The demand for artificial intelligence is promoting unprecedented growth in the ability of data centers. According to Goldman Sachs, a 50% increase is expected by 2027 and an energy consumption that could be doubled by 2030. However, the entity warns that the investment rate could be deflated quickly If the adoption of AI does not comply With current expectations.
The sector lives a “frenzy atmosphere”, in the words of Eric Sheridan, general director of Goldman Sachs, where great technological ones invest “defensive and attack” on the one hand and the other. Today, data centers total 62 capacity gigawatts, with 58% dedicated to the cloud, 29% to traditional loads and 13% to IA. Nevertheless, In just four years, AI could occupy 28% of the totalgaining land especially at the expense of the cloud, although it would do so in an expanding global market.
The investment volume is colossal, it should be noted. The OMDIA consultant equals the GDP of medium economies, with Amazon allocating more than 100,000 million dollars a year, only one of raw capital, since the infrastructure for the multiply by one hundred the scale. If two years ago an avant -garde server integrated eight GPU, in 2027 it is expected that a single rack will reach 576, consuming 600 kilowatts -enough to supply 500 homes in the US.
The opportunity for the semiconductor industry is, therefore, it states: Counterpoint Research estimates that income could double until the billion dollars are exceeded in 2030, driven by servers for increasingly complex applications and by the rise of the “token economy” linked to smart agents.
Said which, this deployment has an environmental cost in which The electricity consumption of data centers It will represent between 1 and 2 percent of the global demand in 2023, at 3 and 4 percent in 2030. And although Goldman Sachs expects that 40% of the increase will be covered with renewable energy and something with nuclear, the remaining 60% will do so with natural gas, with an addition of up to 220 million tons of co? to the atmosphere, the equivalent of an extra 0.6 % of global emissions.
Doubts, in any case, persist. Analysts admit to being “on alert” given the possibility that IA models fail to monetize or that innovations arise that reduce their development, as has already happened. Its base scenario provides an annual growth composed of 17%in the capacity of data centers, up to 92 GW in 2027, but do not rule out adjustments down up to 14%, or up to 20%rise, according to the market.
The warning connects with other critical voices such as Sam Altman himself, CEO of Openai, recalls in The Register, who has already recognized that The industry is going through an AI bubble. Also, from the consultant McKinsey it is remembered that, in reality, nobody knows for sure how the future demand for services will be since the phenomenon is in full expansion.
