In July, Google came under fire when its annual environmental report revealed that its Emissions increased by 48% Thanks to the expansion of our data centers to support AI growth over four years. It also said its goal of reaching net zero emissions across all its operations and value chain by 2030 is now “very ambitious” and “will require (Google) to navigate significant uncertainties.”
For “Sustainable GenAI Development,” the Capgemini Research Institute surveyed executives from 2,000 large organizations worldwide that were already working with GenAI. Almost half (47%) said their organization’s greenhouse gas emissions had increased by an average of 6% in the past year, and a similar proportion (48%) linked their use of AI to growth.
Generative AI demands a lot of energy and water.
The environmental impact of GenAI is aggressive. The technology processes central graphics processing units require rare earth metals that must be mined, releasing greenhouse gases. The hardware behind it also requires frequent upgrades, which studies show can lead to Five million tonnes of e-waste by 2030.
Data centers are expected to be responsible. up to 4 percent of global electricity demand By 2030, driven, at least in part, by AI. OpenAI training GPT-4With 1.76 trillion parameters, a lot of energy was consumed. Equivalent to the annual electricity consumption of five thousand American households. This does not include the power required for inference, where the AI generates output based on new data.
Cooling the servers also requires a lot of water. Assessing 10 to 50 questions on a large language model About 500 ml of water.
See: Sending an email with ChatGPT is like using a bottle of water.
The European Union has a lofty target to reduce the region’s greenhouse gas emissions by 2030. At least 11.7 percent less than expected at the beginning of the decade. However, bit barns are in demand in Europe. Triple Prophecy In the meantime, increasing their share of the region’s total energy demand by 3 percent and putting that goal out of reach.
Businesses don’t know, or care, about the emissions associated with their use of AI
Many businesses now use AI. 80% have increased their investment in it since 2023.According to Capgemini. Nearly a quarter are now integrating generative AI into some or most of their locations or functions, up from 6% in 2023.
See: 31% of organizations using generative AI ask it to write code.
However, the new report highlights that awareness of AI’s electricity and water requirements is poor. Only 38% of executives surveyed claim to be aware of the environmental impact of the GenAI they use, and 12% say their company measures its footprint.
Of those surveyed who are aware of the impact, 51% say the use of AI is one of the main reasons their organization is increasing emissions. They also expect it to increase their emissions ratio by 2.2 percent over the next two years.
The lack of businesses monitoring the environmental impact of their use of GenAI is not for lack of effort. Nearly three-quarters (74%) of those surveyed said this is difficult to do because of limited transparency from hyperscalers and model providers.
Oh Uptime Institute report It found that less than half of data center owners and operators track metrics like renewable energy consumption and water use. The emissions from data centers owned by Google, Microsoft, Meta and Apple are estimated to be around 662% higher than officially reported. The Guardian. This is largely due to renewable energy certificates and carbon offset schemes, which allow companies to claim they use renewable energy when they don’t.
See: Power shortages hamper data center growth in UK, Europe
On the other hand, executives may not be concerned about the impact of using AI on their company’s emissions. Only a fifth of Capgemini survey respondents ranked environmental impact among the top five factors when choosing or building GenAI models.
Cost competitiveness was ranked among the top five considerations by 53% of executives. However, it is primarily tied to energy use, according to Samuel Young, AI practice manager at research firm Energy Systems Catapult.
He said: “When applied at scale, organizations quickly become sensitive to estimated costs. So they have an incentive to adopt low-energy models, which can reduce carbon footprints. ”