AI Driving Fiber-Rich Networks, Data Centers

The substantial and continued growth of artificial intelligence (AI) is lighting up broadband connectivity both within and outside of the data centers that house them. Facilities supporting large-language-model AI applications will require up to five times more connectivity compared to today’s hyperscaler architectures, according to Corning, with demand driving more connections and higher-density fiber connectivity within the data center.

“AI is quickly driving the need for more optical connectivity,” said Michael O’Day, Chief Technology Officer, Corning Optical Communications. “In doing so, it creates the need for new applications across many vertical markets and industries that consumers and businesses will use. And those applications will require more broadband access and connectivity.”

AI is already having a significant impact in many sectors, including healthcare, education, manufacturing, finance, and transportation. “Warren Buffet claims to use AI and that’s how he is making his investment choices now,” said O’Day. “AI holds the promise of helping doctors and clinicians in their life saving work to help detect abnormalities in medical imaging as an example and helping oncologists make more sophisticated data-driven cancer treatment regimens.”

O’Day noted the rapid growth of ChatGPT and its associated applications since launching in 2020. After ChatGPT’s release in November 2022, it amassed 100 million users in January 2023 then jumped to one billion users by February 2023. Such rapid growth in the AI industry is causing a strain on data centers and the AI ecosystem to keep up, due in part to AI’s heavy demands for power and data connectivity that requires the construction of a dedicated facility and network to handle it.

“Our analysis projects significant growth in the capacity deployment [of hyperscaler facilities] between now and 2027,” said O’Day. “By 2027, you’ll see the number of megawatts deployed for AI will be on par roughly with those required to support traditional cloud computing applications, which continue to grow nicely as well.”

Corning expects AI resources as measured in data center megawatt usage to grow at a compound annual growth rate (CAGR) of 42% over the next five years, to support high-density GPU servers that make up the heart of AI application building and usage today, while traditional cloud computing will only grow 13% over the same time period.

To learn more about AI’s hunger for power and fiber, listen to the latest Fiber for Breakfast.