Google and Amazon report record energy use as AI data centers challenge climate goals

Google's AI buildout drove 37% increase in electricity use in 2025

Google reported that its annual electricity consumption rose by 37 percent in 2025 -- the largest increase in the company's history as Silicon Valley's AI data center buildout continues. But the tech giant says it kept operational carbon emissions down by continuing to purchase massive amounts of clean energy. The company's latest sustainability report acknowledges that Google's total electricity usage has increased by more than 250 percent since 2019, which the company attributed to ongoing growth in Google Cloud, YouTube video streaming, and data center construction and operations supporting various AI products and services. The unprecedented 37 percent annual increase is part of an ongoing trend, given how Google's total electricity consumption also grew by 27 percent in 2024. "While the path to achieving our climate ambitions will not be linear -- given our AI infrastructure buildout is currently accelerating faster than the grid is decarbonizing -- we remain focused on scaling abundant and affordable clean power globally and progressing technological innovations that drive down emissions across our operations and the broader industry," according to the Google sustainability report. As the bulk of electricity usage, Google's data centers consumed more than 42 million megawatt-hours of electricity in 2025 compared to 30.6 million megawatt-hours of electricity in 2024. That means Google's data center energy usage rivals the electricity consumption of entire countries such as New Zealand, Denmark, and Nigeria. More self-reported details from Google are available in the full sustainability report, including water consumption numbers for various data center campuses. Despite the steep rise in electricity consumption, Google reported actually reducing operational emissions by 2 percent over the same year period. Such apparent "decoupling of electricity-related emissions" from growth in electricity usage is potentially promising, but Google said it would need to boost "clean energy investments and closer partnerships with local stakeholders in the years ahead." The company also noted that its supply chain emissions from its contracted manufacturers and suppliers actually grew by 25 percent because of "an Asia-Pacific supply chain operating on grids that remain undersupplied with carbon-free energy." The carbon-free energy goal Google has been matching 100 percent of its electricity consumption with renewable energy purchases every year across its global operations, a practice that it has adopted for nine years running. In 2025, the company's purchase agreements encompassing 12 gigawatts of "net-new clean energy" represented the largest annual total in the company's history. But it also acknowledged that companies can merely claim to rely on 100 percent renewable power while still using power generated by fossil fuels in local power grids. Accordingly, Google says it has renewed its emphasis on a "24/7 carbon-free energy ambition" that gets into more granular accounting by focusing on certificates representing hourly and local matches for clean energy. To its credit, Google has become one of the world's largest investors in clean energy technology, according to Michael Thomas, CEO of the Cleanview data platform that tracks renewable energy and data center projects. But his analysis of Google's data center power strategy suggests that Google has pivoted to an "Everything Everywhere All at Once" strategy that encompasses everything from renewables to natural gas generation. Google's broader investments include efforts to commercialize "advanced nuclear, fusion energy, enhanced geothermal, long-duration energy storage, and natural gas with carbon capture and storage." The company's sustainability report highlighted investments of more than $3.8 billion between 2010 and 2025 that are expected to bring 7.5 gigawatts of clean energy online. However, in an April 2026 newsletter, Thomas pointed to Google's $40 billion investment in Texas data centers as including a campus potentially powered by a 933-megawatt natural gas power plant without carbon capture technology. The natural gas plant's turbines could produce 4.5 million tons of carbon dioxide annually. A Google spokesperson told Thomas that the company has not yet signed an agreement for how much power Google's data center might draw from the natural gas plant.

