OXMIQ Labs Inc., the GPU‑and‑AI architecture startup founded by industry veteran Raja Koduri, announced the close of a $35 million Series A financing round, bringing its total funding to $60 million. The capital, co‑led by Fundomo and Samsung Catalyst Fund, will be used to scale OxCore™, OXMIQ’s licensable GPU core, and related software and chiplet technologies for semiconductor firms and AI system builders.
OxCore™ Architecture and Licensing Model
OxCore™ is a scalable, licensable GPU core that integrates three compute engines—a CUDA®‑compatible GPU engine, a tensor processing engine, and an orchestration CPU engine. By tightly coupling functions that are traditionally spread across separate chips, OxCore aims to reduce data movement and improve energy efficiency for AI workloads, especially those that require near‑memory compute. The architecture is designed to scale from single‑core deployments to large‑scale datacenter configurations and is currently running on FPGA with live demonstrations available.
Complementing OxCore, OXMIQ’s OxQuilt™ chiplet integration framework allows heterogeneous compute chiplets and memory to be combined in a single package. Unlike many AI silicon designs that lock customers to a specific foundry or memory type, OxQuilt provides configuration tools that support multiple logic process nodes, memory technologies, interconnect standards, and advanced packaging options. The company also plans to incorporate emerging interconnects such as silicon photonics as they mature.
On the software side, OXMIQ offers a stack that includes OxCapsule™ for high‑level orchestration and OxPython™ for running existing CUDA® and PyTorch® code on OxCore without modification. OxPython has been validated on third‑party platforms and is positioned to deliver “day‑zero” support for new AI models.
Investor Backing and Strategic Partnerships
The $35 million round was co‑led by Fundomo, a New York venture firm focused on frontier compute infrastructure, and Samsung Catalyst Fund, Samsung Electronics’ evergreen venture arm that invests in deep‑tech AI infrastructure. Participants also included MediaTek, AM Intelligence Labs, Pegatron Venture Capital, CDIB‑TEN, Darwin Ventures, Morgan Creek Digital, and Intel Capital, among others.
Investors highlighted the strategic fit of OXMIQ’s architecture with their own roadmaps. Samsung Catalyst Fund’s SVP & Managing Director Dede Goldschmidt said the company’s “novel AI core and software platform enable heterogeneous compute for efficient, custom inference solutions serving large‑scale agentic workloads.” Fundomo partner Rajeev Surati noted that OXMIQ’s approach “flips a cost center into leverage” by allowing customers to design around the chip rather than the other way around.
Leadership Expansion and Market Positioning
OXMIQ added two high‑profile figures to its board and advisory network. Jim Keller, CEO of Tenstorrent and a noted chip architect, joined the board alongside existing member Dr. Ker Zhang. Dr. Valluri (Bob) Rao, a former Intel process‑technology Fellow, became an advisor. Keller emphasized the importance of an “open GPU architecture” for breaking the concentration around a few incumbents, while founder‑CEO Raja Koduri stressed that a licensable, open core can broaden access to custom AI silicon and lower compute costs.
The company’s go‑to‑market focus includes semiconductor manufacturers, “neocloud” providers, AI system builders, and physical AI/robotics firms that seek to own their compute roadmap. OXMIQ is positioning its IP‑first model as capital‑efficient, generating revenue from licensing and integration services while preserving cash for further stack development.
Key Takeaways
- OXMIQ closed a $35 million Series A round, bringing total capital raised to $60 million; the round was co‑led by Fundomo and Samsung Catalyst Fund.
- OxCore™ integrates a CUDA‑compatible GPU engine, a tensor engine, and an orchestration CPU engine into a single licensable core, currently demonstrated on FPGA.
- New board members include Jim Keller (Tenstorrent) and advisor Dr. Valluri Rao (former Intel Fellow), reinforcing OXMIQ’s silicon expertise.
TechInsyte's Take
The financing gives OXMIQ the resources to move its OxCore and OxQuilt technologies from prototype to broader customer adoption, a step that could appeal to enterprises needing custom AI silicon without the expense of full‑chip development. However, the company has not disclosed specific customer commitments or timelines for silicon tape‑out, leaving the pace of market traction uncertain. CIOs and CTOs evaluating custom AI accelerators should monitor OXMIQ’s licensing terms, FPGA demo outcomes, and any early design wins as indicators of practical viability.
Source: Businesswire
