A Q&A With Corning's Michael Crook on Data Center Growth
Data center growth is exploding. Is the telecom industry ready?
I asked Michael Crook, Market Development Manager at Corning Optical Communications some questions about what’s on the horizon for data center networks and what role Corning is playing in the revolution.
What network challenges are unique to data centers and what challenges are ahead for them?
Data centers are confronted with unique challenges due to the rise of generative AI. One of these challenges is time. The speed at which data centers need to be built or expanded is increasing rapidly. Hyperscale data centers are swiftly adapting their architectures to accommodate more fiber for AI applications. Now, more than ever, they require pre-connectorized solutions for denser, faster, and simpler deployments.
Additionally, space within data centers and the size of components are also significant challenges. Compared to traditional data centers that primarily use a Central Processing Unit (CPU) based network, AI requires a second backend network built on Graphical Processing Units (GPUs) which allows an application like a chatbot to learn from a set of data and then create new data based on what it learned.
These GPUs not only connect to the front-end network but also interconnect, creating the dense neural network required for Gen AI. In fact, these large GPU clusters can require over 10 times more optical fiber in the same space than legacy server racks, leading to a need for smaller and denser solutions to fit more fiber into existing spaces.
Data centers also face the challenge of future-proofing their networks and accounting for scalable growth. As AI continues to evolve, the backend networks will also need to adapt and grow. We see fiber connections scaling up as data rate per Graphical Processing Unit scales up—meaning more and more fiber optic connects needed to build larger AI clusters.
In the multi-tenant data center space, we are noticing more hyperscalers leasing space in multi-tenant data center operations to enable AI operators faster.
What, if anything, has changed in the data center space in the last five years?
Generative AI has changed the game for data centers. We have seen hyperscale operators deploy a significant amount of fiber into their networks to facilitate current and future demands of AI. In the multi-tenant data center space, we are noticing more hyperscalers leasing space in multi-tenant data center operations to enable AI operators faster. Additionally, we are seeing a need for future-forward fiber innovations to allow networks to quickly and efficiently evolve as processing rates increase.
What role will Corning play in the current and future AI-driven growth of data centers in the U.S.?
As the original inventor and manufacturer of the first low-loss optical fiber and a constant forward-thinking innovator, Corning Optical Communications is uniquely positioned to create the future-forward fiber solutions needed to keep networks agile within U.S. and global operations. Thanks to our trust-based relationships with the world’s largest hyperscalers and our commitment to research and development, we can anticipate unforeseen challenges and innovate proactively to address them.
Could you tell me about any innovations from Corning in recent years to meet data center requirements, and how Corning designs for the scalability demands of data centers?
Corning has focused on the vectors of size, speed, and simplicity to create our next generation of fiber and cable solutions to help operators manage scalable network growth. We're seeing strong adoption of these solutions, which are helping hyperscale operators get these GPU clusters up and running faster and with less labor, while also increasing the density of the network.
Size: We have reinvented our fiber to be 40% smaller than legacy fibers, allowing for tighter bends and greater flexibility in deployments. This smaller fiber paves the way for a revolution in cable density -- doubling the fiber count in the same cable diameter. For our customers, that reduces the need for additional infrastructure buildouts.
Speed and simplicity: We’ve put smaller fibers, smaller cables, and smaller connectors together into solutions customized for each Gen AI data center at large scale. Not only do these unique solutions help to solve the density problem, but they also offer seamless plug-and-play integration. They move time-consuming work from the installation site to the factory by means of pre-terminated fiber links -- an area in which Corning has been an innovation leader for decades.
How do Corning products contribute to energy-efficient data centers?
We invented a distribution system with a single high-fiber-count cable pre-connectorized at the end of the row, with custom connectorized assemblies dropping off at each rack location. Compared to legacy solutions, our solution is 70% faster to install with fewer skilled people and carries 55% less embodied carbon.
For more information on what Corning is up to in the data center space, visit Corning.com.
Joe Gillard | Executive Editor
Joe Gillard is a media professional with over 10 years of experience writing, editing, and managing the editorial process across a spectrum of innovative industries. Joe strives to deliver the best possible editorial product by focusing on the needs of the audience, utilizing the data available, and collaborating with a talented team.