HyperLight raises $80M to scale TFLN photonics as AI bottlenecks shift from chips to interconnects

HyperLight, a Harvard spinout based in Cambridge, Massachusetts, has closed an $80 million Series C to push its thin-film lithium niobate photonics into AI data centers. The round was led by chip designer MediaTek and included Foxconn, Jabil, UMC, Singapore's EDBI, Taiwan's CDIB-TEN Capital, and the Qatar Investment Authority.

The AI industry is racing to replace copper interconnects with optical links as GPU clusters grow past hundreds of thousands of units. Copper's distance and power limits have become a bottleneck that photonics can address. HyperLight's TFLN platform converts electrical signals to optical ones at high speed with low power and low loss, and its Chiplet design covers both short data center hops and longer telecom links in one manufacturable design.

Supply chain investors signal ecosystem alignment

The cap table is the real signal. MediaTek leads, but the presence of Foxconn and Jabil (contract manufacturers), UMC (foundry), and the Qatar Investment Authority means the firms that would build and buy the technology are now invested. CEO Mian Zhang called it about ecosystem alignment, not just capital.

Key facts about the round and technology:

• Products at 200G per lane are already shipping; 400G-per-lane parts are sampling.
• TFLN is an alternative to silicon photonics, which rivals like Nvidia and Marvell are pursuing.
• HyperLight's single platform is designed to scale into volume production where others have stalled.
• The new funding will go to factory capacity, customer qualification, and deeper foundry partnerships.
• The round joins a wave of investment in photonics for AI, as copper cannot feed data to growing clusters without excessive power.

What the investment means for the broader AI infrastructure build-out

Optics is one of the hottest corners of the AI infrastructure boom. Nvidia has tied up with Marvell on silicon photonics, and startups keep claiming big efficiency gains from photonic networks. HyperLight's material bet stands out because TFLN can handle higher speeds with less power and loss, particularly important as clusters scale and interconnect distances grow.

The technical claims are the company's own, and its backers stand to gain if TFLN wins. The market, not the press release, will decide whether lithium niobate becomes AI's optical workhorse. Still, the supply chain alignment gives HyperLight a production pathway that many photonics startups lack. The new cash will put that theory to the test.

Meanwhile, the broader race to connect AI infrastructure is visible in other massive commitments. France recently announced another €13 billion for its Tibi program, which directs insurers and pension funds into venture and growth funds. And in India, billionaire Mukesh Ambani is weaving AI into telecom services used by more than 500 million people. Optical interconnects are a critical layer enabling all of these ambitions.