Oxmiq Labs, founded by former Intel chief architect Raja Koduri, plans to launch licensable GPU technology focused on AI computing.

The company has raised US$20 million in seed funding from angel and corporate investors, including MediaTek.

Based in Campbell, California, Oxmiq’s GPU IP is designed to scale from single-core use in robotics to thousands of cores for cloud data centers. Its architecture can be customized for specific computing needs.

The company has also built a tool that allows software written for Nvidia’s CUDA to run on other hardware without code changes.

Oxmiq chose to license IP instead of building chips to avoid high development costs.

Koduri previously held senior roles at Intel, AMD, and Apple.

🔗 Source: Reuters

🧠 Food for thought

1️⃣ IP licensing strategy addresses massive chip development costs

Oxmiq’s decision to license GPU intellectual property rather than manufacture complete chips reflects the enormous financial barriers in modern semiconductor design.

Cutting-edge chip design can cost more than $500 million, making it prohibitively expensive for startups to compete directly with established players1.

This cost reality has driven the semiconductor industry toward specialized business models, with the overall market reaching $627 billion in 2024 and projected to hit $1 trillion by 20302.

The IP licensing approach allows Oxmiq to monetize their technology innovations without the massive capital expenditure and manufacturing risks that come with full chip production.

Companies are projected to invest $185 billion in capital expenditures just to expand manufacturing capacity in 2025, highlighting the scale of investment required for hardware production3.

2️⃣ Breaking NVIDIA’s software ecosystem dominance requires CUDA compatibility

Oxmiq’s emphasis on CUDA compatibility without code modification addresses NVIDIA’s strongest competitive moat beyond pure hardware performance.