The cost of the AI boom: Amazon emissions jump 16% as company stands by net-zero pledge

Amazon's carbon footprint jumped 16% last year after several years of little or no increase. The company emitted nearly 80.9 million metric tons of carbon dioxide equivalent in 2025. By comparison, that's slightly higher than the nation of New Zealand's emissions. Amazon disclosed its climate-related data in its most comprehensive sustainability report to date, which includes a breakdown of its carbon sources, water use and other environmental impacts. Not surprisingly, energy use showed the biggest rate of increase in the 2025 carbon tally as Amazon and other tech companies are working to rapidly expand their data center capacity to meet AI computing demand. For the first time since 2019, the company also reported an uptick in its "carbon intensity" -- a measure of how much carbon was emitted relative to each dollar of revenue. Amazon has promoted this metric as a sign that it can decouple its growth from its climate impacts. Despite emissions moving in the wrong direction and ongoing data center-driven challenges, the Seattle-area company remains committed to its pledge of net-zero carbon emissions by 2040. When it comes to that goal, "I remain confident and optimistic in the overarching vision and the long-term progress we continue to make toward it," said Kara Hurst, Amazon's chief sustainability officer, in the foreword to the company's annual report. The report highlights areas of success that include: * Data center efficiency: Amazon's data centers are 9% more efficient than the public cloud average and 30% more efficient than on-premises data centers at directing energy toward computing rather than cooling, lighting or overhead. * Data center water use: Amazon is seven times more efficient in its water use than the industry average thanks to its use of air cooling at most sites, most of the year. * 100% clean energy overall: For the third year running, Amazon matched its company-wide electricity use with an equivalent volume of purchased clean energy, although it technically still draws on fossil fuels for some of its energy. * Electric vehicle fleet: It has the largest corporate EV fleet in North America, with more than 52,700 delivery vans worldwide. It's halfway to meeting its 2030 goal of 100,000 EVs. The company also reported improvements in reducing packaging and plastic use in delivered items; increasing use of low-carbon building materials in data center construction; and progress toward becoming water positive at its data centers, meaning it aims to replenish more water to communities than it uses. The Amazon-backed Climate Pledge -- an effort to get other organizations to commit to net-zero carbon emissions by 2040 -- has grown to 656 signatories after adding 107 companies this year. It marks a notable increase at a time when companies are growing quieter about climate commitments, with some stepping back from earlier goals. But the surge in data center investment shows little sign of slowing, which will keep complicating Amazon's path to lower emissions. CEO Andy Jassy said Amazon expects to spend a record $200 billion in capital expenditures this year, including "AI, chips, robotics, and low-Earth orbit satellites." Not all reactions to that buildout have been positive -- even within the company. Members of Amazon Employees for Climate Justice this month testified before the Seattle City Council in favor of data center requirements for renewable energy and labor protections, though Amazon doesn't operate any data centers within city limits. In the report, Amazon CSO Hurst acknowledged that AI-fueled advances could catalyze sustainability solutions or slow progress toward climate goals. "But what alternative do we have," she said, "but to continue to invest, learn, and move forward to try to solve one of the world's most challenging issues?"

