Eliyan Raises $145 Million to Keep Expensive AI Chips From Sitting Idle

Buying faster AI chips does not help if those chips spend their time waiting for data.

Eliyan has raised $145 million in a Series C round at a $1 billion valuation to address that problem. The Santa Clara company is developing interconnect technology that moves data between processors, memory and other components inside AI systems.

The company plans to make money in two ways: license the technology to chip designers and sell modular chiplets that can be integrated into custom processors. Initial chiplet shipments are expected during 2026.

Eliyan is positioning itself as an independent alternative for companies that do not want their connectivity layer controlled by a competing accelerator supplier. Nvidia addressed the problem through Mellanox. Google and Amazon work with companies including Broadcom and Marvell. Eliyan is targeting the growing group of companies designing their own AI silicon.

Cisco and Lumentum participated in the round, which was led by Seligman Ventures. The investor mix matters because this is not only a semiconductor-design problem. It touches networking, optics, packaging and the systems required to keep large AI clusters utilized.

Eliyan says many GPUs operate at only 30% to 40% utilization because they cannot receive data quickly enough. That is a company claim, not an independent industry benchmark. Its projection of hundreds of millions of dollars in revenue by the end of 2027 should be treated the same way: possible, but not proven.

The larger signal is clear. AI infrastructure investment is moving outward from accelerators into memory, optics, networking and chip-to-chip connectivity. The processor gets the attention. The interconnect determines how much of that processor a customer can actually use.