Oxmiq Labs raises $35M Series A for licensable AI chip architecture targeting sovereign AI
The AMW Read
Oxmiq introduces a novel licensable GPU model in AI infrastructure, updating the player map and addressing sovereign AI demand; novelty is high as no major AI chip IP licensing startup has emerged at this scale, significance is segment-level with potential cross-segment impact on compute economics.
Oxmiq Labs raises $35M Series A for licensable AI chip architecture targeting sovereign AI
Oxmiq Labs, the AI chip architecture startup founded by former Intel graphics chief Raja Koduri, has raised $35 million in a Series A funding round co-led by Samsung Catalyst Fund and Fundomo. The round brings Oxmiq's total funding to $60 million, with participation from MediaTek, Pegatron Venture Capital, Intel Capital, and others. Oxmiq is building a licensable GPU architecture called OxCore, positioning itself as an "Arm for AI GPUs" — enabling semiconductor firms, neocloud providers, and governments to build custom AI processors without starting from scratch or relying solely on Nvidia. The company's initial commercial focus is India and Southeast Asia, and it has partnered with AM Intelligence Labs to architect a 2 GW renewable-powered AI compute platform in Uttar Pradesh.
Why it matters: Oxmiq's licensable GPU model exemplifies the emerging "context-engineering moat" pattern in the AI infrastructure segment — where the value shifts from raw silicon to the IP layer that enables customization at lower cost. More importantly, the company directly addresses the sovereign AI theme that is reshaping compute economics in emerging markets: Raja Koduri explicitly argues that India needs to own critical AI chip IP to reduce AI compute costs by 50-100x. If Oxmiq succeeds in licensing GPU designs to governments and regional semiconductor players, it could fundamentally alter the capital cycle dynamics of the AI infrastructure buildout, offering an alternative to the hyperscaler-distribution model that currently funnels most AI compute spend through Nvidia, AWS, and Azure.
Grounded expert take: This is a structurally significant bet on two fronts. First, the licensable IP model — borrowing Arm's playbook — directly challenges the dominant vertically integrated GPU model (Nvidia) and the hyperscaler in-house chip push (Trainium, TPU). Second, the sovereign AI angle, coupled with India/Southeast Asia focus, taps into a geopolitical force that is accelerating as governments seek to reduce dependency on US-controlled compute supply chains. The $35M Series A, while modest relative to the capital required for chip tape-outs, positions Oxmiq as a player in a space where China-based entrants like Cambricon and Biren have faced export control headwinds, potentially creating an open window. However, the road to commercial silicon is capital-intensive and execution-heavy — Oxmiq will likely need to demonstrate customer tape-outs and software ecosystem adoption before the model truly threatens incumbents.