Economist Predicts Energy Deadlock in the West Due to AI

The current model of artificial intelligence development in the West, where many companies are developing similar neural networks, threatens to lead to an energy deadlock, economist Konstantin Tserazov told RIA Novosti.

"The entire cumbersome technical architecture of AI agents' operation involves millions of tokens — units of queries, information processing, and response generation. Under the current AI development model in the West, with its emphasis on mindless and mass adoption of AI agents, an energy deadlock will arrive very soon," the expert believes.

He noted that the formal abundance of AI players in the United States has a downside: they are solving the same problems in parallel — for example, training similar models on the same datasets — without coordinating their efforts.

Currently, due to such uncoordinated behavior among market participants in the US, processors in AI infrastructure are operating significantly below full capacity. During the neural network training phase, they run at maximum load, consuming large amounts of electricity, but during further operation, downtime reaches 65%, and in some cases — 95%, Tserazov noted.

"A universal 'know-it-all' is economically unprofitable — what is needed are industry-specific neural networks for concrete tasks, which would reduce training time and eliminate repeated processing of queries," he added, emphasizing that companies should work together to achieve an effective project, but the US has not been able to do this.

He believes that Russia has a better chance of transitioning to such a model of AI development, given the level of coordination on various economic projects demonstrated, for example, by the BRICS+ association.

"And that means — and I am confident of this — within BRICS+, it is possible to reach an agreement on the distribution of efforts in AI development. And Russia has something to offer in this regard: energy, competencies, and advanced developments for the effective development of AI," the economist concluded.