AI's Energy Appetite: Data Centers Struggle with Sustainability Amid AI Boom

AI supercharges data centre energy use - straining the grid and slowing sustainability efforts

At 2.9 watt-hours per ChatGPT request, AI queries require about 10 times the electricity of traditional Google queries, according to the Electric Power Research Institute, a nonprofit research firm. Emerging AI capabilities such as audio and video generation are likely to add to this energy demand. The energy needs of AI are shifting the calculus of energy companies. They're now exploring previously untenable options, such as restarting a nuclear reactor at the Three Mile Island power plant that has been dormant since the infamous disaster in 1979. Data centres have had continuous growth for decades, but the magnitude of growth in the still-young era of large language models has been exceptional. AI requires a lot more computational and data storage resources than the pre-AI rate of data centre growth could provide. AI and the grid Thanks to AI, the electrical grid - in many places already near its capacity or prone to stability challenges - is experiencing more pressure than before. There is also a substantial lag between computing growth and grid growth. Data centres take one to two years to build, while adding new power to the grid requires over four years. As a recent report from the Electric Power Research Institute lays out, just 15 states contain 80 per cent of the data centres in the US. Some states - such as Virginia, home to Data Centre Alley - astonishingly have over 25 per cent of their electricity consumed by data centres. There are similar trends of clustered data centre growth in other parts of the world. For example, Ireland has become a data centre nation. Along with the need to add more power generation to sustain this growth, nearly all countries have decarbonization goals. This means they are striving to integrate more renewable energy sources into the grid. Renewables such as wind and solar are intermittent: The wind doesn't always blow and the sun doesn't always shine. The dearth of cheap, green and scalable energy storage means the grid faces an even bigger problem matching supply with demand. Additional challenges to data centre growth include increasing use of water cooling for efficiency, which strains limited fresh water sources. As a result, some communities are pushing back against new data centre investments. Better tech There are several ways the industry is addressing this energy crisis. First, computing hardware has gotten substantially more energy efficient over the years in terms of the operations executed per watt consumed. Data centres' power use efficiency, a metric that shows the ratio of power consumed for computing versus for cooling and other infrastructure, has been reduced to 1.5 on average, and even to an impressive 1.2 in advanced facilities. New data centres have more efficient cooling by using water cooling and external cool air when it's available. Unfortunately, efficiency alone is not going to solve the sustainability problem. In fact, Jevons paradox points to how efficiency may result in an increase of energy consumption in the longer run. In addition, hardware efficiency gains have slowed down substantially, as the industry has hit the limits of chip technology scaling. To continue improving efficiency, researchers are designing specialised hardware such as accelerators, new integration technologies such as 3D chips, and new chip cooling techniques. Similarly, researchers are increasingly studying and developing data centre cooling technologies. The Electric Power Research Institute report endorses new cooling methods, such as air-assisted liquid cooling and immersion cooling. While liquid cooling has already made its way into data centres, only a few new data centres have implemented the still-in-development immersion cooling. Flexible future A new way of building AI data centres is flexible computing, where the key idea is to compute more when electricity is cheaper, more available and greener, and less when it's more expensive, scarce and polluting. Data centre operators can convert their facilities to be a flexible load on the grid. Academia and industry have provided early examples of data centre demand response, where data centres regulate their power depending on power grid needs. For example, they can schedule certain computing tasks for off-peak hours. Implementing broader and larger scale flexibility in power consumption requires innovation in hardware, software and grid-data centre coordination. Especially for AI, there is much room to develop new strategies to tune data centres' computational loads and therefore energy consumption. For example, data centres can scale back accuracy to reduce workloads when training AI models. Realising this vision requires better modelling and forecasting. Data centres can try to better understand and predict their loads and conditions. It's also important to predict the grid load and growth. The Electric Power Research Institute's load forecasting initiative involves activities to help with grid planning and operations. Comprehensive monitoring and intelligent analytics - possibly relying on AI - for both data centres and the grid are essential for accurate forecasting. On the edge The US is at a critical juncture with the explosive growth of AI. It is immensely difficult to integrate hundreds of megawatts of electricity demand into already strained grids. It might be time to rethink how the industry builds data centres. One possibility is to sustainably build more edge data centres - smaller, widely distributed facilities - to bring computing to local communities. Edge data centres can also reliably add computing power to dense, urban regions without further stressing the grid. While these smaller centres currently make up 10 per cent of data centres in the US, analysts project the market for smaller-scale edge data centres to grow by over 20 per cent in the next five years. Along with converting data centres into flexible and controllable loads, innovating in the edge data centre space may make AI's energy demands much more sustainable. (The Conversation) PY PY

