Analysis
Eliyan closed a $145 million Series C led by Seligman Ventures, with Cisco Investments and Lumentum among the participating investors, pushing the company to unicorn status at a $1 billion valuation. The round funds scaling of Eliyan's NuLink PHYs and NuGear chiplet families, silicon designed to solve one of the AI infrastructure buildout's least-discussed but most expensive problems.
The bottleneck is straightforward: in dense AI clusters, GPUs typically run at just 30-40% utilization, not because the chips themselves are slow, but because they can process data far faster than they can send or receive it across the cluster. That means some of the most expensive silicon in the world sits idle a majority of the time waiting on data movement -- a problem that gets worse, not better, as cluster sizes scale into the hundreds of thousands of GPUs that frontier labs are now deploying.
“Eliyan's chiplet interconnects are designed to widen that data pipe rather than speed up the compute itself.”
Eliyan's chiplet interconnects are designed to widen that data pipe rather than speed up the compute itself. The company expects to begin initial chiplet shipments this year and forecasts hundreds of millions of dollars in sales by the end of 2027, up from low single-digit millions in 2025 revenue -- an aggressive but not unreasonable trajectory if AI cluster build-outs continue at their current pace.
For infrastructure-focused investors, Eliyan's raise is part of a broader rotation of venture capital toward the physical interconnect and packaging layer of AI compute -- alongside ChipAgents' Nvidia-backed chip-design round and GlobalFoundries' federal photonics grant -- rather than the GPUs themselves, where competition among a handful of well-capitalized players makes new entrants a harder bet. What to watch: whether Eliyan's first chiplet shipments this year hit committed customer timelines, and whether utilization-rate improvements at early customers become public reference data.
