Tech giants are rushing to prevent a carbon time bomb caused by huge data centers they build all over the world.
The technology, pioneered by Google, is becoming increasingly popular. power-hungry artificial intelligence goes online: using software to find clean electricity in parts of the world with abundant solar and wind on the grid, and then ramping up data center operations there. This can reduce carbon emissions and costs.
There is a pressing need to figure out how to operate data centers in a way that maximizes the use of renewable energy, said Chris Noble, co-founder and chief executive officer of Cirrus Nexus, a cloud computing manager that uses data centers owned by Google, Microsoft and Amazon. .
Climate risks driven by artificial intelligence computing are far-reaching—and will get worse without a significant shift from fossil fuel-derived electricity to clean energy. CEO of Nvidia Corp. Jensen Huang said that artificial intelligence the “turning point” has arrived. He also said that cost of data centers will double within five years, which will encourage the emergence of new software.
According to International Energy Agency. Together they are responsible for annual emissions of about the same amount of carbon dioxide as Brazil.
Hyperscalers—as major data center owners such as Google, Microsoft and Amazon are known—have set climate goals and face internal and external pressure to meet them. These lofty goals include decarbonizing their operations.
But the development of artificial intelligence is already damaging these goals. GPUs have played a key role in the emergence of large language models and consume more power than central processing units used in other forms of computing. According to the data, training an AI model requires more energy than 100 households per year. IEA estimates.
“The growth of artificial intelligence is far outpacing the ability to produce clean energy for it,” he said.
Moreover, AI power consumption is inconsistent and looks more like a sawtooth graph than the smooth line most data center operators are accustomed to. This makes decarbonization challenging, let alone ensuring grid stability.
According to Dave Sterlace, director of global data center operations at Hitachi Energy, AI growth is being driven by North American companies that are concentrating computing power and energy consumption there. It’s a trend he didn’t expect two years ago.
To reduce data center CO2 emissions, hyperscalers and other data center providers have financed huge numbers of solar or wind farms and used the credits to offset the emissions. (As for loans, some of them didn’t have significant impact on emissions.)
But this alone will not be enough, especially given the rise in the use of AI. That’s why operators are turning to the strategy used by the Alphabet Inc. unit. Google, called “load redistribution”. The idea: reduce emissions by changing the way data centers operate.
Today, most data centers strive to operate in a “steady state,” meaning their power consumption is fairly stable. This leaves them at the mercy of the grid they are connected to and regardless of the daily mix of natural gas, nuclear and renewable energy generation, given the lack of transmission lines between regions. To break their dependence on dirtier grids, tech giants are looking to shift daily or even hourly data center operations around the world to absorb excess renewable energy production.
Google has made an early effort to match energy use in some data centers with zero-carbon energy on an hourly basis, aiming to make its machines run on clean energy 24 hours a day, 7 days a week. No one has yet fully achieved this goal. And, of course, the strategy of moving loads around the world can be complicated by countries pushing for data sovereignty policies that attempt to limit and protect the flow of data across borders. But what Cirrus Nexus and Google are testing could still be an important piece of the emissions reduction puzzle.
Manhattan-based Cirrus Nexus surveys the world’s energy grids and measures emissions at five-minute intervals to find the least polluting computing resources for itself and its clients in industries ranging from pharmaceuticals to accounting. Last summer, the company had the opportunity to put this search into practice.
The Netherlands had its sunniest June on record, causing the cost of solar power on the grid to fall. This has made running servers cheaper and less carbon-intensive. Cirrus Nexus then moved its computing load to California as soon as the sun set in the Netherlands, allowing it to take advantage of the solar power that had just come online for the day in the Golden State.
By chasing the sun from Europe to the US West Coast and back, the company was able to reduce computing emissions for certain workloads for itself and customers by 34%, rather than relying solely on servers in any location, according to company data provided by Bloomberg Green. Ensuring operational flexibility comes with both benefits and risks.
The ability to source spare megawatts with zero carbon emissions could help reduce stress on the grid during a heat wave or cold winter storm, for example. But data centers must work with utilities and grid operators because large fluctuations in demand can disrupt electrical systems, increasing the likelihood of blackouts. Dominion Energy, which seeing a sharp rise in demand for data centers Its Virginia utility is working on a load-shifting program in its data centers to reduce grid load during extreme weather events.
In recent years, Google and Amazon have tested changing the use of data centers for their own operations and for customers using their cloud services. (Cirrus Nexus, for example, uses cloud services offered by Amazon, Microsoft and Google.) In Virginia, Microsoft signed an agreement with Constellation Energy Corp. that guarantees that more than 90% of the energy for its regional data center will be produced with zero carbon emissions. energy. However, achieving 100% remains a challenge for him and other hyperscalers.
Google’s data centers run on carbon-free energy about 64% of the time, with 13 regional sites achieving 85% and seven achieving just over 90% globally, said Michael Terrell, who leads Google’s 24-hour carbon-free energy strategy , 7 days a week. .
“But if you’re not displacing fossil assets, you’re not fully achieving your climate goals,” Terrell said.