Kopin NDA rumors with Nvidia, Broadcom

- Kopin and Fabric.AI announced on April 28 a real deal — a joint program to build MicroLED optical interconnects for AI data centers. - The hard number is $15 million: Fabric.AI agreed to fund prototype development, and its own launch filing said Kopin will own 19.9%. - The Nvidia and Broadcom angle is still rumor, not filing-backed fact — but the data-center optics pitch is now concrete.

Optical interconnects are the plumbing problem inside AI data centers. GPUs keep getting faster, but moving data between them burns power, creates heat, and starts to choke the whole system. That is the gap this story is about. The actual news is not an Nvidia or Broadcom partnership announcement. It is that Kopin and Fabric.AI disclosed a real joint-development deal on April 27 and April 28, 2026, built around MicroLED-based links for GPU-to-GPU and rack-to-rack communications. (kopin.com) ### What actually got announced? Kopin said it is collaborating with Fabric.AI to develop a MicroLED-based optical interconnect platform called Neural I/o, aimed at replacing traditional copper wiring between GPUs and other high(kopin.com)a Joint Development and License Agreement dated April 27, with Fabric agreeing to pay up to $15 million tied to prototype milestones. (kopin.com) ### Why do people care about interconnects? Because AI systems are now constrained by data movement almost as much as compute. Copper links work, but they waste energy and run into signal and thermal limits as bandwidth scales u(kopin.com)se: more bandwidth per watt, with less heat and potentially less complexity than laser-based optics. (kopin.com) ### What is Kopin bringing here? Kopin is not a new AI chip startup. It is an older optics and microdisplay company that has spent years in defense and display hardware, and it has been talking up its NeuralDisplay architecture (kopin.com) the interconnect product. (kopin.com) ### Where did the 19.9% number come from? That part is real too. Fabric.AI’s own April 28 launch filing said that, as part of the development deal, Kopin will own 19.9% of Fabric.AI. So the equity angle is not just social-media chatter. It is in the company disclosure. What is missing is any filing-backed evidence that Nvidia or Broadcom are counterparties, customers, or NDA partners in this program. (sec.gov) ### So what about Nvidia and Broadcom? Right now, that is the speculative part. I could not find any Kopin filing, Fabric.AI filing, or company press release naming Nvidia or Broadcom in connection with this interconnect effort. Fabric.AI does reference Jensen Huang indirectly by saying “AI factory” is a term coined at Nvidia GTC, but that is branding context, not proof of a commercial relationship. (sec.gov) ### Why would rumors jump there anyway? Because Nvidia and Broadcom sit close to the exact bottleneck this product targets. Nvidia dominates AI accelerator systems. Broadcom is deeply tied to networking, switching, and custom silicon inside hyperscale infrastructure. If someone claims a new optical link could matter fo(sec.gov)me as confirmed. The catch is that NDAs, by definition, are easy to speculate about and hard to verify. (kopin.com) ### How early is this? Very early. The companies are talking about prototype and demonstration milestones, not production deployments. The $15 million is development funding, not evidence of volume shipments into Nvidia racks or Broadcom-based systems. Even the bullish company language frames this as technology designed to replace copper and laser systems, not as something already adopted at scale. (kopin.com) ### Bottom line? The real story is stronger than a pure rumor and weaker than the tweet implies. Kopin does have a disclosed Fabric.AI partnership, a disclosed path to up to $15 million in development funding, and a disclosed 19(kopin.com)design wins. (kopin.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.