NcodiN, a deep-tech startup in Paris, just secured €16 million in seed funding and launched one of the most ambitious hardware innovations in the global AI ecosystem. The company develops nanophotonics technology that replaces traditional copper interconnects inside advanced AI processors. The founders describe this problem as the “copper wall” because copper traces inside chips can no longer support the bandwidth and energy efficiency that modern AI workloads demand. NcodiN wants to break that wall and reshape the hardware landscape that fuels the world’s fastest-growing AI systems.
The Copper Wall Slows Down AI Growth
AI models grow in size, complexity and dataset volume every year. New models demand enormous memory bandwidth and extremely fast communication between chiplets, accelerators and memory stacks. Engineers design advanced chips with dozens or even hundreds of chiplets, but copper interconnects struggle to handle this architecture.
Copper wires limit data speeds because resistance and capacitance slow signals down. They also produce heat, consume large amounts of power and degrade signal quality over distance. Engineers try to shorten wire lengths or thicken traces, but these tricks no longer solve the core issue. AI processors now depend on rapid chip-to-chip communication more than ever, and copper cannot sustain this growth. The AI industry often claims that “compute scales faster than input/output,” and NcodiN builds its entire mission around solving this imbalance.
NcodiN calls this bottleneck the copper wall. If chipmakers cannot move data quickly enough across the chip package, then even the most powerful compute cores cannot reach their potential. AI-training clusters already waste energy while they wait for data movement. The copper wall blocks the next leap in AI hardware efficiency, and NcodiN wants to tear it down.
NcodiN Introduces the World’s Smallest Laser for On-Chip Photonics
NcodiN approaches the problem with a photonic interposer that replaces copper connections with optical links. The company integrates nanolasers directly onto silicon and routes data through light instead of electricity. This solution uses a platform the team calls NConnect. The platform combines an ultracompact semiconductor laser with a photonic interposer that supports high-density optical input/output.
The engineers at NcodiN design nanolasers that occupy extremely small surface areas and consume very little energy. They build these lasers directly on silicon rather than on compound semiconductor substrates. This choice allows tighter integration with existing chip manufacturing and packaging processes. The team demonstrated energy efficiency below 0.1 picojoules per bit, an achievement that traditional electrical interconnects cannot approach at scale.
The NConnect interposer routes light through waveguides instead of routing electrons through metal. Light does not generate heat in the same way copper does, so the platform significantly reduces power consumption. The optical links also carry signals across longer distances without distortion, which frees engineers to design larger chiplet layouts. NcodiN claims that its photonic architecture supports next-generation chips that integrate dozens of memory stacks and accelerator units on a single substrate.
The Technology Shifts the Limits of AI Hardware
NcodiN’s approach unlocks multiple new possibilities for chip designers:
- Higher Bandwidth
Engineers can push data through optical links at far higher rates than copper. The company enables architectures that require massive parallel memory access, which modern AI models demand during both training and inference. - Lower Energy Consumption
Copper interconnects drain energy because they resist electrical flow and generate heat. Photonics avoids this problem and reduces power consumption per bit. - Longer Reach Inside Packages
Optical signals travel farther inside multi-chip modules. This capability allows larger wafer-scale or near-wafer-scale designs. - Better Thermal Stability
Lasers and waveguides produce far less heat than copper traces under similar loads. This benefit lowers cooling costs inside data centers. - Compatibility with Chiplet Ecosystems
NcodiN designs NConnect for the chiplet era. Engineers can combine logic cores, memory stacks and accelerators inside one advanced package and link them through light rather than electricity.
This shift does not only optimize performance. It redefines the architecture of future AI processors. AI companies already explore wafer-scale compute, memory-rich chiplets and extremely large accelerator modules. NcodiN gives this new generation of hardware the physical communication layer it needs to scale.
Investors Signal Strong Support for Deep-Tech Hardware
NcodiN’s €16 million seed round attracted top European and global deep-tech investors. Several venture funds that specialize in photonics, semiconductor manufacturing and AI infrastructure joined the round. Investors view NcodiN not as a niche research project but as a foundational technology provider for future AI hardware.
The company now plans to expand its cleanroom operations, strengthen its engineering team, refine its integration with 300 mm CMOS processes and begin industrial-scale prototyping. The founders will also build new partnerships with chipmakers, packaging firms and data-center operators. AI companies increasingly look for breakthroughs in system-level hardware, not just in model architecture. Investors believe that innovations like NConnect will influence the next decade of AI performance.
NcodiN’s Roadmap and Upcoming Milestones
NcodiN now enters a critical phase. The team must translate its laboratory breakthroughs into manufacturable, reliable photonic systems that operate at industrial scale. The roadmap includes several important steps:
- A Working 300 mm Production-Ready Flow
The company must prove that its nanolaser technology integrates cleanly with standard semiconductor fabrication processes. The team also needs to maintain high yield across full wafers. - Full Reliability and Thermal Testing
AI accelerators operate under extreme thermal pressure. NcodiN must demonstrate long-term laser stability and low-loss waveguide performance under such conditions. - Partnerships with Chiplet Designers and Hyperscalers
Success requires close collaboration with companies that design next-generation accelerators. Partnerships with AI labs and cloud providers will validate the real performance gains of NConnect. - Scaling Optical Input/Output Density
The company aims to pack more optical channels into the interposer and reach bandwidth levels that far exceed anything copper can support. - Cost Reduction Through Integration
NcodiN must lower cost through wafer-level integration and simplified packaging. If the company succeeds, it will position optical interposers as a mainstream solution.
A Global Impact on AI Hardware
NcodiN’s technology influences markets far beyond Europe. AI growth in the United States, India, China and the Middle East depends heavily on new hardware breakthroughs. NcodiN offers a scalable way to remove one of the biggest limitations in modern AI systems. Hardware engineers across the world will pay close attention to the company’s progress. If NConnect enters mass production, it will reshape chiplet design, datacenter architecture and global semiconductor strategy.
Conclusion
NcodiN stands at the center of a major shift in AI hardware design. The company attacks the copper wall with a photonic interposer and an incredibly compact on-silicon laser. This breakthrough promises higher bandwidth, lower energy consumption and far more scalable chip architectures. With €16 million in new funding, NcodiN now moves from research to industrial development. The AI world eagerly watches this transition, because success at this level could transform the entire foundation of artificial-intelligence computing.
Also Read – Top 10 Most Weirdest Startup Ideas That Worked