AI supercharges data centre energy use - straining the grid and slowing sustainability efforts

At 2.9 watt-hours per ChatGPT request, AI queries require about 10 times the electricity of traditional Google queries, according to the Electric Power Research Institute, a nonprofit research firm. Emerging AI capabilities such as audio and video generation are likely to add to this energy demand. The energy needs of AI are shifting the calculus of energy companies. They're now exploring previously untenable options, such as restarting a nuclear reactor at the Three Mile Island power plant that has been dormant since the infamous disaster in 1979. Data centres have had continuous growth for decades, but the magnitude of growth in the still-young era of large language models has been exceptional. AI requires a lot more computational and data storage resources than the pre-AI rate of data centre growth could provide. AI and the grid Thanks to AI, the electrical grid - in many places already near its capacity or prone to stability challenges - is experiencing more pressure than before. There is also a substantial lag between computing growth and grid growth. Data centres take one to two years to build, while adding new power to the grid requires over four years. As a recent report from the Electric Power Research Institute lays out, just 15 states contain 80 per cent of the data centres in the US. Some states - such as Virginia, home to Data Centre Alley - astonishingly have over 25 per cent of their electricity consumed by data centres. There are similar trends of clustered data centre growth in other parts of the world. For example, Ireland has become a data centre nation. Along with the need to add more power generation to sustain this growth, nearly all countries have decarbonization goals. This means they are striving to integrate more renewable energy sources into the grid. Renewables such as wind and solar are intermittent: The wind doesn't always blow and the sun doesn't always shine. The dearth of cheap, green and scalable energy storage means the grid faces an even bigger problem matching supply with demand. Additional challenges to data centre growth include increasing use of water cooling for efficiency, which strains limited fresh water sources. As a result, some communities are pushing back against new data centre investments. Better tech There are several ways the industry is addressing this energy crisis. First, computing hardware has gotten substantially more energy efficient over the years in terms of the operations executed per watt consumed. Data centres' power use efficiency, a metric that shows the ratio of power consumed for computing versus for cooling and other infrastructure, has been reduced to 1.5 on average, and even to an impressive 1.2 in advanced facilities. New data centres have more efficient cooling by using water cooling and external cool air when it's available. Unfortunately, efficiency alone is not going to solve the sustainability problem. In fact, Jevons paradox points to how efficiency may result in an increase of energy consumption in the longer run. In addition, hardware efficiency gains have slowed down substantially, as the industry has hit the limits of chip technology scaling. To continue improving efficiency, researchers are designing specialised hardware such as accelerators, new integration technologies such as 3D chips, and new chip cooling techniques. Similarly, researchers are increasingly studying and developing data centre cooling technologies. The Electric Power Research Institute report endorses new cooling methods, such as air-assisted liquid cooling and immersion cooling. While liquid cooling has already made its way into data centres, only a few new data centres have implemented the still-in-development immersion cooling. Flexible future A new way of building AI data centres is flexible computing, where the key idea is to compute more when electricity is cheaper, more available and greener, and less when it's more expensive, scarce and polluting. Data centre operators can convert their facilities to be a flexible load on the grid. Academia and industry have provided early examples of data centre demand response, where data centres regulate their power depending on power grid needs. For example, they can schedule certain computing tasks for off-peak hours. Implementing broader and larger scale flexibility in power consumption requires innovation in hardware, software and grid-data centre coordination. Especially for AI, there is much room to develop new strategies to tune data centres' computational loads and therefore energy consumption. For example, data centres can scale back accuracy to reduce workloads when training AI models. Realising this vision requires better modelling and forecasting. Data centres can try to better understand and predict their loads and conditions. It's also important to predict the grid load and growth. The Electric Power Research Institute's load forecasting initiative involves activities to help with grid planning and operations. Comprehensive monitoring and intelligent analytics - possibly relying on AI - for both data centres and the grid are essential for accurate forecasting. On the edge The US is at a critical juncture with the explosive growth of AI. It is immensely difficult to integrate hundreds of megawatts of electricity demand into already strained grids. It might be time to rethink how the industry builds data centres. One possibility is to sustainably build more edge data centres - smaller, widely distributed facilities - to bring computing to local communities. Edge data centres can also reliably add computing power to dense, urban regions without further stressing the grid. While these smaller centres currently make up 10 per cent of data centres in the US, analysts project the market for smaller-scale edge data centres to grow by over 20 per cent in the next five years. Along with converting data centres into flexible and controllable loads, innovating in the edge data centre space may make AI's energy demands much more sustainable. (The Conversation) PY PY

