AIBRIEF
The two year old London startup designs chips, lasers, and the optical network connecting them, and closed a Series B at $3.3B with Arm and Reed Hastings on the cap table.
OLIX, a two-year-old London AI chip startup, raised a $312M Series B at a $3.3B valuation this week. Fundomo led the round, with Arm, Hudson River Trading (a quantitative trading firm), and Reed Hastings joining as new investors. Existing backers Hummingbird Ventures, Crane, Plural, Creandum, Phoenix Court, and Transition all increased their commitments.
The round is funding a specific bet about how AI inference should be built. OLIX argues today's datacenters run every stage of inference on the same general-purpose chip, and that uniformity is the efficiency ceiling. The startup designs the chips, the lasers, and the optical network that connects them, and ships complete racks tuned for inference.
The X-1 platform, OLIX's rack-scale inference system, splits token production across specialized stages, with each chip focused on one part of the model rather than the whole. A "slow and wide" optical interconnect moves data between chips using light rather than copper, and a deterministic compiler schedules work across racks. The first chip, the DX-1, targets decoding, the stage where a model generates output. OLIX claims the DX-1 delivers more than 10,000 tokens per second per user for 100B-parameter models.
That figure is OLIX's own benchmark. The company has not announced volume customers or production deployment. The Series B also appoints Professor Nick McKeown to the board.