Copper cables that link GPUs and TPUs in data centres are starting to buckle under soaring data demands, wasting power and creating a major bottleneck for AI workloads. Enlightra addresses this by using tiny, chip-scale lasers that transmit data via light rather than electricity, enabling information to move many times faster while consuming only a fraction of the energy.
The Swiss-American deeptech startup has just come out of stealth with $15 million in funding from investors including Y Combinator, Runa Capital, Pegasus Tech Ventures and Protocol Labs, and has already built prototypes that can move data reliably for AI systems in a market that consulting firm McKinsey expects to reach $24 billion by 2030.
Cutting the energy footprint of AI infrastructure
Enlightra was founded in 2022 by photonics expert Maxim Karpov, a PhD and MIT Innovator Under 35, and physicist John Jost, whose earlier research in laser and quantum technologies contributed to two Nobel Prizes. Both founders saw a gap between cutting-edge optical research and the slower-moving world of data centre hardware and wanted to bring lab-grade laser technology into practical, large-scale use.
Enlightra’s chips use a single specialised laser to generate more than 50 distinct colours of light, with each colour acting as its own high-speed lane on a data highway. These colours can be controlled independently on standard silicon manufacturing lines, enabling production of millions of units per year while using far fewer lasers, less chip area, and less power than traditional designs.
Compared with companies like Ayar Labs and optics giant Lumentum, which typically rely on one or a small number of laser wavelengths, Enlightra’s multi-colour approach offers more parallel data channels without needing custom factories, and has already demonstrated multi-channel transmission over kilometres of fibre without errors
What about diversity?
On diversity, Karpov told TFN, “Male-to-female ratio is around 2:1, and we have more than 10 different nationalities.”
What’s next?
The next phase begins with pilot production in 2027, when Enlightra plans to test its optical links within GPU clusters similar to those used by major AI companies today. From there, the company aims to scale into full rack-level connections inside data centres, undersea fibre links and eventually chip-to-memory connections, supported by a 25-person engineering team with experience from Y Combinator and Intel’s Ignite programme
