Today, the biggest challenge in AI is not the chip itself but how chips connect to one another. As AI models get larger and more complex, moving data between processors, accelerators, and memory has become the main obstacle.
CamGraPhIC, a company founded in 2018 at the University of Cambridge’s Graphene Centre by Professor Andrea Ferrari and Dr Marco Romagnoli, has worked on this problem for seven years.
This week, it received €211 million in Italian state aid, approved by the European Commission, to bring its graphene-based optical interconnect technology to industry. This is one of the largest public investments ever made in an Italian deep-tech startup.
The company makes graphene-based optical transceivers that use light, not electrons, to send and receive data. In these devices, graphene replaces the silicon used in standard photonic interconnects.
CamGraPhIC says its transceivers use 80% less energy than traditional silicon versions, offer higher bandwidth and lower latency, and operate reliably across a wide temperature range without requiring complex cooling. The first use for this technology is the GPU-to-HBM (high-bandwidth memory) connection, where data bottlenecks are a big problem in AI systems.
Its direct competitors include Ayar Labs and Lightmatter. What makes the graphene approach different is its claimed energy savings over silicon and its possible fit with current semiconductor factories. If CamGraPhIC can show it can produce at scale using existing factories, the cost savings could be large.
The €211 million grant is meant to test this ability. A pilot facility near Milan, set to open in 2028, will show whether the technology works with commercial semiconductor and photonics factories for production.
