OXMIQ Labs, a developer of licensable GPU IP and full-stack AI compute architecture, has raised $35 million in Series A funding co-led by Fundomo and Samsung Catalyst Fund. The round brings total funding to $60 million. OXMIQ provides OxCore GPU IP, OxQuilt chiplet architecture, OxPython CUDA compatibility layer, and OxCapsule fleet management platform, allowing semiconductor companies to build custom AI accelerators without vendor lock-in.
Supply Chain Investors Signal Strategic Bet
The timing aligns with growing demand for alternatives to closed GPU ecosystems. Tenstorrent raised $693 million in Series D in December 2024 and $800 million in Series E in November 2025. Groq secured $750 million in September 2025 and an additional $650 million in June 2026. OXMIQ differentiates by offering true licensable GPU IP with CUDA compatibility rather than selling chips or edge NPUs.
AI Compute Costs Demand 50x Efficiency Gains
Current AI infrastructure faces severe constraints. Power accounts for the majority of operating costs, and single-vendor dependency concentrates over 60 percent of AI compute. The company targets a 30-50x reduction in energy per token to make large-scale inference economically viable. Existing solutions require hundreds of millions in development costs and years of timeline for custom silicon.
Full-Stack Architecture Targets Inference Era
OXMIQ integrates hardware IP with software compatibility. OxCore combines RISC-V orchestration, tensor processing, and CUDA/PTX engines on a single core. OxQuilt enables composable chiplets across any foundry or memory type. OxPython runs unmodified PyTorch and Hugging Face workloads on non-NVIDIA hardware. This vertical integration differs from competitors focused on either CPU IP, legacy GPU designs, or inference chips.
"A licensable core with an open architecture means design teams everywhere can build the custom AI silicon their work needs." — Raja Koduri, CEO
Board Addition Validates Technical Credibility
The round brought Intel Capital and MediaTek as participants alongside co-leads Fundomo and Samsung Catalyst Fund. Jim Keller, CEO of Tenstorrent and architect of multiple landmark chips, joined the board. This investor mix provides direct access to foundry, memory, and system integration expertise rather than pure financial capital.
AI Chip Market Expands to Multi-Trillion Scale
The AI chip market stands at $100 billion in 2026 and is projected to reach $2.1 trillion by 2040. The AI inference market is expected to grow from $106 billion in 2025 to $255 billion by 2030. Semiconductor IP licensing represents a capital-efficient path to capture this growth without manufacturing overhead. Sovereign AI initiatives and hyperscalers increasingly seek licensable GPU alternatives to reduce dependency.
Raja Koduri and Jim Keller Pair on GPU Architecture
Founder Raja Koduri previously led GPU architecture at Intel, AMD, and Apple. The addition of Jim Keller to the board pairs two of the industry's most experienced silicon architects. The team brings over 500 years of combined GPU and AI experience with prior products generating more than $100 billion in revenue impact.