The artificial intelligence boom is triggering a massive infrastructure overhaul in data centers, and the next battleground isn't chips or cooling systems,it's the invisible fiber optic cables that connect them. Goldman Sachs analysts recently projected that the market for AI-related networking hardware will expand nearly ninefold, from approximately $15 billion in 2026 to $154 billion by 2028, driven by hyperscalers migrating from traditional copper connections to high-speed optical fiber.

Why Are Data Centers Switching to Optical Fiber?

As AI models grow larger and more complex, data centers need to move enormous amounts of data between thousands of graphics processing units (GPUs) simultaneously. Copper wiring, which has been the industry standard for decades, simply cannot keep pace with the speed and volume requirements of modern AI infrastructure. Optical fiber, which transmits data as pulses of light, can handle far greater bandwidth and move information faster over longer distances without signal degradation.

The shift is happening at two critical levels within data centers. Scale-out networking connects massive AI clusters containing 100,000 or more GPUs across multiple racks and facilities. Scale-up networking links GPUs within a single computing unit. Both require the kind of high-speed, high-capacity connections that optical technology provides.

"We're going through the single largest infrastructure buildout in human history," one industry CEO told CNBC, adding that artificial intelligence "is going to become fundamental infrastructure all over the world."

CEO, quoted on CNBC, May 7, 2026

What Components Will Drive the $154 Billion Market?

Goldman Sachs identified three categories of optical networking hardware that will capture the bulk of this spending surge. The analysis breaks down where the investment will concentrate:

• Co-packaged optics (CPO): These integrated modules combine optical transceivers directly with computing chips, eliminating the need for separate connections. Goldman projects CPO alone will account for $91 billion of the total $154 billion market by 2028.
• Optical modules: Standalone transceivers that convert electrical signals to light and back again, enabling long-distance fiber connections between data center racks and facilities.
• Related components: Supporting infrastructure including fiber cables, connectors, and management systems that enable the entire optical network to function.

The scale of this transformation is staggering. Goldman Sachs calculated that the aggregate dollar content of networking hardware per computing unit will increase 29 times over the next few years. This means that for every dollar spent on networking equipment today, companies will spend $29 by 2028 as they build out AI infrastructure.

How Are Companies Preparing for the Optical Networking Boom?

Several technology companies are already positioning themselves to capitalize on this shift. MACOM Technology Solutions, a semiconductor and optical networking firm, reported strong momentum in its data center segment. In the second quarter of 2026, the company generated $98.2 million in data center revenue, up 14.5% from the previous quarter, driven primarily by surging demand for pluggable optical modules and higher cable production volumes.

"All three of our business segments contributed to the growth, with the Data Center segment leading the way as AI infrastructure spending accelerated," explained Steve Daly, President and Chief Executive Officer at MACOM Technology Solutions.

Steve Daly, President and Chief Executive Officer, MACOM Technology Solutions Holdings, Inc.

MACOM's gross margin expanded to 58.5% in the quarter, up 90 basis points from the previous period, as the company benefited from manufacturing scale, improved production efficiencies, and a growing share of higher-margin data center products. The company's adjusted earnings per share reached $1.09, up from $1.02 in the prior quarter and $0.85 a year earlier.

Another player in the space, Clearfield Inc., demonstrated the real-world impact of optical networking infrastructure. The company recently helped rebuild the British Virgin Islands' communications network after it was devastated by two Category 5 hurricanes in 2017. Using Clearfield's modular, pre-connectorized fiber technologies, the reconstruction team cut deployment times by 50% and significantly accelerated post-storm network restoration.

What Does This Mean for the Broader AI Infrastructure Race?

Industry executives are calling this moment a major supercycle, comparable to previous transformative infrastructure buildouts. One CEO told CNBC on April 23 that demand for AI and cloud services is fueling massive growth in optical networking, with forecasts showing optical and IP networks growing up to 20% year over year in 2026.

The timing is critical. As companies like Microsoft, Google, and Amazon race to build massive AI data centers to support generative AI applications, they cannot simply add more GPUs and hope the existing network infrastructure keeps up. The bottleneck has shifted from computing power to data movement. Without sufficient optical networking capacity, even the most powerful chips will sit idle, unable to exchange information fast enough to train and run large language models efficiently.

Goldman Sachs' projection of a ninefold market expansion reflects confidence that this transition is not a temporary trend but a fundamental restructuring of how data centers will operate for the next decade. As AI becomes embedded in everything from cloud services to enterprise software, the infrastructure required to support it will only grow more complex and demanding. Optical networking is no longer a luxury upgrade; it has become the essential backbone of AI infrastructure.