Ukrainian drones hit Russia’s top data centre in attack on AI ambitions

DIFFICULTIES PURCHASING CHIPS

Both companies have invested heavily in the development of their models but lag global leaders in the United States and China by up to nine months, according to Sberbank’s own estimates, partly due to Western sanctions which hinder Russia’s access to the most advanced hardware.

“It’s clear what difficulties we face when purchasing modern graphics processing units. Both we and Yandex suffer from this,” Gref said on Sep 29, calling it the biggest reason why Russia lags behind the US and China in AI.

Sberbank is in talks to buy chips from Chinese makers, but in an interview with Reuters on Jun 3, Vedyakhin said that US chipmaker Nvidia’s NVDA.O monopoly on its CUDA software platform, which has become an AI industry standard, is an even greater problem.

Yandex said previously that its supercomputers operated on Nvidia A100 chips, a model launched in 2020 and since then succeeded by more advanced H100 chips.

Sberbank has two functioning supercomputers. The most advanced one, called Christofari Neo, ranked 167th in the Top 500 list of the world’s most powerful supercomputers, is located in Moscow and is based on Nvidia A100 chips.

Yandex’s most advanced Chervonenkis supercomputer is ranked 101st in the same list.

IT expert Mikhail Klimarev told Reuters it could take a long time to replace the chips due to Western sanctions, while damage to the supercomputers could set back the Russian AI industry, which already suffered from sanctions.

“The equipment is being imported, but it’s done at a high cost, through indirect routes. First, you need to negotiate so that someone is willing to sell it to you and is not afraid of sanctions,” he said.