This week in Washington, D.C., NVIDIA kicked off the Data Center World 2025 function with a strong vision for the future of AI system.
The concept of AI-scale information centers was introduced in his presentation by Wade Vinson, the main data center engineer at NVIDIA. These massive, energy-efficient facilities would help meet the growing demand for accelerated computing. NVIDIA envisions expansive” Artificial factories” powered by DGX SuperPODs and Blackwell GPUs, supported by Vertiv and Schneider Electric’s advanced heating and energy systems.
There is no denying that the information center industry is adopting AI businesses.
completing step one of a Texas AI stock
Vinson pointed to Crusoe Energy Systems ‘ Lancium Clean Campus, which is located close to Abilene, Texas. He continued,” As he said:
- The 200 MW in two buildings, second phase of this AI manufacturer, is generally finished.
- It may grow to 1.2 GW after the second phase. By the middle of 2026, it should be finished.
- Rear-door heat exchanger, air chilling, and direct-to-chip wet chilling are included in the design.
- Six more properties will make up the facility’s four million square feet, bringing it to four million.
- To give on-site energy, 10 gas turbine will be deployed.
Also, each tower may run up to 50, 000 NVIDIA GB200 NVL72s GPUs on a single integrated community material, expanding the range of data center scale and design for AI training and inference workloads.
Vinson claimed that some AI factories may use on-site strength, while others will profit from locations where there is already power on site. He referred to outdated mill, factories, and retail establishments that are already connected to the grid.
An old store in San Francisco can be converted into an AI factory in decades, for instance, rather than the lengthy steps needed to complete new construction and acquire value interconnects and enables. Big buildings are frequently present at these locations that can house solar power plants.
turning existing data centres into AI businesses.
How about the current statistics locations? Aging constructions may struggle to provide AI and NVIDIA equipment. Vinson thinks that many colocation facilities ( colos ) are well positioned to become AI factories.
Any colony constructed in the last ten years has enough energy and cooling to turn into an AI factory, he claimed. ” AI businesses should be viewed as a profit opportunity rather than an expense.”
He thinks that AI could increase productivity in business and individuals by 10 % or more, contributing$ 100 trillion to the global economy.
It represents a larger output change than it did as a result of the flood of global electrification that started about 100 years ago, Vinson said.
Planning is essential to the success of AI factories.
Vinson urged those interested in creating or operating their own Artificial companies to be aware of the value of planning. Modeling is crucial, and it’s important to take into account the different factors involved.
He praised the Omniverse modeling tool from NVIDIA as a method for effectively planning an AI factory. It uses electronic mini systems to provide accurate modelling of data center equipment and design marketing. Lacking the ability to model and approximate numerous possible scenarios can cause inefficiencies in areas like energy consumption and may shorten construction times.
Through integrated energy management, simulations enable data centers to increase operational performance, according to Vinson.
SEE: With Just About 30 Lines of Code, Data Centers You Reduce Energy Use By Up To 30 %
For instance, some veteran data middle managers may find it challenging to switch from the traditional concepts of racks, aisles, and servers to GPU equipment that is cooled by liquid and has the appropriate power and power supply equipment.
Because AI shop designs will have much more energy and cooling equipment inside than server racks, layouts may be radically different. After all, GPU-powered SuperPODs generate more heat than normal information centers, which is higher.
Expect significant rack merger, according to Vinson. Eight of the ancient racks might well turn into one more advanced plate with GPUs in. Because the racks inside AI factories will be significantly different from what the majority of data centers are used to, it is crucial to create a streamlined energy and cooling construction.