Managing data centre energy and water use in the AI era

As artificial intelligence (AI) use surges, the data centre energy demand needed to support this growth is creating opportunities across many industries - from job creation to wealth building across commercial real estate construction, technology and venture funding. But with the global data centre sector set to expand at 14% (the compound annual growth rate) through 2030, it will require up to $3 trillion in infrastructure investment. The data centre boom is also causing a stir in local communities, becoming a hot political issue. Along with the disruption to neighbourhoods from construction, data centres use a tremendous amount of energy and water. Gartner analysts predict worldwide data centre electricity consumption will double by 2030, and the Environmental and Energy Study Institute reports that some large data centres require up to 5 million gallons of water each day. These demands are only expected to continue to increase as more data centres are built. But there is an opportunity to conscientiously build data centres for the future. A careful, thoroughly researched and thoughtfully executed data centre plan should enable the acceleration of AI-driven innovation without environmental consequences. Balancing the needs of data center stakeholders and concerned citizens is possible. Three effective approaches to address the current data centre energy challenges include: Liquid cooling and heat reuse replaces traditional air blowing to absorb and transfer data centre heat. Instead of venting that heat into the atmosphere, it is repurposed for other needs in the building or distributed to a broader network to lower the energy needs of nearby businesses or residences. Co-locating data centres on the same or an adjacent renewable energy site such as a wind or solar farm draws less power from the grid. Rather than taking power that would be used by the local community, the data centre has its own source. Predictive optimization technology uses AI and machine learning to anticipate and address potential data centre issues such as overheating before they have a negative impact on operations. This reduces energy loss from inefficient airflow and cable degradation. It also captures usage and heating data that shows real-time issues and spots trends over time. This data can support more efficient load balancing, maintenance scheduling and overall equipment optimization. Introducing new methodologies and technologies into fresh data centre builds or retrofitting existing buildings is only part of the solution, however. Supporting AI means rethinking the fundamental design of the data centre building because there are still gaps that squander energy through inefficient airflow, cable degradation and reliance on time-consuming manual installation and equipment checks. One reason for this is that traditional computing workloads have changed. Where CPUs (central processing units) were once the default standard for running workloads, they are fading into the background as today's AI demands higher powered, energy-intensive GPUs (graphics processing units). Accommodating the new GPU hardware often requires installing new equipment, which can take months and must be followed by a manual safety inspection process. These inspections rely on the human eye and handheld infrared cameras that scan an area to capture temperature readings. As AI generates heat loads that are denser, less stable and can shift, it is difficult to determine precisely where the most heat is being generated. This is due to the lack of visibility into data centre buildings' thermal energy performance and limits. As a result, these manual inspections cannot keep up with the need for continuous monitoring of the entire data centre for issues such as overheating. A lack of skilled data centre professionals compounds these challenges. This is one of the biggest and most overlooked risks in data centre management and construction today. The Uptime Institute reports that half of data centre owners and operators are tracking the metrics needed to assess their sustainability and meet regulatory requirements. Data centres are also now capturing thermal energy performance using thermal-based spatial intelligence. This approach can spot anomalies such as uneven heat distribution or hot spots that would otherwise go unrecognized by the human eye or would not be detected by an algorithm until the situation becomes critical. As a sensor without a camera, thermal-based spatial sensing provides data centre operators with real-time insight on energy consumption and heat output, down to the most granular levels. When this data is integrated into a dashboard depicting building performance, it helps ensure energy capacity flows only where it is needed. These insights can lead to faster responses and more strategic planning based on a more accurate depiction of data centre conditions. This also solves a precision issue that emerges when digital twins, virtual replicas of a facility's physical infrastructure, are used. The digital twin models are only as accurate as the sensor data input. Without continuous, physical sensing, the twin reflects intent versus reality, which can lead to costly surprises. Spatial intelligence insight, as part of a comprehensive data centre efficiency strategy, can reduce energy demand. Peer-reviewed research confirms that usage-based building controls can reduce energy consumption by 20-30% by aligning cooling output to actual demand. This would help to balance a data centre's needs with those of its neighbours. It would also help data centres to show any meaningful reductions in resource consumption. It is the responsibility of those that build and maintain data centres to optimize the infrastructure for energy efficiency. Only then can we have more productive conversations about planning the data centres of the future.

