Back to Hyderabad

OXMIQ Labs raises $35 million to scale AI GPU development in Hyderabad

OXMIQ Labs raises $35 million to scale AI GPU development in Hyderabad

OXMIQ Labs, a California-based GPU and AI architecture startup with a development site in Hyderabad, has raised $35 million in a Series A funding round. Founded by former Intel leader and chip veteran Raja Koduri, the company has now brought its total capital raised to $60 million. The startup, which operates its development site in Hyderabad and has its headquarters in California, plans to utilize the newly secured capital to scale its custom AI silicon architectures, which include the OxCore and OxQuilt platforms.

The Series A funding round was co-led by Fundomo and Samsung Catalyst Fund. The investment round also drew participation from a diverse group of strategic and financial investors, including MediaTek, AM Intelligence Labs, Pegatron Venture Capital, CDIB-TEN, Darwin Ventures, and Morgan Creek Digital. In addition to these financial and strategic investors, Intel Capital is involved with the startup as a strategic intellectual property partner.

A primary focus for the funding will be to scale OxCore, which is OXMIQ Labs' licensable GPU architecture. This architecture is specifically aimed at semiconductor companies, AI system builders, and sovereign computing programmes that are seeking custom AI silicon but want to avoid the resource requirements of a full chip programme.

The OxCore architecture integrates a CUDA-compatible GPU engine, a tensor processing engine, and a CPU-based orchestration engine. These three components are combined into a scalable core that has been designed for near-memory compute. According to the company, the OxCore architecture is already running on FPGA for live demonstrations.

In addition to OxCore, the startup is also building OxQuilt, a chiplet integration architecture. OxQuilt is designed to combine heterogeneous compute chiplets and memory together in a single package. The architecture is engineered to provide system builders with flexibility across various chip foundries, memory types, interconnects, and packaging choices.

Share