AI supercharges data centre energy use-straining the grid and slowing sustainability efforts

Boston University Boston (US), Jul 13 (The Conversation) The artificial intelligence boom has had such a profound effect on big tech companies that their energy consumption, and with it their carbon emissions, have surged. The spectacular success of large language models such as ChatGPT has helped fuel this growth in energy demand. At 2.9 watt-hours per ChatGPT request, AI queries require about 10 times the electricity of traditional Google queries, according to the Electric Power Research Institute, a nonprofit research firm. Emerging AI capabilities such as audio and video generation are likely to add to this energy demand. The energy needs of AI are shifting the calculus of energy companies. They're now exploring previously untenable options, such as restarting a nuclear reactor at the Three Mile Island power plant that has been dormant since the infamous disaster in 1979. Data centres have had continuous growth for decades, but the magnitude of growth in the still-young era of large language models has been exceptional. AI requires a lot more computational and data storage resources than the pre-AI rate of data centre growth could provide. Also read: How Indian data centre operators are mitigating AI power consumption woes AI and the grid Thanks to AI, the electrical grid - in many places already near its capacity or prone to stability challenges - is experiencing more pressure than before. There is also a substantial lag between computing growth and grid growth. Data centres take one to two years to build, while adding new power to the grid requires over four years. As a recent report from the Electric Power Research Institute lays out, just 15 states contain 80 per cent of the data centres in the US. Some states - such as Virginia, home to Data Centre Alley - astonishingly have over 25 per cent of their electricity consumed by data centres. There are similar trends of clustered data centre growth in other parts of the world. For example, Ireland has become a data centre nation. Along with the need to add more power generation to sustain this growth, nearly all countries have decarbonization goals. This means they are striving to integrate more renewable energy sources into the grid. Renewables such as wind and solar are intermittent: The wind doesn't always blow and the sun doesn't always shine. The dearth of cheap, green and scalable energy storage means the grid faces an even bigger problem matching supply with demand. Additional challenges to data centre growth include increasing use of water cooling for efficiency, which strains limited fresh water sources. As a result, some communities are pushing back against new data centre investments. Better tech There are several ways the industry is addressing this energy crisis. First, computing hardware has gotten substantially more energy efficient over the years in terms of the operations executed per watt consumed. Data centres' power use efficiency, a metric that shows the ratio of power consumed for computing versus for cooling and other infrastructure, has been reduced to 1.5 on average, and even to an impressive 1.2 in advanced facilities. New data centres have more efficient cooling by using water cooling and external cool air when it's available. Unfortunately, efficiency alone is not going to solve the sustainability problem. In fact, Jevons paradox points to how efficiency may result in an increase of energy consumption in the longer run. In addition, hardware efficiency gains have slowed down substantially, as the industry has hit the limits of chip technology scaling. To continue improving efficiency, researchers are designing specialised hardware such as accelerators, new integration technologies such as 3D chips, and new chip cooling techniques. Similarly, researchers are increasingly studying and developing data centre cooling technologies. The Electric Power Research Institute report endorses new cooling methods, such as air-assisted liquid cooling and immersion cooling. While liquid cooling has already made its way into data centres, only a few new data centres have implemented the still-in-development immersion cooling. Flexible future A new way of building AI data centres is flexible computing, where the key idea is to compute more when electricity is cheaper, more available and greener, and less when it's more expensive, scarce and polluting. Data centre operators can convert their facilities to be a flexible load on the grid. Academia and industry have provided early examples of data centre demand response, where data centres regulate their power depending on power grid needs. For example, they can schedule certain computing tasks for off-peak hours. Implementing broader and larger scale flexibility in power consumption requires innovation in hardware, software and grid-data centre coordination. Especially for AI, there is much room to develop new strategies to tune data centres' computational loads and therefore energy consumption. For example, data centres can scale back accuracy to reduce workloads when training AI models. Realising this vision requires better modelling and forecasting. Data centres can try to better understand and predict their loads and conditions. It's also important to predict the grid load and growth. The Electric Power Research Institute's load forecasting initiative involves activities to help with grid planning and operations. Comprehensive monitoring and intelligent analytics - possibly relying on AI - for both data centres and the grid are essential for accurate forecasting. On the edge The US is at a critical juncture with the explosive growth of AI. It is immensely difficult to integrate hundreds of megawatts of electricity demand into already strained grids. It might be time to rethink how the industry builds data centres. One possibility is to sustainably build more edge data centres - smaller, widely distributed facilities - to bring computing to local communities. Edge data centres can also reliably add computing power to dense, urban regions without further stressing the grid. While these smaller centres currently make up 10 per cent of data centres in the US, analysts project the market for smaller-scale edge data centres to grow by over 20 per cent in the next five years. Along with converting data centres into flexible and controllable loads, innovating in the edge data centre space may make AI's energy demands much more sustainable.

