Fabric.AI said it expects to demonstrate its Neural I/o MicroLED-based optical interconnect platform by the end of 2026, a system the company is developing with Kopin Corporation to speed data movement between GPUs, accelerators, memory systems and servers inside AI data centres.
The demonstration is pitched as a response to an emerging bottleneck in AI infrastructure and Fabric.AI, a fabless semiconductor infrastructure company, said it has signed two non-disclosure agreements with leading chipmakers to discuss integration while holding approximately $30 million in cash and cash equivalents, which it believes is sufficient to fund operations through the planned demo.
Related reading
- OpenAI closes $122bn funding round at $852bn valuation as IPO pressure builds
- OpenAI deploys GPT-5.5 across Nvidia on Blackwell infrastructure in enterprise-scale AI agent rollout
- Google unveils two new AI chips at Cloud Next as hyperscaler hardware race intensifies
“The industry increasingly understands that AI performance is no longer determined solely by compute, it is determined by how efficiently data can move throughout the system,” Fabric.AI Chief Executive Officer Josh Silverman said in the announcement.
Fabric.AI says Neural I/o’s MicroLED architecture may deliver lower power consumption, lower latency, improved thermal efficiency, greater density and superior scalability versus traditional copper and laser-based approaches, the company framed the late-2026 demonstration as the next milestone toward building a foundational interconnect layer for what it calls AI factories and said it believes the market opportunity for next‑generation AI interconnects could ultimately exceed $100 billion annually.
