November 13, 2024
Artificial intelligence adoption rates are booming.
From 2023 to 2024 alone, the use of this technology by businesses jumped from 55% to 72%, and AI is expected to see an annual market growth rate of 36.6% through 2030.
We are living in what’s known as the “AI spring” and surveys show massive early adoption by consumers. Looking at two popular applications as examples, already the vast majority of drivers use satellite navigation when driving in major U.S. cities and digital voice assistants outnumber the entire human population globally.
That’s today. Moving forward, the promise of AI technologies has leading figures like OpenAI’s Sam Altman forecasting that “in the next couple of decades, we will be able to do things that would have seemed like magic to our grandparents” and looking ahead to “astounding triumphs – fixing the climate, establishing a space colony, and the discovery of all of physics.”
The oncoming applications sound like magic indeed until you consider that middle-aged people can clearly remember when pagers were the standard for on-the-go reachability and streaming didn’t exist. Forget real-time satellite navigation; road trips as recent as the early 2000s often featured hard copies of turn-by-turn directions printed from the internet.
We can count on technology to increasingly enable transformative efficiencies and life-changing experiences, and it’s a safe bet that personal and institutional adoption of AI will continue to boom. At the same time, this growth counts on advances in subfields like machine learning and deep learning – technologies dependent on massive amounts of data and powerful hardware.
The promise of AI and the potential for its growth assume that data center infrastructure will be sufficient to support its many demands.
AI Demands Extreme Capacity
AI and cloud computing are surging data center density requirements. In the near term alone, the average server rack density is expected to double in two years: from 60kW/rack in 2023 to 120kW/rack in 2025.
Longer term? “The reality is, we just don’t know how quickly these (applications) will scale, and we want to have the capacity in place in case they scale very quickly,” Meta’s Mark Zuckerberg said, adding “it's actually quite hard to forecast.”
In an early-2024 survey by Lightwave, 95% of respondents reported that AI is impacting the rate of capacity planning for their data centers. Almost two-thirds of the operators are accelerating their capacity expansion by 2x to 5x and more than a third’s expansion plans are in the 10x to 20x range.
AI Demands Infrastructure Change
Artificial intelligence and machine learning call for parallel 
processing between graphics processing unit (GPU) servers and the formation of data center clusters, which are groups of GPU servers working together and in parallel, setting them up to efficiently perform AI learning and inference workloads. This infrastructure evolution is a significant change to data center architecture that requires an overhaul of cabling and connectivity standards.
With data center clusters, all GPU servers are connected to a switch (row or room) and every server is additionally connected to switch fabric, storage, and out-of-band (OOB) management. This node-to-node connectivity requires twice as many connections as operators typically deploy with traditional data center 
architecture.
Operators are now faced with the extraordinary challenge of meeting the hard-to-forecast-but-extreme increases in density and connectivity needed between servers within ultra-high-density (UHD) environments and AI clusters, often confined by an existing facility footprint, the need to use fewer servers per rack to lessen power and heat requirements, “rapid turn-up” deployment timeframes, and limited labor workforce with the necessary skills and experience.
And zooming out on the AI boom, we know that these increased demands and larger workloads will translate to a host of performance challenges, longer processing times, and emerging security threats.
AI Demands Innovative Connectivity Solutions
To address the most common UHD and AI network infrastructure challenges, Sumitomo Electric Lightwave has developed a truly innovative and comprehensive end-to-end connectivity solution that optimizes data center capacity, time to install, loss budgets, power efficiency, network security, staffing, and sustainability.
Our data center solutions stand out in the rapidly evolving digital landscape for a user-centric focus on network scalability, modularity, and reliability.
Contact us to learn how a holistic combination of standard and customized products can help optimize your data center for the AI boom.