“What if electronic bottlenecks didn’t exist, and you could move data seamlessly between chips?” With that question, Cédric Huyghebaert, CTO and co-founder of Black Semiconductor, captured the ambition behind one of Europe’s most daring photonics startups. Speaking at PIC Summit Europe 2025 in Eindhoven, he explained how his company is harnessing the unique properties of graphene to merge photonics directly into semiconductor production. “We call it a new category of chips,” he said. “Because now, your CMOS can talk two languages: electronics and photonics.”

From the lab to the fab: the long road of graphene

“When we started studying graphene,” Huyghebaert recalled, “everyone thought we’d found the new silicon.” With its extraordinary mobility and conductivity, researchers imagined replacing CMOS transistors entirely. “But fortunately,” he smiled, “graphene doesn’t have a band gap. It has a weird band structure, and that turned out to be our advantage.”

Early experiments, he explained, were simple: a single atom-thick graphene layer exfoliated in the lab. “We put light to it and discovered it absorbed 2.3% of the light; a lot for a monolayer, but not enough to build devices.” The key insight came when scientists laminated graphene onto a waveguide, enabling light to repeatedly interact with the graphene layer. “Then you could tune the absorption, even absorb all the light if you wanted,” said Huyghebaert.

That realization opened the door to graphene modulators and photodetectors, devices that can both manipulate and detect light at extraordinary speeds. “It worked fantastically in the lab,” he said. “But how do you connect it to electronics? How do you bring that to industry scale? That’s where Black Semiconductor started.”

A new chip category: EPICs, not just PICs

Traditional approaches to combining photonics and electronics rely on complex and costly advanced packaging: bonding, thinning, and stacking multiple dies. “It’s very complicated,” Huyghebaert said. “What we propose instead is an EPIC: an electronic-photonic integrated circuit. We integrate both technologies, using standard industrial processes. No lift-off, no exotic tools.”

The result is revolutionary simplicity: photonics that can be fabricated within standard CMOS lines, enabling data movement by light instead of electrons. “Whatever you make on your chip below,” he said, “suddenly it can talk two languages. It can compute electronically and communicate optically. That’s a game-changer.”

The challenges: quality, reproducibility, scalability

Graphene’s potential is matched by its headaches. “You can’t just go to a foundry and ask them to integrate graphene - at least, we didn’t find any that could,” Huyghebaert admitted. The main issues? Quality and reproducibility. “Graphene is known for being fantastic and unreliable,” he said. “So we’re developing single-crystal graphene, grown on 200-millimeter wafers and scaling to 300 millimeters. That’s crucial for stability.”

The transfer process, too, is tricky. “We grow graphene on a template, then move it to the wafer, but keeping it pristine is hard. In CMOS history, reproducibility was also the key bottleneck until the industry learned how to control the gate oxide. Now we must do the same for graphene.”

The third challenge is scalability. “Even if it works, can we make 100,000 wafers a year? A million?” he asked. “If you want photonics on every chip, you need a plan for mass manufacturing.”

Europe’s graphene moment

Aachen-based Black Semiconductor’s answer came through European collaboration. “Europe has a deep legacy in graphene research,” Huyghebaert reminded the audience, referencing the EU’s Graphene Flagship program. “So in 2021 we started talking with policymakers, convincing them that Europe should keep this technology here.”

Their pitch succeeded. In June 2024, Black Semiconductor secured an IPCEI grant (Important Project of Common European Interest) to build a 300-millimeter pilot line for integrated graphene photonics. “We’re designing the fab now,” Huyghebaert said. “It will be operational by mid-2026, fully up and running in early 2027, with product sampling soon after.”

The company’s growth has been just as rapid. “We were two people in 2022,” he said. “We’re 130 today and will be 240 by next year.” That pace brings its own challenges. “Hiring talent is hard,” he admitted, “but aligning that talent is even harder. It takes time for new people to become efficient; that’s what we’re working on now.”

Beyond silicon photonics

Today’s silicon photonics (SiPh) technology - the workhorse of optical interconnects - faces limits when integrated into the back-end-of-line (BEOL) of chip manufacturing. “SiPh requires high temperatures and materials that aren’t BEOL-compatible,” Huyghebaert explained. “Graphene, by contrast, fits perfectly. It’s low-temperature, CMOS-compatible, and enables both high-speed modulators and detectors.”

That compatibility, he argued, could redefine computer architecture. “It allows us to rethink how chips communicate; directly through light, within the same stack. That’s how we break the interconnect bottleneck holding back AI.”

In the company’s vision video, he summarized it succinctly: “Graphene photonics eliminates electronic bottlenecks for limitless data throughput.”

The next step: glass and light

Black Semiconductor is also developing glass panel interposers, an emerging platform for next-generation compute. “Glass reduces signal loss, improves bandwidth, and enables more complex system architectures,” said Huyghebaert. “Combined with integrated graphene photonics, it creates a seamless optical fabric between chiplets.”

In early lab tests, their graphene modulators already show promising speeds: 5 GHz today, with a roadmap toward 20–25 GHz modulation and 60 GHz photodetectors. “We’re just taking our first steps,” he said, “but we know the path forward.”

Toward the post-silicon era

To close, Huyghebaert pointed to his last academic paper before founding Black Semiconductor. “It was called ‘2D materials can make the back end of line intelligent.’ That’s what we’re doing: making the back end smarter.”

The goal, he said, is not to replace CMOS, but to augment it, turning the back of the chip into an active optical layer. “It’s the first version of a new category of chips. Electronics and photonics, truly integrated. That’s how we rethink computing.”

Huyghebaert’s message was clear: the next revolution in computing won’t come from faster transistors alone; it will come from light, flowing through a single layer of carbon.