AI data centers have a power problem that is not in the chip. The GPUs get the headlines, but moving data between processors and memory burns energy that could otherwise run computation. The industry calls this the interconnect bottleneck, and it is becoming the binding constraint as AI clusters scale.
The standard fix is silicon photonics: transceivers that send and receive data using light instead of electrons, reducing the energy cost of each bit that crosses a chip-to-chip link. Companies including Ayar Labs and Lightmatter already sell silicon photonic transceivers in volume. CamGraPhIC, a startup founded in 2018 at the University of Cambridge's Graphene Centre, says graphene can do the same thing more efficiently — with 80 percent less energy draw than silicon versions Tech Funding News. This week, the European Commission approved €211 million in Italian state aid for CamGraPhIC to prove it can manufacture the technology at commercial scale European Commission. One of the largest single public investments in an Italian deep-tech startup is now betting on a seven-year-old company's answer to a question its competitors have already started answering with a different material IEU Monitoring.
CamGraPhIC's graphene transceivers are designed for the GPU-to-HBM connection — the link between a graphics processor and its high-bandwidth memory, where data bottlenecks are currently worst in AI systems. The company was founded in 2018 by Andrea Ferrari, director of the Cambridge Graphene Centre, and Marco Romagnoli, who leads photonic integration research at CNIT, the Italian national telecommunications consortium 2D Photonics. It operates as a subsidiary of 2D Photonics Group, a holding company backed by the NATO Innovation Fund, Sony Innovation Fund, and Bosch Ventures. The company has spent seven years developing the material science. The €211 million grant, administered by Italy and approved under EU State aid rules, funds a pilot facility near Milan expected to open in 2028 Tech Funding News.
The stated target is fab compatibility. CamGraPhIC's pitch is not just that graphene outperforms silicon in the lab — it is that graphene photonic components can be manufactured using existing semiconductor and photonics factories without major retooling. If that claim holds, the cost of switching from silicon photonics to graphene would be lower than building a new supply chain from scratch. The company's direct competitors — Ayar Labs and Lightmatter — have both argued silicon photonics can already solve the bandwidth problem at acceptable cost, and both have commercial deployments with major customers.
The €211 million is a bet on an answer that does not yet exist. CamGraPhIC has not shipped at commercial volume. Its 80 percent energy reduction figure is a company claim, not an independently verified number. The 2028 pilot facility will test whether the technology works in a production environment — not just in the lab conditions where graphene photonics has been demonstrated repeatedly. That gap between laboratory performance and factory-floor reliability is where most material bets quietly die.
The Commission approved the aid under the EU Chips Act framework, framing graphene photonic interconnects as part of the bloc's strategic semiconductor R&D objectives European Commission. Italy is providing the direct grant. The project involves collaborations with universities and research organizations in Pisa and Bergamo, alongside technology companies that have not been publicly named. Frontier IP Group, a UK-listed intellectual property firm, holds a 9.1 percent stake in 2D Photonics IEU Monitoring.
The interesting question is not whether light can move data more efficiently than electrons. It can, and it does. The question is whether a novel material — one that has been promising in research settings for a decade — can enter a commercial supply chain before the existing alternative gets cheaper and more entrenched. Ayar Labs is already shipping. Lightmatter is already shipping. The window for a material newcomer to establish a position before silicon photonics becomes commoditized is narrowing. The EU has just bet €211 million that CamGraPhIC can beat the clock.