AI is making Google and Microsoft big contributors to climate change

The generative artificial intelligence boom is boosting revenue for major tech companies -- but the massive demand for energy is making these companies big-time contributors to climate change. Earlier this month, Google said its carbon emissions have risen by 48% since 2019, mostly due to energy consumption by data centers and supply chain emissions. The company's carbon emissions were up 13% year-over-year in 2023, according to its 2024 Environmental Report. In 2021, Google had set a goal of net-zero emissions across its operations and value chain by the end of the decade. However, Google said in the report that starting in 2023, it was "no longer maintaining operational carbon neutrality," and would instead focus on other "carbon solutions and partnerships" to fulfill the net-zero goal. Meanwhile Microsoft had set a similar goal in 2020 to be "carbon negative" by the end of the decade. In May, however, Microsoft said in its 2024 Environmental Sustainability Report its carbon emissions were almost 31% higher than in 2020, mostly due to building more data centers for AI workloads, and hardware such as semiconductors and servers. "Our challenges are in part unique to our position as a leading cloud supplier that is expanding its datacenters," Microsoft said in a statement. "But, even more, we reflect the challenges the world must overcome to develop and use greener concrete, steel, fuels, and chips." In April, Ami Badani, chief marketing officer of British chip designer Arm, said data centers powering AI chatbots such as OpenAI's ChatGPT account for 2% of global electricity consumption, and that demand could eventually slow down AI progress. Despite Google's massive carbon footprint, a Goldman Sachs study found a query on ChatGPT needs almost 10 times as much electricity as a Google search. By 2030, data centers could consume up to 9% of electricity in the U.S. -- more than double what is being used now, according to the Electric Power Research Institute. However, a third of nuclear power plants based in the U.S. are reportedly discussing deals with tech companies to be electricity suppliers. And in April, OpenAI chief executive Sam Altman was among investors in Exowatt, a startup developing modules that store energy as heat and produce electricity for AI data centers.

