A new collaboration between Kopin and Fabric.AI aims to fundamentally reshape how data moves inside AI data centers by replacing traditional copper wiring with light-based optical interconnects, potentially solving one of the industry's most pressing challenges: excessive power consumption in GPU communication. The companies have jointly developed Neural I/o, a MicroLED-based optical interconnect technology designed to move data between graphics processors, circuit boards, and equipment racks using photons instead of electrons, with Fabric.AI placing a $15 million initial development order.

Why Are Data Centers Struggling With Power Consumption?

Today's artificial intelligence infrastructure relies on dense clusters of graphics processing units (GPUs) that must communicate with each other constantly. These GPUs currently use copper wiring to exchange data, a process that consumes enormous amounts of energy to maintain high-bandwidth connections and cool the resulting heat. As AI workloads scale globally, this approach is hitting a wall. According to a 2024 Department of Energy report, data centers are expected to consume between 6.7 and 12 percent of all U.S. electricity by 2028. The problem is particularly acute because traditional copper interconnects generate significant heat while transferring data, creating a vicious cycle where more power is needed both for communication and for cooling systems.

The challenge extends beyond just raw power consumption. High-density GPU clusters create rapid, megawatt-scale load swings that challenge traditional data center power designs and utility interconnections. As one analyst noted, the two biggest obstacles facing large-scale AI deployments are power and bandwidth, and finding ways to enable chip-to-chip connectivity without taxing the power budget has been a persistent problem.

How Does MicroLED Optical Interconnect Technology Work?

Neural I/o repurposes programmable MicroLED pixels as ultra-high-speed optical transceivers capable of moving data at extremely fast rates while consuming significantly less power per bit than existing copper-based solutions. Each MicroLED pixel functions as a high-speed transmitter, sending digital bits at rapid speeds and enabling real-time GPU-to-GPU data exchange at massive scale. By using photons instead of electrons, the technology eliminates the need for copper interconnects and expensive laser-based systems entirely.

Kopin brings deep expertise in MicroLED materials, process development, and manufacturing to the partnership, while Fabric.AI contributes system-level design and focus on AI factory infrastructure. The collaboration positions Kopin as the exclusive manufacturer of Neural I/o chipsets, and the company owns 19.9 percent of Fabric.AI.

Steps to Understanding Data Center Power Optimization

• Identify the bottleneck: Traditional copper wiring between GPUs consumes enormous energy for both data transfer and cooling, making it the primary target for efficiency improvements in modern AI infrastructure.
• Evaluate alternative technologies: Optical interconnects using light-based transmission offer significantly lower power consumption per bit compared to electron-based copper systems, with the added benefit of eliminating expensive laser infrastructure.
• Consider manufacturing and supply chain: Kopin's position as a U.S.-based MicroLED producer provides partners with a secure, reliable, and scalable supply chain, increasingly critical as MicroLEDs become foundational to next-generation AI infrastructure systems.

What Makes This Technology Significant for the AI Industry?

The Neural I/o technology addresses a critical gap in current data center architecture. As AI workloads continue to scale, the industry needs faster, more efficient optical interfaces capable of supporting GPU-to-GPU communication at massive scale. Kopin's CEO Michael Murray stated that the marriage of MicroLED technology with the company's bi-directional NeuralDisplay architecture is exactly what the industry needs to break through current interconnect bottlenecks.

"The ability to enable chip-to-chip and system-to-system connectivity in a way that enables the full throughput of the accelerator without taxing the power budget has been a persistent challenge. With its Neural I/o technology, built on MicroLED technology, Kopin presents a unique, compelling value proposition," said Matt Kimball, Principal Analyst at Moor Insights & Strategy.

Matt Kimball, Principal Analyst at Moor Insights & Strategy

For Kopin, this collaboration represents a dramatic expansion of its market opportunity. The company originally developed its MicroLED and NeuralDisplay capabilities for virtual reality and augmented reality applications in defense and industrial markets. By extending these technologies into AI infrastructure, one of the fastest-growing segments of the technology market, Kopin gains access to an enormous ecosystem desperately needing faster, lower-power performance.

How Does This Fit Into Broader Data Center Power Solutions?

The Neural I/o development is part of a larger industry shift toward addressing data center power constraints. Other companies are tackling the problem from different angles. For instance, Micron Technology announced that it has started shipping its largest commercially available solid-state drive, which allows more memory to be stored with lower power demands compared to traditional hard-disk drives. Meanwhile, ON.energy received a patent for medium voltage uninterruptible power supply (UPS) systems specifically designed for AI data centers, featuring advanced controls that enable operators to dynamically reduce grid consumption and support peak shaving.

The University of Utah is also experimenting with nuclear power for AI infrastructure, planning to produce electricity with its TRIGA reactor for the first time this summer to power a mini AI data center. These parallel developments suggest the industry recognizes that solving the power challenge requires innovation across multiple layers of data center infrastructure, from the chips themselves to the power delivery systems that support them.

"MicroLED-based interconnects are the leading edge in infrastructure for AI data centers. Kopin's bi-directional MicroLED technology is the foundation of our optical interconnect architecture. Their expertise in MicroLED materials and fabrication, combined with our innovative system-level design for AI factories, creates a patent-protected technology position that we believe will define the next generation of data-center communication," said Josh Silverman, Chief Executive Officer of Fabric.AI.

Josh Silverman, Chief Executive Officer of Fabric.AI

The Neural I/o technology represents a fundamental shift in how data moves through AI infrastructure. By replacing copper with light, the technology promises to deliver the same functional outcome with a fraction of the power, addressing one of the most pressing challenges facing the rapidly expanding AI industry. As data center power consumption continues to climb, innovations like this optical interconnect technology may prove essential to making large-scale AI deployment sustainable and economically viable.