As artificial intelligence (AI) and high-performance computing (HPC) systems scale, the semiconductor industry is hitting a severe physical barrier: the data-movement bottleneck. Traditional copper wiring simply cannot keep up with the bandwidth and energy efficiency demands of modern computing architectures.
To solve this critical challenge, French deep-tech startup NcodiN has announced a strategic collaboration with CEA-Leti, a globally recognized research institute. Together, they are embarking on a mission to industrialize NcodiN’s groundbreaking optical interposer technology on a commercial-grade 300 mm integrated photonics process.
At AarokaTech, we closely monitor the technologies bridging the gap between legacy hardware and future computing demands. This partnership marks a massive leap toward scalable, in-package, long-reach optical links that will eventually replace copper interconnects in advanced AI chips.
Moving Beyond Copper: The AI Interconnect Bottleneck
The relentless demand for data processing power in AI systems requires orders of magnitude increases in both bandwidth and energy efficiency. While processors have become incredibly fast, the physical copper connections moving data between them generate too much heat, consume too much power, and limit overall system speed.
The industry’s solution is a fundamental shift toward optical interconnects—using light instead of electricity to transmit data. However, previous photonic components have often been too bulky or inefficient for large-scale, dense integration.
Backed by a recent €16 million seed financing round secured last November, NcodiN is positioned to change this narrative by bringing unprecedented miniaturization to integrated photonics.
NConnect: The World’s Smallest Laser on Silicon
The core of NcodiN’s innovation is the NConnect platform, an integrated optical interconnect system powered by what the company touts as the world’s smallest laser on silicon.
This proprietary nanolaser technology shatters previous physical limitations, boasting a footprint that is 500 times smaller than today’s industry-standard photonic devices. This extreme miniaturization enables:
- Ultra-Dense Integration: The ability to pack more than 5,000 nanolasers per square millimeter (>5,000/mm²).
- Record-Low Energy Operation: Achieving unmatched power efficiency at approximately 0.1 picojoules per bit (~0.1 pJ/bit).
By utilizing nanolaser-enabled photonic interposers, chip designers can facilitate massive data transfer rates directly within the semiconductor package, drastically reducing latency and power consumption.
The 300 mm Wafer Transition: A Turning Point for Commercialization
Proving a technology works in a lab is only half the battle; manufacturing it at a commercial scale is the true test. By leveraging CEA-Leti’s extensive expertise in advanced photonics integration, NcodiN is transitioning its nanolaser technology to a standard 300 mm silicon photonics platform.
“NcodiN’s nanolaser-enabled photonic interconnects overcome the long-standing bottleneck of bulky, inefficient photonic components that have prevented large-scale adoption,” explained Francesco Manegatti, Co-founder and CEO of NcodiN. “Our collaboration with CEA-Leti aims to demonstrate NConnect’s compatibility with 300 mm wafers, which is essential for commercial-scale production and cost-effective adoption in AI-centric processors and high-bandwidth computing systems.”
Sébastien Dauvé, CEO of CEA-Leti, echoed this sentiment, highlighting that transitioning to a 300 mm CMOS-compatible process is a definitive turning point for the industry.
“This is a turning point for optical interconnects that can finally be produced at the scale, cost, and reliability the AI industry requires,” Dauvé stated. “This collaboration highlights a key part of CEA-Leti’s mission: transferring advanced semiconductor and microelectronics technologies to industry, where they serve a range of vital markets.”
As AI infrastructure continues its explosive growth, the shift from electronic to photonic data transmission is no longer just a theoretical roadmap—it is becoming a manufacturing reality.
