Nvidia teams up with Google Cloud to launch AI-driven hardware samples

In partnership with Google, Nvidia at the moment launched its L4 platform, a brand new cloud {hardware} product it says is optimized to run productive AI functions like OpenAI’s DALL-E 2.

Obtainable in non-public preview on Google Cloud by way of Google’s G2 digital machines, Nvidia says its L4 platform is designed to speed up “AI-powered” video efficiency. Serving as a normal function GPU, the L4 gives transcoding and video streaming capabilities in addition to video decoding.

Past offering entry to the L4 platform by way of Google Cloud, Google is integrating L4 with Vertex AI, its managed machine studying service for enterprise clients.

The L4 might be out there this 12 months from Nvidia’s networking {hardware} companions comparable to Asus, Cisco, Dell, Hewlett Packard Enterprise and Lenovo.

L4 websites alongside different AI-focused {hardware} options Nvidia introduced at the moment, together with the L40, H100 NVL, and Grace Hopper for Suggestion Fashions. The L40 is optimized for graphics and AI-enabled 2D, video and 3D picture manufacturing, whereas the H100 NVL helps the deployment of main language fashions comparable to ChatGPT. (Because the identify suggests, Grace Hopper for Suggestion Fashions focuses on the advice mannequin.)

The L40 is offered via Nvidia’s {hardware} companions talked about above. Nvidia expects the Grace Hopper Superchip and H100 NVL GPU to launch within the second half of the 12 months, in the meantime.

In associated information, at the moment is the launch of Nvidia’s DGX Cloud platform, which supplies corporations entry to infrastructure and software program to coach fashions for productive and different types of AI. Introduced earlier this 12 months, DGX Cloud permits companies to hire Nvidia {hardware} clusters month-to-month, beginning at $36,999 per occasion monthly.

Every occasion of DGX Cloud incorporates eight NVIDIA H100 or A100 80GB Tensor Core GPUs for a complete of 640GB GPU reminiscence per node paired with storage. Clients even have entry to AI Enterprise, Nvidia’s software program layer that features AI frameworks, pre-trained fashions, and “accelerated” information science libraries.

Nvidia says it has partnered with “main” cloud service suppliers to host DGX Cloud infrastructure, beginning with Oracle Cloud Infrastructure. Microsoft Azure is predicted to start out internet hosting DGX Cloud subsequent fiscal quarter, and the service will quickly broaden to Google Cloud.

Nvidia’s aggressive push for AI infrastructure comes as the corporate strikes away from unprofitable investments in different areas comparable to gaming {and professional} virtualization. Nvidia’s newest earnings report confirmed that its information middle enterprise, which incorporates chips for AI, continues to develop ($3.62 billion), advised that Nvidia might proceed to capitalize on the productive AI growth.