AI summary
Summary of article:
Huida released the container-type modular AI data center DSX POD in Berlin, aiming to reduce deployment time from several months to 14 days, accelerating the launch of AI factory and generating revenue as soon as possible.
Artificial Intelligence (AI) chip giantHuida (Nvidia, also translated as NVIDIA) The annual technology event “GPU Technology Conference” (GPU Technology Conference) is being held in Berlin, Germany. A container-type AI data center called DSX POD was unveiled at the exhibition. founder of HUIDAJen-Hsun Huang (Jensen Huang) said on the 21st that if the deployment of data centers can be reduced from months to weeks, it will be a breakthrough that will completely change the rules of the game for all AI factories.
Yahoo Finance reported that it could take months or even years for the data center to be built to come online, and will not generate revenue until it begins processing AI tokens. Therefore, the longer it takes to go online, the higher the cumulative cost. Huida hopes to reduce the time customers wait for the AI system to generate revenue. A container-sized modular data center is expected to reduce factory construction time to a few weeks.
In an exclusive interview with technology anchor Deirdre Bosa, Jen-Hsun Huang said that the AI factory is very large and takes several months. “From the shipment of the system to the completion of deployment and start of operation, if we can save a few months and shorten it to a few weeks to go online, it will be a game-changing breakthrough for all AI factories.”
Powered by “Vera Rubin” next-generation AI chip computing platform, the DSX POD can be plugged in and used immediately after being delivered to the site, and can be connected to cooling, networks and other infrastructure. The entire deployment process has been reduced from half a year to just 14 days.
Huida said the DSX POD is not intended to replace traditional data centers, but rather can be used as a way to run AI applications before physical facilities are ready.
Jen-Hsun Huang said in the statement: “Through Nvidia DSX POD, we bring the precision and speed of factory manufacturing to the creation of AI infrastructure, reducing deployment time from months to weeks, allowing customers to more rapidly expand production capacity, and allowing revenue-generating production lines to start up sooner.”
Huida said that DSX supports POD liquid cooling technology (liquid cooling), but due to the dry cooler design, it can save 54 million liters of water per day compared to ordinary standard data centers.
Huida said it is currently collaborating with Anduril, Armada, Bull, Crusoe, Dell Technologies, Foxconn and other partners, and expects to begin delivery of DSX PODs in early 2027.
Summary FAQ
-
DSX POD primarily seeks to solve the problem of traditional data center construction being too slow, significantly shortening the deployment process that originally took months or even a few weeks, allowing customers to run AI applications faster and generate revenue.
-
Because data center AI cannot process words and cannot begin to recover costs until it is online, the longer the delay in deployment, the higher the cumulative expense; If it can be activated several months in advance, it can enter the revenue phase even earlier.
-
The DSX POD supports liquid cooling technology and saves water through a dry cooler; Huida has also cooperated with Anduril, Armada, Bull, Crusoe, Dell Technologies, Hon Hai, etc., and is expected to deliver in early 2027.
