → chips
In the US, that hidden water use can be twice the amount a data centre consumes on its own. asserted
centre → hide → own
💬Give feedback
🕘History 🎫Support