Tech

NVIDIA Teams Up With Equinix to Bring Generative AI Training to Data Centers


Data center services company Equinix is collaborating with NVIDIA to provide private AI with NVIDIA DGX on a fully managed AI data center solution, both companies announced on Wednesday, Jan. 24.

Equinix Private AI with NVIDIA DGX is available Jan. 24 in the U.S. and Frankfurt, Germany. Equinix said it will expand to 250 data centers spread across North America, South America, Europe Asia and Africa “soon.”

Equinix and NVIDIA announce Private AI with NVIDIA DGX solution

The Equinix Private AI with NVIDIA DGX solution is based on NVIDIA DGX SuperPOD, a supercomputing architecture for creating, training and running generative AI models. Then, it is combined with the NVIDIA AI Enterprise software stack. Equinix provides data center management for Equinix Private AI with NVIDIA DGX.

The new solution will be coordinated through the Equinix Managed Service team and can be scaled up from existing Equinix multi-cloud offerings.

“What we’re seeing is that once customers start out with AI, they continue to scale,” Charlie Boyle, vice president of DGX Systems NVIDIA, told TechRepublic in an email. “With this offering, customers can start out at any size and grow with their success.”

The new solution aims to help organizations spin up AI infrastructure faster

Cost, speed of deployment and sustainability are among the concerns enterprises often list regarding generative AI deployment that NVIDIA and Equinix want to solve. The Equinix Private AI with NVIDIA DGX solution is a “turnkey offering for companies to get the AI infrastructure they want in a very predictable, short lead time manner,” said Boyle, during a pre-briefing on Jan. 23.

Any data center infrastructure handling proprietary data fed into generative AI needs to be scalable, with low throughput and low latency, Boyle said. In essence, customers want to “move the AI closer to the data,” he said.

Data security is a barrier to generative AI adoption

Generative AI models have the potential to transform how organizations do business, but data security is a major barrier to entry; enterprises do not want large language model makers to have access to their proprietary data.

This solution is of particular interest to businesses because of its position in the wave of custom and proprietary generative AI models for enterprise. Generative AI could be used in healthcare and finance, but both are fields in which privacy is incredibly important. Financial service organizations that are already customers of both Equinix and NVIDIA have been exploring generative AI technology for customer service and fraud detection; in particular, retrieval-augmented generative AI is in demand.

Governance, security and attestation are at the top of customer needs, said Jon Lin, executive vice president and general manager, data center services at Equinix.

In addition, enterprises want their data to be up-to-date and not constrained by a model trained on publicly available information from a year ago. Equinix Private AI with NVIDIA DGX was made to address those problems and to help organizations – especially those that already use Equinix or NVIDIA infrastructure – speed up and smooth out the process of creating custom generative AI tools trained on proprietary data.

SEE: Microsoft is doubling its AI data center footprint in the U.K. (TechRepublic)

Enterprises “want to own their own AI future,” said Boyle. “And what largely that translates to is they want to own the models that their business is dependent on. Whether that is they want to own the model for data security, IP security, auditability, all of those aspects.”

Competitors to Equinix Private AI with NVIDIA DGX

Other large companies offering generative AI deployment for enterprises using private data include cloud services such as Google’s Vertex AI Workbench and Google Cloud AI Infrastructure, IBM’s watsonx, Microsoft’s Azure AI and Amazon’s SageMaker, Lambda (which uses NVIDIA DGX) and generative AI deployment training platforms provided by organizations such as Hugging Face, MosaicML and Databricks.



Source Credit

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button