Brief IA
›

Nvidia Promises AI Data Centers in Containers by 2027

Models & LLM·Tom Levy·

Nvidia Promises AI Data Centers in Containers by 2027

Nvidia Promises AI Data Centers in Containers by 2027
Key Takeaways
1Nvidia announces AI containerized data centers ready for connection (DSX POD)
2Up to 100% less cooling water claimed, with limits depending on climate
3First shipping window targeted for early 2027, installation announced in 14 days
Why it matters : Nvidia claims that faster deployment would allow for earlier revenue generation.

Nvidia is preparing pre-assembled AI data centers in containers, designed around its Vera Rubin systems. The company claims a deployment time of 14 days instead of six months and boasts up to 100% less cooling water usage thanks to a closed-loop system, while acknowledging additional needs during heatwaves. The first shipments are targeted for early 2027.

Up to 100% Less Water, but Extra Support Needed in Extreme Heat

Nvidia states that its DSX platform can operate in a closed loop and dissipate heat into the air without evaporation, potentially allowing for up to 100% less cooling water usage. However, the company admits that certain climates will require additional cooling during extreme heat events. The first DSX POD units are announced for global shipments in early 2027.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Modules Ready to Connect On-Site After Factory Assembly

The DSX POD involves delivering fully pre-assembled AI data centers in containers. Partners like Dell, HPE, Lenovo, and Schneider Electric are expected to integrate servers, networking, cabling, power, and liquid cooling in the factory before on-site testing. Once delivered, the units would only require connections to power, cooling systems, and the network of a prepared location. The architecture is based on Nvidia's Vera Rubin systems.

Nvidia Promises 14-Day Installation and Earlier Revenue

According to Nvidia, this deployment method reduces installation time from six months to 14 days. The company indicates that faster execution of AI models would allow the operator to start generating revenue sooner. The DSX POD reference design was unveiled at GTC Berlin.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.