Five ways data centers can save water

Key technologies and design choices that reduce water and energy demands. It's no secret that the data centers that form the hidden engine of today's AI boom can consume huge amounts of water, mostly for cooling purposes, and dealing with this issue is critical for any organisation in the technology sector. Worldwide, the International Energy Agency (IEA) estimates that the data center sector consumes 560 billion liters of water per year today, and projections suggest this figure could soar over the coming years, reaching as much as 1,2 trillion liters per year by 2030, according to a recent government report on data center water use. Taking measures today can help to mitigate this. It's very possible for data centers to reduce the drain on local freshwater resources, while also curbing their energy use, and all without reducing performance. Here are five key issues which are driving water consumption in data centers today, and steps that data center operators can take to mitigate them. 1 Open-loop evaporative cooling is a problem Today, many data centers still rely on 'open-loop' evaporative cooling systems, where hot water is sprayed onto pads to be cooled. This older form of cooling architecture means that data centers consume significant volumes of water, which has to be continuously replenished as it evaporates. But more modern sealed liquid cooling systems offer a way to mitigate this source of water loss, by ensuring water is reused efficiently, while not impacting performance. Today's closed-loop systems reuse the same water over and over again, using liquid-to-air heat exchangers for cooling, rather than evaporating water. These warm-water cooling systems are sealed, so that after the system is filled in the first place, water loss is extremely low under normal operating conditions. Instead, the systems recirculate the same water inside a sealed environment, so no water is consumed at all during normal operations. Water is not discharged from the system, and it never evaporates: instead, it absorbs heat directly from high-performance components and transfers it out of the system. Closed-loop systems are designed not to be consumption-driven infrastructure but instead are recirculating, self-contained systems. This technology will be vital to reducing the water demands of data centers worldwide. 2 Data centers still attempt to cool the room In too many data centers, cooling is highly inefficient, meaning that up to a third of the data center's total energy use can be 'spent' on keeping components cool, with fans working to keep the data center's temperature low as well as cooling individual machines. Modern technologies such as direct-to-node and direct-to-chip cooling are offering a new way to curb this inefficient energy and water use. In direct-to-chip cooling, cold plates are attached to heat-generating components such as GPUS, with coolant liquid flowing through the plates to absorb heat and transfer it out to a heat exchanger. Liquid cooling techniques such as direct-to-node cooling can remove up to 98% of heat from servers, operating on a closed-loop system which means water is not wasted. By capturing heat directly at the source, at the processor level, rather than attempting to cool entire rooms full of hot air, data center operators can reduce energy demand and also place less strain on local water systems. This also produces higher-temperature outputs which are easier to reuse. 3 Generative AI is driving an increase in compute density One key reason for the increasing water use of data centers is simply that the AI boom is driving more demand for cooling. Generative AI is powered by Graphics Processing Units (GPUs), which can consume between five and 10 times more energy than Central Processing Units (CPUs). But we are also seeing an increase in compute density, as more and more of these components are packed into smaller areas, using techniques such as 3D silicon stacking. The demands of AI workloads mean that warm water closed-loop cooling is coming into its own as a low-energy solution. Even at high rack densities, warm water systems (which can use inputs at much higher temperatures) enable high performance while also curbing water use and removing evaporative cooling. To take one example, DreamWorks Animation adopted warm-water cooling systems for high performance computing (HPC) and recorded a 20% performance increase alongside reduced cooling requirements. 4 Projects to reuse data center heat are still too rare In data centers, almost all of the electricity used is ultimately converted into heat, and today that is a resource that is underused. In fact, the heat has been traditionally treated as an unwelcome byproduct to be removed and then discarded. But this can be transformed into a usable resource, which can deliver heat for residential communities, commercial buildings and even district heating. This reuse is still too rare, but pioneering projects across Europe are already showing the potential of the idea for reducing reliance on fossil fuels, cutting energy costs, and boosting resilience. Projects in Ireland and Scandinavia are already reusing heat to warm homes and businesses. Older evaporative cooling systems are wasteful, and shifting towards closed-loop systems where water is reused, heat harvested and given to the wider community are a prime example of how waste can be turned into wealth. 5 Data centers are still in the wrong locations At present, too little thought is given to the location of data centers, meaning that the facilities can be located in areas where there is already strain on local freshwater supplies. Going forward, organizations will take this factor into account in order to deliver more sustainable data centers which 'fit in' to their local environment: research found that 45% of IT professionals believe current data center design does not support sustainability goals. In the future, data centers will be put in carefully chosen locations to optimize renewable energy use, and also to ensure their water demands are not at odds with their local environment. This could include data centers built into urban areas with a goal to reusing waste heat to warm local homes - or even to offer spa-like experiences to local people. We've rated the best business cloud storage. This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Amazon's AI expansion drives emissions to record high

