OXMIQ Labs, the unified GPU and AI architecture company founded by Raja Koduri, has closed a $35 million Series A round, bringing its total funding to $60 million. The latest round will help scale OxCore™, OXMIQ’s licensable GPU architecture that enables semiconductor companies and AI system builders to create custom AI silicon without having to launch a full chip program.
The round was co-led by Fundomo and Samsung Catalyst Fund, with participation from MediaTek, AM Intelligence Labs, Pegatron Venture Capital, CDIB-TEN, Darwin Ventures and Morgan Creek Digital, along with other strategic and financial investors.
OXMIQ says its work spans the full AI stack, from renewable power and data center infrastructure to silicon IP, electron-to-token machines, and the software layer that powers AI factories and agents.
Built for the AI stack
OXMIQ was founded around a simple but urgent problem that is the demand for tokens is growing faster than the infrastructure needed to support it.
To address that, the company is rethinking the GPU stack from atoms to agents, combining silicon IP, configurable systems and software to help lower the cost of intelligence across every layer of the stack.
At the center of that vision is OxCore™, a scalable GPU core built for licensing. It combines three compute engines in one architecture: a CUDA-compatible GPU engine, a tensor processing engine and an orchestration engine based on CPU control.
The idea is to bring together capabilities that are usually spread across different chips, while improving efficiency through near-memory compute and reducing the cost of moving data around. OxCore is already running on FPGA, with live demonstrations available.
Flexible chiplet approach
Another key part of the platform is OxQuilt™, OXMIQ’s chiplet integration architecture. It combines heterogeneous compute chiplets and memory in a single package, while giving customers flexibility across logic nodes, memory types, interconnect standards and advanced packaging options.
That flexibility is important because many AI silicon designs are tied to a single foundry or memory choice.
OXMIQ’s approach is designed to work across supply chains, making it easier for design teams to build custom AI silicon packages without the cost and complexity of a full chip program. The architecture is also built to accommodate emerging technologies such as silicon photonics as they mature.
Software that matches the hardware
OXMIQ couples its hardware IP with a software stack that ranges from high-level orchestration through OxCapsule™ to low-level kernel optimization.
Its OxPython™ layer is designed to run existing CUDA and PyTorch code on OxCore without requiring code changes, which gives developers more portability across hardware platforms.
The company says this stack is designed to support next-generation silicon architectures for large-scale inference and to provide day-one support for new models. OxPython has already been validated on third-party platforms, with live demonstrations available.
Capital-efficient model
Unlike full SoC companies, OXMIQ is taking an IP-first approach. By focusing on architecture and licensing rather than building complete chips in-house, the company says it can generate revenue through customer engagements while keeping more capital available to develop the broader platform.
Overall, the new funding gives OXMIQ more room to push its vision of a flexible, licensable GPU and AI architecture stack that is built for the next wave of custom AI silicon.
Leadership Comments
“We are very excited to co-lead OXMIQ’s financing round and back Raja Koduri and the strong team at OXMIQ,” said David (Dede) Goldschmidt, SVP & Managing Director, Head of the Samsung Catalyst Fund. “OXMIQ’s novel AI core and software platform enable heterogeneous compute for efficient, custom inference solutions serving large-scale agentic workloads.”
“Raja has built silicon at every layer of the stack, and he knows exactly where the constraints sit. Most compute IP makes the customer bend their memory, packaging, and foundry around the chip. OXMIQ does the opposite, and that flips a cost center into leverage. We backed this team because they will define how AI compute gets built this decade,” said Rajeev Surati, Partner at Fundomo.
An Expanding Team
OXMIQ has strengthened its board and advisory ranks with two additions that bring decades of silicon pedigree. Jim Keller, CEO of Tenstorrent and among the most influential chip architects in the industry, joins the board of directors alongside existing board member Dr. Ker Zhang. Dr. Valluri (Bob) Rao, a renowned Fellow who retired from Intel’s process technology group, joins as an advisor. Together, they deepen OXMIQ’s leadership as the company moves from architecture to customer integration.
“I am excited to join the OXMIQ board. Raja and this team are creating an open GPU architecture, a much-needed step toward removing the artificial boundaries around AI innovation. As the industry concentrates around a few incumbents, this is more important than ever. OXMIQ’s open, configurable foundation, which developers can build on and own, is exactly where compute should be heading,” said Jim Keller, CEO of Tenstorrent and OXMIQ board member.
Raja Koduri, OXMIQ founder and CEO, added: “A licensable core with an open architecture means design teams everywhere can build the custom AI silicon their work needs. Today, state-of-the-art AI reaches most people through a handful of channels, and the cost of the compute underneath is the reason. Bring that cost down, and you widen who gets to build with it. I believe AI is a force for good when it is a tool everyone can pick up and use, not just the few who can afford to build with it. Closing this round with investors who own the supply chain tells us we can get there.”
To Know More: CLICK HERE
