Russian authorities are working to address a shortage of graphics accelerators needed to operate artificial intelligence systems. According to reports, such accelerators are not produced in Russia and the government is preparing state support measures, including tax breaks for the purchase of such equipment Forbes. This equipment has not yet been clearly distinguished from mining equipment.
Image credit: Igor Omilaev / unsplash.com
A government subcommittee working group is addressing the shortage and high cost of graphics processors. By the end of September, specific standards will be determined as obstacles to the localization of artificial intelligence equipment, and it is necessary to formulate a basis for revising these standards. standards will also be developed “Allows you to clearly distinguish artificial intelligence equipment (GPU, NPU, high-speed network, etc.) from mining and other equipment”. This is necessary to simplify “Management of Support Measures” and eliminate their patterns of abuse.
One of the measures is to prepare proposals before the end of the year “About tax incentives and other support mechanisms for purchasing artificial intelligence equipment to reduce the cost of owning artificial intelligence equipment”. By the end of this year, standards will be developed that will differentiate AI data centers from other data centers. Experts interviewed by Forbes recalled that the basis of artificial intelligence infrastructure consisted of servers with graphics processors. The device can be used to both train artificial intelligence models and launch already trained models. You can separate AI equipment from mining equipment based on the GPU category and its computing power: it is not financially profitable to buy an Nvidia H200 or B300 accelerator for mining.
Formally, there is no difference between mining and artificial intelligence equipment according to the classifier – it is classified according to the HS code 8471 50 000 0. Therefore, it makes sense to tie the benefit not to the fact that the device was purchased, but to the confirmed intended use. It should be emphasized that GPUs are not produced in Russia, so it is recommended to solve the problem step by step: first localize assembly, and then increase localization over the next ten years – too strict requirements may slow down the market development.
If you find an error, select it with your mouse and press CTRL+ENTER.