Why it matters: Amazon's annual environmental report, released a day after Google's report unveiled similar increases, adds to mounting evidence that AI growth is making it harder for tech giants to maintain climate ambition. Driving the news: Just like Google, Amazon's aggregate growth in data center capacity is outpacing its stated efficiency gains. * Amazon said it added more data center capacity globally than any company last year, while insisting it continues to improve efficiency. By the numbers: Amazon's greenhouse gas emissions rose more than 16% last year, a record jump fueled largely by a 34% increase in electricity and a 20% increase from manufacturing supply chains. * In a new disclosure this year, Amazon said its data centers withdrew 2.48 billion gallons of water, meaning at least some of that was returned to the community and not consumed. * It also said a metric used to measure water efficiency at data centers improved 20% since 2024, and more than 50% since 2021. Catch up quick: The AI race is intensifying scrutiny of the growing energy and water demands tied to data centers. Between the lines: Amazon's report is considerably shorter and generally offers less discussion of its environmental metrics than Google's, although the companies emphasize and disclose different measures. * Google's report spans 117 pages versus Amazon's 51. Zoom out: Amazon's environmental footprint extends well beyond data centers because of its global logistics and retail operations, which sets it apart from other tech giants in the AI race. * Amazon also said it has deployed more than 52,700 electric delivery vans, putting it just over halfway toward its goal of 100,000 by 2030. What we're watching: Microsoft's sustainability report -- due out in the coming weeks -- is the next major test revealing how much AI's environmental footprint is growing across the industry.

'A wild testament to the obscene bloat and waste of GenAI': Google's electricity consumption is exponentially increasing

As a Brit, I've been thinking a lot about the climate crisis since last month's historic heatwave. With many buildings across the UK designed to retain heat, we're simply not built for 37°C/98.6°F weather. Causes of climate change are numerous, and I know we didn't get here solely due to the advent of AI, but the steep increase in major players' energy consumption as a result definitely isn't helping. Data analyst Kentan Joshi looked at Google's latest environmental report and wrote about its recent trends in energy consumption. He said on BlueSky, "Two years ago [the company] flipped from linear to exponential growth, and their climate impact is blowing out, too. A WILD testament to the obscene bloat and waste of GenAI." Joshi dives into more detail in a blog post, elaborating, "The company's total electricity consumption jumped from 31 terawatt hours (TWh) in 2024 to 43 TWh in 2025. This is very easily the biggest increase in their electricity consumption ever." Though that's pretty damning all on its own, it's not just Google that's seen such a sharp increase in its energy consumption. Over a series of graphs, Joshi compares Google's power usage to other major players such as Microsoft, Meta, Apple, and even Netflix (it's worth noting Amazon stopped disclosing its power consumption data around 2022, so Joshi can only estimate). Though all of these companies have seen an increase in power consumption to varying degrees since at least 2023, Google's uptick remains stark. Google's power consumption even outstrips the total terawatt hours per year of a number of countries, too. Last year, the company's power demand was greater than that of Slovakia, Ecuador, Ireland, or Nigeria. NEW BLOG FOR YOU Google's energy consumption numbers in their new climate report are mind-blowing. 2 years ago they flipped from linear to exponential growth, and their climate impact is blowing out, too. A WILD testament to the obscene bloat and waste of GenAI: ketanjoshi.co/2026/07/01/g... -- @ketanjoshi.co ( @ketanjoshi.co.bsky.social) 2026-07-02T12:14:43.895Z Now, obviously, energy consumption is not synonymous with CO2 emissions, and a decent chunk of Google's power supply could be coming from renewable sources. However, Joshi's analysis suggests this may be but a drop in the bucket. Even after various adjustments accounting for "exclusions from their supply chain and renewable energy procurement claims," Google's emission trends are heading in the opposite direction from the emissions target it set out back in 2021. This is an unsurprising trend, given that Google's 2024 environmental report revealed its greenhouse gas emissions had increased by nearly 50% in the last 5 years. Joshi gets right to the point with his criticism, quoting Google's own environmental report as he writes, "If 'AI infrastructure buildout is currently accelerating faster than the grid is decarbonising', then Google shouldn't be building AI infrastructure. If they are breaching the boundaries of safe operation on a planet that can only take so much, they should stop and consider whether all of this is worth it." Joshi's entire blog post is full of lovely graphs and analysis that I definitely recommend scrolling through yourself. He's previously offered a debunking of the old 'AI energy consumption is actually very efficient, and could actually help us reduce global emissions' chestnut too. Basically, if you had the sneaking suspicion that claims about how a single AI prompt uses 'five drops' or 'one 15th of a teaspoon' were not meaningful metrics beyond how they attempt to obscure the environmental impact of the wider industry, you may be on to something. Joshi explains that "Text generation by a chatbot is comfortably the lowest energy form of [a] generative system," but obviously AI technologies encompass far more than that. Joshi has previously argued it's part of a strategy of 'selective disclosure,' writing last year, "What seems like a massive step for transparency and disclosure must be seen in the context of the absence of anything before it. Coming off a baseline of disclosing nothing, the industry can now choose to carefully set the narrative around precisely what they desire and get accolades for doing so."

