February 8, 2024
Nvidia and Cisco Systems want to simplify the process of creating in-house AI computing infrastructure with a new joint service offering end-to-end artificial intelligence solutions that aim to allow any enterprise firm to host its own AI data center. Along with its own networking gear, Cisco will globally broker Nvidia AI software and GPU cloud products along with jointly configured “purpose-built Ethernet networking-based solutions.” European cloud services provider ClusterPower is an early customer, using the new offering “to help drive data center operations with innovative AI/ML solutions.”
“For Nvidia, the partnership opens up another channel for its technology. Cisco, meanwhile, gets to capitalize on an AI spending boom that’s helped turn Nvidia into the world’s most valuable chipmaker,” Bloomberg reports, explaining how the new offering aims to bring the massive speed, security and massive cloud computing power of large data centers closer to home.
“It’s part of an effort to spread the deployment of AI hardware beyond its current concentration in data centers owned by” Microsoft, Amazon, Google and Meta Platforms, Bloomberg writes, explaining how Nvidia is hedging its bet with the joint venture. While the aforementioned “hyperscalers” currently use Nvidia chips, “they’re all either working on their own chips or have already begun adopting them in some way,” exposing “a potentially risky concentration of revenue for Nvidia,” according to Bloomberg.
Nvidia details its role in a press release issued from the Cisco Live event in Amsterdam this week. Cisco’s M7 UCS rack and blade servers will use Nvidia’s latest Tensor Core GPUs, to optimize performance across “AI and data-intensive workloads in the data center and at the edge,” Nvidia says.
Nvidia’s AI Enterprise software — with more than 50 frameworks, pretrained models as well as development tools — will be available through Cisco. TechRadar says the partnership will enable “a scalable powerhouse for the extreme workloads generated by AI and data heavy operations.”
Under the agreement, Cisco is making available “both cloud-based and on-premises AI infrastructure and software to handle the intense workloads of generative AI,” TechRadar writes. Customers can “manage their operations and data through the Cisco Networking Cloud which provides simplified management of the AI infrastructure from anywhere.”