-
If the Google search engine used AI in all its answers, its electricity consumption would increase from 106.8 to 1,068 MWh
-
It is already generating unprecedented pressure on the power grid.
Google’s AI Overviews feature complements search results with personalized explanations from a language model. If this artificial intelligence were activated worldwide, it would consume the electricity needed to charge seven electric cars per second.
10 times more energy than a normal search. The answers generated by a large language model consume 10 times more energy than a conventional Google search, Alphabet Chairman John Hennessy told Reuters in an interview.
If a standard Google query uses 0.3 watt-hours, the artificial intelligence in the search engine’s new “AI Overviews” feature increases the consumption to 3 watt-hours, roughly the same amount of electricity used by an LED light bulb when turned on for an hour.
Seven electric cars per second. To put this into perspective, Google receives an average of 356 million queries every hour.
If AI Overviews ends up being activated for everyone and for all queries (for now it is only available in the United States and for certain searches), the search engine’s electricity consumption would increase from 106.8 to 1,068 MWh.
As Gizmodo says, that’s enough to charge the batteries of seven electric cars per second, assuming an average battery size of 40 kWh.
Much greedier data centers. Energy usage in data centres has remained relatively stable in recent years, but consumption could increase by 200% by 2030 due to the large-scale deployment of AI, research warns.
In the event of a full rollout of the search engine, Google’s energy expenditure would exceed that of entire countries the size of Ireland.
The energy pressure of AI. The high-performance chips used to train and run AI models consume large amounts of electricity, which is already putting unprecedented pressure on the power grid.
With data center consumption expected to continue to increase by a factor of 10 every six months, Elon Musk warned at a conference that there are not enough transformers to meet demand.
Nuclear as a response. Meeting the energy demands of AI without falling behind on the goal of cutting CO2 emissions requires clean and highly efficient energy sources.
Meta or Microsoft’s bet is on nuclear energy. Microsoft has invested in both small modular nuclear reactors (SMRs) and nuclear fusion projects that it hopes to launch by 2028.
Image | Google
At Xataka | The power grid is suffering, and not because of electric cars: because of the enormous demand for artificial intelligence