Amazon reveals it used more energy than New Zealand in 2025 as its carbon footprint rises 16%

* Amazon says emissions rose 16% in 2025, purchased electricity emissions up 34% * Projects are using renewable energy matching and lower-carbon materials * Data center PUE continues to improve, but Google does it better In its 2025 Sustainability Report, Amazon revealed its absolute emissions increased 16% to 80.9 million tonnes, meaning that one single company now generates as many emissions as an entire country - New Zealand (77.8 million tonnes in 2024). The company blamed AI for creating unprecedented energy demand, causing emissions to rise significantly in 2025, but it argues long-term investments in renewable energy and data center efficiency keep it on track for its 2040 net zero goal. However, this is Amazon's largest annual emissions increase since it launched its net zero Climate Pledge, and its data center expansion plans continue to develop. Amazon's emissions continue to grow - 2040 net zero goal remains realistic Although AWS revenue grew 20% in 2025, Amazon noted the amount of CO2 per dollar of revenue actually grew 3% year-over-year - though this figure, known as carbon intensity, is still 38% lower than 2019 before aggressive AI-fuelled data center expansions. But while data centers have certainly caused energy consumption and carbon emissions to rise, the company's biggest contributor is actually its supply chain. Around three-quarters (76%) of its total emissions now come from its supply chain, up 20% year-over-year. Despite the 34% increase in emissions from purchased electricity, those emissions only account for around one-twentieth (5%) of Amazon's entire carbon footprint. Amazon also noted that 80 new renewable and carbon-free energy projects in 2025 brought its overall capacity to 42GW across 712 projects, with 61 of its construction projects last year using lower-carbon materials. But with a Power Usage Effectiveness (PUE) score of 1.14, its data centers fall short of Google's 2024 figure of 1.09. Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.

