Fabric.AI Files Two Patent Applications Covering Fiber-Coupled and Connector-Free MicroLED Optical Interconnect, Developed with Kopin

Both applications derive from a single optical approach designed to hold assembly complexity flat as channel counts rise; Company remains on track to show first demonstration hardware in January 2027

Fabric.AI, an AI infrastructure company developing a suite of fabless semiconductor technologies for next-generation AI factories, announced that it has filed two patent applications with the United States Patent and Trademark Office (USPTO) covering its micro-LED optical interconnect architecture.

Both applications derive from the same design choice: imaging an entire micro-LED array through a novel optical system, rather than aligning light channel by channel. Through their partnership, Fabric.AI and Kopin Corporation (Nasdaq: KOPN) are jointly developing MicroLED-based optical interconnect technology designed to address bandwidth and energy-efficiency constraints in next-generation AI data centers.

Fiber-coupled: one optic, however many channels

The first application covers the Company’s approach to fiber-coupled optical interconnect, using a novel optical system to image an entire micro-LED array onto the cores of a multicore optical fiber. The architecture is designed to hold the number of precision alignment steps constant as channel count rises, rather than requiring alignment channel by channel.

Fabric.AI believes this is the property that allows dense parallel optical links — thousands of channels within a single square millimeter of die — to scale between boards, trays and racks without a corresponding rise in assembly cost.

Also Read: AiThority Interview with Gou Rao, co-founder and CEO at NeuBird AI

Free-space: the same optic, without the fiber

The second application takes the same optical approach to its limit. Rather than imaging the array onto a fiber face, it transmits parallel data across a short air gap, eliminating the need for any physical connectors between the optical endpoints.

The technology uses compact optics to image the micro-LED array directly onto a matching detector array, holding the channels separate. Because the coupling is imaged rather than bonded, the fiber becomes optional: for short, fixed gaps inside a package or between adjacent boards, removing the connector removes both a failure point and an assembly step.

“Both applications come from one architectural choice: we image the whole array using special optics instead of aligning light channel by channel. That is what we believe keeps assembly complexity flat as we scale from hundreds of optical channels toward thousands — and what lets us remove the connector entirely where the gap is short enough,” said James Altucher, Fabric.AI Senior Strategic Advisor. “We continue to collaborate with Kopin in refining our technology and look forward to sharing these innovations in more detail with those potential customers currently under NDA discussions, as well as to providing more updates as they develop.”

These applications also underscore Fabric.AI’s continued collaboration with Kopin on Neural I/o™, its optical interconnect architecture designed to address the increasing data-movement, bandwidth and energy-efficiency demands of next-generation AI infrastructure. The Company expects to demonstrate working Neural I/o hardware in January 2027.