It’s time to upgrade AI infrastructure
If 2023 was the year in which artificial intelligence (AI) transformed software and digital services, this year will be the one with an increasing focus on hardware and infrastructure.
At the Cisco Live conference in Amsterdam this week, the global networking equipment leader unveiled new capabilities in security and automation that would help make AI part of the fabric of most business operations.
The event attracted 14,000 IT professionals, representing an industry that is impatient to see generative AI move from a consumer fascination to a boardroom priority.
Several keynote addresses suggested that moment had arrived.
“There is no AI without a network,” Jonathan Davidson, executive vicepresident and GM of Cisco Networking, said during his address. “It’s like driving a Formula One race car: you are in control of something revolutionary, as long as the car is not controlling you, and you need a pit crew that can make split-second decisions.
“For AI, think of us as watching the road ahead for you. We give you insights to avoid threats and anything that is going to slow you down. To do this, we are simplifying and securing networking everywhere for everyone at every scale. The platform is powered by AI to simplify, secure and scale your operations across the entire network infrastructure.”
A new Cisco AI readiness index, which surveyed more than 8,000 private sector, business and IT leaders across 30 countries, found that while 95% of respondents had an AI strategy in place or under development, only 14% were ready to integrate AI into their businesses.
The key reason was not a lack of willingness to use AI, but a lack of infrastructure. That was a word that cropped up almost as frequently as AI in discussions at the event.
In response to the need, Cisco has unveiled new technologies to help businesses develop and optimise infrastructure to support AI. Among those, it has announced an AI and machine-learning blueprint for data centre networks, and new hardware products for environments in which customers need connectivity and AI at the edge — the endpoints furthest from data centres. Cisco has also extended a long-running partnership with Nvidia, which in the past year has been propelled into the elite of tech companies with a market value of more than $1-trillion (about R19-trillion), thanks to its cutting-edge graphics processing units (GPUs) — computer chips that have underpinned the generative AI revolution.
Cisco and Nvidia this week announced plans to deliver AI infrastructure solutions for data centres that are easy to deploy and manage, enabling the huge computing power that enterprises will need as they deploy their own AI solutions.
Davidson told Business Times that Cisco has been working with the chipmaker to make sure it has the full Nvidia portfolio at its disposal. He also believes Cisco is central to Nvidia’s continued success.
“We want to make it really easy for our customers to take the Nvidia GPUs that are in our servers, as well as storage, networking infrastructure, and the software on top of that, and wrap that all together in a single package,” he said.
“We have a Cisco-validated design for that. Nvidia is doing great, their numbers are great, but they don’t have 16,000 or 18,000 resellers like Cisco has, so I think we have a big opportunity to be able to help them go into the broad-based enterprise market.”
Jeetu Patel, executive vice-president and GM of security and collaboration at Cisco, told Business Times that providing effective infrastructure for AI-specific workloads and data centre configurations is a natural evolution for his organisation.
“We power a lot of the world’s data centres; we power a lot of the world’s enterprises with a networking fabric and security capability and observability capability. These are going to be unique requirements for AI,” Patel said.
“One beauty about what’s happening with generative AI is that in the past, humans needed to learn the language of machines. And now, machines are learning the language of humans. That changes all assumptions we’ve made about what is in the bounds of possibility for getting to have 8-billion people effectively use computing, and use it efficiently and more naturally.”