AI infrastructure emissions: AI race weakens climate pledges at Google, Amazon

Google and Amazon this week reported sharp increases in greenhouse gas emissions, driven by the frantic construction of artificial intelligence infrastructure that is pushing the tech giants further from their carbon neutrality pledges. Amazon's emissions, published on Wednesday, have risen 58 percent over the same period, and more than 16 percent last year, despite a pledge to reach carbon neutrality by 2040. San Francisco, Jul 03, 2026 - Google and Amazon this week reported sharp increases in greenhouse gas emissions, driven by the frantic construction of artificial intelligence infrastructure that is pushing the tech giants further from their carbon neutrality pledges. Google's total emissions, disclosed on Tuesday, have jumped 82 percent since 2019, and more than 18 percent just last year, even though it has committed to cutting them in half by 2030. Amazon's emissions, published on Wednesday, have risen 58 percent over the same period, and more than 16 percent last year, despite a pledge to reach carbon neutrality by 2040. Another sign that the problem is getting worse: both companies now pollute more for every dollar they generate in revenue. In other words, their emissions are rising faster than their sales -- a first since at least 2021 for Amazon. "Our AI infrastructure buildout is currently accelerating faster than the grid is decarbonizing," Kate Brandt, Google's chief sustainability officer, said in a blog post announcing the company's annual environmental report. Her counterpart at Amazon, Kara Hurst, likewise said in the ecommerce giant's own report that demand for AI products could "slow us down" when it comes to the company's environmental ambitions. In total, Google emitted 18.8 million tonnes of CO2 equivalent last year, which comes partly from its data centers and offices but primarily from its supply chain for chip and server manufacturing and the construction of new centers by its suppliers. Amazon emitted 80.85 million tonnes, from the same cloud computing activities, plus its warehouses, logistics fleet, and deliveries around the world. "One thing we can count on with companies is that they will pursue profits," Sytske Wijnsma, an assistant professor at UC Berkeley's graduate school of business, told AFP. Her research is focused on supply chain sustainability, particularly the growing demand for critical minerals. One the one hand, that creates an incentive for corporations to cut their operational costs - like energy. Companies will invest in more sustainable options if it reduces costs, Wijnsma said, which can be "a win-win" for them and the environment. The bigger issue, she continued, is their supply chains, which they don't directly control. "They need to find a way to fill that gap between the demand and the supply" of resources, like energy and critical minerals, Wijnsma said. "The gap can be filled by making your chips and data centers more efficient" or by tapping into more readily available options like fossil fuels. - A global effort - Earlier this month, a United Nations report found that data centers worldwide use so much energy that only 10 countries each consume more, making AI the 11th largest energy consumer globally. By 2030, AI data centers are expected to become the sixth largest energy user around the world. "It is time to come clean," UN Secretary General Antonio Guterres said during a speech on June 23 during London Climate Week. "If AI is to help build a better future, it must be honest about what it costs us now." Guterres launched an AI Environmental Transparency Initiative and urged every major AI company to measure and publicly disclose their environmental impact as well as commit to powering every data center with renewable energy by 2030. The issue affects the entire sector, including Meta and Microsoft which are expected to publish their own reports soon. The global AI race, which accelerated after the meteoric launch of ChatGPT in late 2022, is pushing tech giants to multiply data centers-facilities that consume large amounts of electricity, water for cooling, concrete, steel and chips. Google's electricity consumption has doubled in three years and nearly matches that of a country like Greece. Amazon's emissions linked to the construction of data centers have soared by more than 40 percent in a single year. In their voluminous annual reports, both companies tout their efforts. Google says it signed a record volume of "clean" electricity contracts last year and invests in nuclear and geothermal power. Amazon, for its part, presents itself as the world's largest buyer of renewable energy for the sixth year in a row, invests in small nuclear reactors, and claims to have more than 52,000 electric trucks.