Data centres: Balancing AI needs with sustainability

IBM summarised the conundrum: "Organisations have long grappled with energy consumption challenges that accompany traditional data centres, which can lead to increased operational costs and a notable environmental impact." The rise of generative artificial intelligence (AI) experiences added to the demand for more computing power and data processing. Khalid Wani, senior director for sales in India at Western Digital, told HT, "AI is now essential in data centre operations through machine learning and predictive analytics. It's for real-time infrastructure monitoring, proactive issue identification, resource optimisation, and equipment failure prediction." Statista Market Insights research estimated the global generative AI market would be worth $207 billion by 2030, up from $44.89 billion in 2023. If AI has worsened an energy and emissions problem, it may also be helping solve it in a proverbial double-edged sword. In particular, the AI factor could help in data centre management, which requires sensors and cooling systems. Wani added that a reduction in errors means more efficiency because of "reduced downtime and enhanced security via threat detection." Data centres could be cooled using various methods, including perimeter cooling, close-coupled cooling, and newer direct liquid cooling (DLC) methods. According to the Uptime Institute's 2023 Global Data Center Survey, 56% of 1,575 surveyed data centres globally still relied on perimeter cooling, but they believed the continued increase in silicon power would necessitate a switch to DLC methods. Newer generation AI-optimised chips played a crucial role. Google confirmed in its latest sustainability report that Trillium, its sixth-generation Tensor Processing Unit (TPU), was 67% more energy-efficient than its predecessor. Kate Brandt, chief sustainability officer at Google, said, "A Google-owned and operated data centre is, on average, approximately 1.8 times as energy efficient as a typical enterprise data centre. In 2023, the average annual power usage effectiveness for our data centres was 1.10." Power usage effectiveness (PUE) was a metric to define the efficiency of a data centre. The Uptime Institute's survey pegged the global average for data centres at 1.58 PUE, an improvement from 1.67 in 2019 and 1.98 in 2011. Noelle Walsh, corporate vice president, Cloud Operations and Innovation at Microsoft, said in 2022, "Our newest generation of data centres have a design PUE of 1.12 and, with each new generation, we strive to become even more efficient." Microsoft's Azure Maia 100 and Cobalt 100 chips, Amazon's second-generation Trainium, Meta's MTIA, and Google's TPUs illustrated an industry-wide momentum to build AI chips as the foundation of data centres, cloud services and AI training. Ramanujam Komanduri, country manager at Pure Storage India, said, "AI not only means new and more powerful chips but a very large amount of data being ingested, processed and stored." For India, there is an opportunity with momentum towards digitisation and an increasingly connected population accessing the internet and apps. Estimates pegged data centre-specific consumption at around 2% of total electricity use. He pointed out, "Meta's 100-petabyte AI Research Super Cluster deal with Pure Storage noted an 80% lower total cost of ownership due to our efficient storage infrastructure. This efficiency means fewer hardware components, less power consumption, and floor space required." Captain Ishver Dholakiya, founder and managing director of Goldi Solar, an Indian tech company that makes renewable power solutions for data centres, said, "The digital expansion has made data centres integral to modern life. With the advent of AI, data centres' power consumption is projected to increase from 853 MW in 2023 to 1,645 MW by 2026." Research firm CareEdge Ratings estimated that India's data centre industry, then in a 'growth phase', would double its capacity to 2,000 MW by 2026. Commercial real estate and investment firm CBRE, in a progress report in late 2023, suggested India's data centre demand was led by fintech, healthcare, education, social media, as well as, content delivery. "The data centre segment's accelerated growth is likely to continue over 2023-24, with nearly 500 MW currently under construction," said the report. Microsoft, Yotta Data Services, Atlassian, Equinix (working with Oracle Cloud), Amazon Web Services, and Google had outlined data centre plans for India.

AI makes Google and Microsoft big contributors to climate change - ExBulletin

The boom in generative artificial intelligence is boosting revenues for big tech companies, but its massive energy demands make these companies a major contributor to climate change. Spotify introduces comment feature to podcasts Google said earlier this month that its carbon emissions have increased 48% since 2019, mostly due to data center energy consumption and supply chain emissions. The company's 2024 Environmental Report shows that its carbon emissions in 2023 will increase 13% from the previous year. In 2021, Google had set a goal of achieving net-zero emissions across its operations and value chain by the end of the decade. But in the report, Google said it would no longer remain operationally carbon neutral after 2023 and would instead focus on other carbon solutions and partnerships to reach its net-zero goal. Meanwhile, Microsoft set a similar goal in 2020 to become carbon negative by the end of the decade. But in May, Microsoft announced in its 2024 Environmental Sustainability Report that its carbon emissions had increased by about 31% over 2020, mainly due to increased construction of data centers and hardware such as semiconductors and servers for AI workloads. "The challenges we face are unique to our position as a leading cloud supplier with an expanding data center footprint, but more importantly they reflect the challenges the world must overcome to develop and use greener concrete, steel, fuels and chips," Microsoft said in a statement. In April, Ami Badani, chief marketing officer at British chip designer ARM, said that data centers running AI chatbots such as OpenAI's ChatGPT account for 2% of global electricity consumption, a demand that could ultimately slow AI progress. Despite Google's massive carbon footprint, a Goldman Sachs study found that a ChatGPT query requires nearly 10 times more electricity than a Google search. According to the Electric Power Research Institute, data centers could consume up to 9% of U.S. electricity by 2030, more than double today. But a third of U.S. nuclear plants have reportedly signed contracts with tech companies to become their power suppliers. And in April, OpenAI CEO Sam Altman became an investor in Exowatt, a startup developing modules to store energy as heat and generate electricity for AI data centers. What Are The Main Benefits Of Comparing Car Insurance Quotes Online

