Jensen Huang: Container-based data centers cut time and costs

[Financial News] NVIDIA, an artificial intelligence (AI) semiconductor company, has unveiled a modular data center solution that can dramatically shorten the time needed to build AI data centers and bring forward the point at which customers begin generating revenue.
According to a Yahoo Finance report on the 21st (local time), NVIDIA CEO Jensen Huang unveiled DSX POD, a reference design for a modular AI data center the size of a standard shipping container, at the GPU Technology Conference (GTC) in Berlin, Germany.
The solution is focused on minimizing the time it takes AI systems to process tokens and generate revenue, thereby reducing the enormous cost burden caused by delays in data center construction.
Conventional brick-and-concrete data centers can take anywhere from several months to several years to reach commercial use. By contrast, DSX POD, powered by NVIDIA’s next-generation Rubin chipset, can be connected to power, cooling systems and network infrastructure and put to use as soon as it arrives on site. This could dramatically reduce the data center deployment period, which previously took at least six months, to around 14 days.
Huang said, “It takes months to get a massive AI factory (data center) up and running,” and stressed, “If we can reduce the time from system shipment to actual operation from months to weeks, that will be a complete game changer for the entire AI infrastructure market.”
NVIDIA said DSX POD will not completely replace existing large data centers, but will serve as a “stepping stone,” allowing AI applications to be run and generate revenue before permanent data center facilities are completed.
DSX POD supports liquid cooling (water cooling) and also incorporates a dry-cooler design, giving it the environmental benefit of saving up to 54 million liters of water per day compared with conventional standard data centers.
Experts believe that faster data center deployment will ease bottlenecks in AI infrastructure expansion, benefiting customers and further accelerating NVIDIA’s AI chip sales. However, even if data center construction speeds up, power supply issues—including the construction of power plants—will remain a challenge for the AI industry.

[email protected] Yoon Jae-joon Reporter