Oxmiq raises $35M Series A to license next‑gen AI GPU cores

Oxmiq is betting that the next wave of AI hardware innovation won’t come from a single Nvidia rival, but from a long tail of companies that want custom silicon without building full chip teams. By offering OxCore as licensable IP—more Arm than Nvidia—the company is trying to turn advanced AI GPUs into a modular component anyone with a serious budget can slot into their own devices, data centers or appliances.

If this model works, it could significantly broaden the set of actors with access to near‑frontier inference performance. That matters for the AGI race because it would loosen Nvidia’s stranglehold on architecture and push some of the innovation to the edge, where smaller players can optimize for latency, energy and cost in very specific use cases. It also gives countries and corporates anxious about supply‑chain sovereignty a way to stand up 'their own' chips while still leaning on proven cores.

The strategic wildcard is execution: IP cores live or die on ecosystem and tooling. Oxmiq will need reference designs, compilers and a healthy developer community to avoid becoming another impressive but under‑used architecture. Still, seeing MediaTek, Samsung’s venture arm and Pegatron on the cap table signals that parts of the Asian hardware ecosystem are willing to experiment with a post‑Nvidia world.