The Real Cost of AI: Water Usage that Big Tech May Not be Reporting

Reports suggest that tech giants may be using far more water than they report and this could vary based on how the datacentres are powered Big tech companies like Amazon, Google and Microsoft are slated to spend over $1 trillion on AI infrastructure over the next couple of years. We also know that this industry hogs power and water like no other tech before it. What has fallen between the regulatory cracks is the fact that water consumption might be way beyond what we're given to understand. First indications of this dichotomy came in the sustainability reports released by Google and Amazon last week. These tech giants had pledged to zero-out carbon emissions in the coming years but with AI, the goals have stretched way beyond. Google's carbon emissions are up 25% since last year while Amazon reports a 16% spike. A report in the Wall Street Journal notes that even in the sustainability reports, water usage numbers may be less than honest. It says only Meta tallies the water used at the power stations that feed them electricity alongside the quantum it uses on-site for cooling down the datacentre machinery. In other words, these tech giants may be using far more water than they report. And this could vary based on how the datacentres are powered. Now, if that sounds scary, the WSJ also adds that overall water consumption is said to grow further. Both Google and Amazon may have to make some serious and potentially costly adjustments to their operations to get anywhere close to their own net-zero targets. Unless of course, President Trump throws these out of the window, as he has done in the past with anything remotely concerning climate change. When it comes to water usage, these companies need to care since they're aren't mandated to report the full scope of their water use - both direct and indirect. However, the WSJ article quotes a 2024 analysis by the Lawrence Berkely National Labs to suggest that datacentres historically have 12 times more indirect water consumption as they directly consume. Google reported that they consumed 10.9 billion gallons of water in 2025 compared to a year ago, representing a 34% spike. What's more, almost all of it went to datacentre cooling. But it doesn't state how much indirect water is consumed. A research paper from University VU at Amsterdam suggests it could be around three times its direct water consumption. This figure for Meta 19 billion gallons in 2024, which was a whopping 20 times higher than its direct water usage. The company waxes eloquent about becoming "water positive in 2030" via water-restoration projects, but is silent on how it plans to massively cut down on indirect water consumption. While it is true that Big Tech pays for renewable energy to match the energy they use, but experts have questioned this calculation. Last year, 16 state attorneys general in the US wrote a letter claiming that offsetting energy from fossil fuels isn't the same as replacing these source of energy. In other words, renewable-energy credits aren't the same as an offset for water consumed in a given area, they had noted. It is not as though these tech giants are callously wasting water. Be it Microsoft, Google or Amazon, all have announced measures they propose to take to reduce water consumption and replenish more water than they use. Then there are claims like the one from Amazon that said its datacentres uses water seven times as efficiently as the industry average. Sounds like a plan, but there limited data to suggest targets and achievement. Given the penchant for these tech giants to be economical with the truth about water usage, we have no way other than believing in independent research that presents a grim picture. Take the case of a 2025 report from nonprofit organisation Ceres which claimed that total direct and indirect water demands in Phoenix alone stood at 3% of total annual water use. Seems manageable? Well, Ceres adds that this number could hit 20% in the next five years, which would be what residents of Phoenix use to water all of the region's lawns and other landscaping. It also warns that neighbouring New Mexico, which supplies all of Phoenix's energy needs, the water demand could grow even more rapidly. Now it sounds scary! Google and Amazon acknowledge their higher levels of energy use over the past twelve months when AI utilisation has also gone up. What's intriguing is that they also speak the language of China around carbon intensity. Like Beijing, these tech giants too say that the benefits of AI far outweighs the impact on the environment. In this already confusing milieu, there is also the small matter of carbon pollution arising from energy purchases as well as their impact of Scope-3 emissions which maps pollution that a company directly does not control. For both Amazon and Google, this could include GPU acquisitions as well as use of smartphones, tablets and other electronic paraphernalia. In fact, Google has combined two categories for Scope-3 - capital goods and use of sold products. Given that the latter's contribution may be minute, it stands to reason that datacentres were their main driver. Their Scope-3 emission grew by 2.1 MMT, double of what they had reported back in 2019 - Google's baseline year. As for Amazon, their Scope-3 emissions came from capital goods, fuel and energy. This could include datacentres and warehouses in the first category, as their number was higher than Google's . "To meet strong customer demand, in 2025 we added more datacentre capacity globally than any other company, including more than 1.2 GW in Q4 alone," Amazon said. Small wonder then that net-zero targets have been pushed ahead. In the past, energy for offices and small datacentres had pushed up the carbon footprint that was cancelled out via purchase of renewable energy. With AI, this approach cannot work though Amazon and Google can still use renewables and batteries to power their datacentres. Not gas turbines as SpaceX recently did and got itself in some trouble. Going forward, these companies need to report the quantum of emissions from construction and outfitting of datacentres, given that steel and cement industry are major polluters. Also, they need to look at GPUs and memory chips, mostly made in Asia where electrical grids are dominated by fossil fuels. Which is where reports like those from Nvidia having solved the datacentre water issue must be taken note of. Their closed loop cooling system (which we reported) claims to have no need for fresh water once filled and the design appears to be a win-win as it zeroes out direct water use and reduces the total energy used for cooling. Even Microsoft has taken some steps in this direction claiming two years ago that their new datacentres would start using close-loop technology by 2027. For now, most datacentres use evaporative cooling systems that are energy-efficient but thirsty, the Lawrence Berkely report says while noting that retrofitting might prove to be very expensive. Josh Parker, head of sustainability at Nvidia calls it out. He is quoted by the WSJ to suggest that solving water footprint at the datacentre operational level is just the start. But the real issue comes from water consumed for power generation. The amount of energy and related water consumed by AI datacentres needs a relook in this context. But he also points out that the electricity for any other industry isn't scrutinised as much. Maybe, that's why everyone is looking to AI to fix the problem. Afterall, haven't the AI czars presented the technology as a magic wand to solve all human challenges? Maybe a bit more transparency around numbers and probably solutions might help clear the minds of doomsday predictors.