47% of enterprises weaken their sustainability targets due to productive artificial intelligence.
According to a new report from Capgemini, businesses measure sustainability commitments rather than missing the benefits of productive artificial intelligence rather than missing. 47% of those who implement technology in most or all functions had to “have to” re -use “in their original environmental targets.
In July, Google came under fire after the annual environment report increased by 48% in four years thanks to the expansion of data centers to support AI developments. He also stated that by 2030, the goal of achieving all operations and net zero emissions in the value chain will require a “extremely ambitious” and “(Google) to a significant uncertainty”.
The Capgemini Research Institute searched for executives working with GENAI from 2,000 major organizations worldwide to “develop sustainable Genei”. Almost half (47%) said that greenhouse gas emissions of the organizations increased by 6%last year and a similar rate (48%) attributed an increase with AI usage.
Productive AI requires a significant amount of energy and water
Genai has an aggressive environmental effect. The graphic processing units at the center of the operation of the technology require rare soil metals that need to be removed by releasing greenhouse gases. The hardware behind it also requires frequent upgrades and studies suggesting that it can create Five million tons of e-Atık up to 2030.
It is estimated that data centers will be responsible. Up to 4% of global power demand By 2030, at least partially directed by AI. Openai’s GPT-4, education with a 1.76 trillion parameter consumed some energy Equivalent to the annual use of five thousand US households. This does not even include the electricity required for the inference that AI produces outputs based on new data.
A significant amount of water is required to cool the servers. 10 to 50 queries in a large language model About 500 ml of water.
See: sending an e -mail with chatgpt is the equivalent of consuming a bottle of water
The EU has a supreme target to reduce the region’s 2030 greenhouse gas emissions. At least 11.7% lower At the beginning of a decade. However, the demand for lice stables in Europe It was estimated as a triple At that time, to increase the region’s share in total energy demand by 3% and this goal cannot be achieved.
Businesses may not know or even care about emissions due to AI usage
Many businesses are now using AI, 80 % have increased their investments since 2023According to Capgemini. Approximately one quarter integrates the productive AI from 6% in 2023 to most or most of its location or functions.
See: 31% of the organizations using the producer AI want to write code
However, the new report emphasizes that the awareness of AI’s electricity and water demands is irregular. Only 38% of the surveyers claim that they are aware of the environmental impact of Genei they use, and 12% says they measure the footprint of their companies.
51% of the respondents who are aware of the impact are one of the main reasons for the increase in the organization of AI use. In addition, they expect to increase the rate of emissions from internal operations by 2.2%in the next two years.
The lack of enterprises following the environmental impact of Genei usage stems from the lack of effort. Almost three (74%) of the respondents, said that they were challenging due to limited transparency of hyper -scale and model providers.
A Report from the Working Duration Institute Data center owners and operators have found that less than half of the operators followed metrics such as renewable energy consumption and water use. The emissions of Google, Microsoft, Meta and Apple’s data centers are likely to be more than 662% higher than officially reported. Guardian. This is largely due to renewable energy certificates and carbon offset schemes, which allows companies to claim that they use renewable energy when they do not.
See: Power famine stop data center growth in England, Europe
On the other hand, managers may not worry about the impact of use of AI on their emissions. Only one -fifth of the participants in the Capgemini survey listed the environmental footprint between the top five factors when choosing or creating Genei models.
Cost competitiveness was among the top five issues by 53% of the managers. However, according to Samuel Young, AI application manager of the research company Energy Systems Catapult, this is basically linked to energy use.
“When implemented on a scale, organizations become rapidly sensitive to the costs of inference. Therefore, there are less energy intensive models that can reduce carbon effect.”
