Nvidia unveils containerized data center that cuts deployment time to 14 days

Nvidia CEO Jensen Huang speaks at the annual GTC developer conference in Berlin, Germany, on Wednesday (local time). [AP]
Nvidia CEO Jensen Huang speaks at the annual GTC developer conference in Berlin, Germany, on Wednesday (local time). [AP]

Nvidia CEO Jensen Huang called on European industries to accelerate their adoption of AI and expressed an optimistic outlook on AI infrastructure investment on Wednesday (local time), Bloomberg reported.

Speaking at the annual GTC developer conference in Berlin, Huang said that for Europe to benefit from AI and build a genuine AI ecosystem and economy, "every end industry must adopt AI." Citing the automotive, financial and healthcare sectors as examples, he stressed that "all of these industries need to become AI companies."

Huang said Nvidia had invested about $10 billion in Europe over the past year and that the company has built more than 12,000 partnerships across the continent since its founding.

The European tour came as investor skepticism over the massive sums being poured into AI infrastructure has grown and AI and semiconductor share prices have taken a hit over the past two weeks. The Philadelphia Semiconductor Index recorded its longest consecutive losing streak since 2008.

Huang's speech was briefly interrupted by Greenpeace protesters, who questioned him about reports that Nvidia chips had been used in Russian drones before being escorted out by security personnel.

Huang said building a 1 GW "AI factory" — a large-scale data center — costs about $50 billion. He argued that if such a facility is expected to generate $40 billion in annual sales, a three-month delay in deployment would mean $10 billion in lost revenue. Nvidia also unveiled the DSX Pod, a compact data center that comes pre-installed and pre-tested with servers, cooling systems and other equipment inside a shipping container.

Nvidia said on-site deployment time drops from six months to 14 days with the new system. It is suited for smaller-scale deployments of less than 1 GW.

Nvidia also announced plans to expand its supply chain. To ease bottlenecks in AI chip supply, the company is working with six substrate, packaging and testing partners — including Samsung Electro-Mechanics — to scale up production capacity. Ibiden, Samsung Electro-Mechanics and Unimicron will help secure substrates, while Amkor Technology will quadruple its chip packaging supply by the end of 2028.

Nvidia plans to double its semiconductor testing capacity in 2027 compared with 2026, with expansions taking place in the United States, Taiwan and Singapore.


yckim6452@heraldcorp.com