Google and Amazon's climate pledges are being derailed by AI- Fast Company

Hundreds of companies have vowed to reduce their carbon emissions in recent years. But with AI on the rise, carbon-neutral goals are only getting more challenging to achieve. In 2019, before AI made its way to the forefront, Amazon cofounded The Climate Pledge, which committed more than 500 companies to net-zero goals by 2040. The following year, another 90 companies joined The Better Climate Challenge, which proposed lowering emissions by 50% by 2030. While the goals are headed in the right direction, climate emissions are not, especially now that many of those same companies are focused on AI efforts. Google, for instance, was once widely viewed as one of the leaders of the clean energy movement, but now its emissions are increasing rather than shrinking. The company revealed in a new report that it had vowed to go net-zero but that its emissions grew 13% in 2023. Compared to 2019, emissions soared even more -- by a staggering 48%. "As we further integrate AI into our products, reducing emissions may be challenging," the report reads. The company says AI is the culprit, as it puts a greater demand on data centers that require a ton of electricity to keep up and running. The company is no longer claiming the "net-zero" label and says it plans to achieve this now by 2030.

Big Tech has a problem thanks to energy-gobbling AI. These stocks may help solve it

The tech sector faces a moment of truth in its ambitious climate targets, as the growing power needs of artificial intelligence data centers jeopardize the industry's promise to slash carbon emissions. Alphabet 's Google unit reported last week that its greenhouse gas emissions rose 13% in 2023 compared to the prior year, "primarily due to increases in data center energy consumption and supply chain emissions." Google admitted in its most recent environmental report that "reducing emissions may be challenging due to increasing energy demands from the greater intensity of AI compute, and the emissions associated with the expected increases in our technical infrastructure investment." And Microsoft disclosed in May that its indirect emissions rose by 31% compared to 2020, primarily due to "the construction of more datacenters and the associated embodied carbon in building materials." Goals at risk The energy needs of AI computing pose a significant challenge to Google's and Microsoft's ambitious climate goals. Google aims to achieve net-zero emissions by 2030 through around-the-clock carbon-free energy on every electric grid where it operates. Microsoft wants to become carbon negative in the same year, matching 100% of its electricity consumption with carbon-free energy and investing in carbon removal technologies. Whether this remains possible is unclear. Google and Microsoft, for example, have not pushed back their climate goals. But at the same time, forging ahead with energy-intensive generative AI is a commercial imperative for Microsoft, Keith Weiss, Morgan Stanley's U.S. software analyst, told clients in a Thursday note. Morgan Stanley is recommending a suite of stocks -- AES Corporation , Bloom Energy , Legrand , Schneider Electric , Holcim , and Sika -- that can help the tech sector move away from relying on fossil fuels while also building out its data centers. The investment bank rates all but one the equivalent of a buy; Schneider is rated neutral. AES Corp. has large portfolio of renewable energy assets, is a leader in the battery technology that's crucial for dispatching and storing wind and solar power and is a major supplier to tech companies. Bloom Energy makes makes fuel cells that can run on natural gas or hydrogen. The cells are installed onsite at facilities including data centers and can operate independently of the power grid. Holcim and Skia offer construction solutions that improve energy efficiency and promise to reduce the carbon footprint of buildings. Schneider Electric and Legrand help make data center servers more efficient through power management and temperature solutions. HCMLY YTD mountain Holcim ADRs in 2024.