The Readiness of the New Bynet Data Centers Facilities to Absorb AI computing

Picture of Moti Tau

Moti Tau

Contact with us

The growing AI world in recent years has allowed various customers to conduct advanced computing research and perform advanced AI processes using advanced GPU servers, both for running different learning models and for implemented solutions in the field.

Since 2022, there has been a dramatic increase in the consumption of such services, as the entry of GPUs(NVIDIA’s H100, H200 and advanced servers from other companies) enables processing and parallel calculations at much larger rates than what was previously possible with regular CPU servers that performed serial processing.

Alongside the improvement in the performance of the cards, the consumption of these servers, both in terms of electricity and air conditioning, has increased significantly and requires adapting the infrastructure to the new needs. If until a few years ago the power was about 5 KW per Rack  averagely, the new Racks consume only tens to hundreds of kilowatts per single Rack.

This dramatic change necessitates special preparations from all facility infrastructures:

  • The ability to supply electricity of up to hundreds of kilowatts per single Rack.
  • Dealing with the nature of the equipment’s electricity consumption – rapid increases and decreases in power.
  • Integrated cooling capability using air, water, and liquid for racks according to the nature of the equipment.

As an advanced and leading technology Data centers company, Bynet Data Centers currently operates dozens of Racks that include GPU servers and various types of AI servers in the company’s active facilities, most of them in the main facility “Jerusalem of Gold.”

In planning the company’s next facilities (one facility will operate at the end of 2025 and the second at the end of 2026), many preparations have been incorporated to deal with advanced cooling technologies and the planned power consumption of the developing AI world. These preparations included full flexibility of the infrastructure down to the Rack level, allowing for dealing with Racks with future powers of 100KW and above – advanced cooling design, unique high-power distribution strips, and more. These facilities will be able to handle all existing and future cooling technologies – In-Rack/in-Row cooling, Direct Liquid Cooling (DLC) Single Phase, Dual Phase and Immersion cooling.

Additionally, in the coming months, several high-power cooling Racks for AI of up to hundreds of KW will be installed in the “Jerusalem of Gold” facility, allowing the company and its clients to experiment with such equipment even before the establishment of future halls and enabling better implementation of infrastructures in the next facilities.