OXMIQ Labs has raised $35 million in Series A funding to build and commercialize its licensable AI GPU architecture. The startup was founded by Raja Koduri, a well-known chip industry leader who earlier worked at Intel, AMD, Apple, ATI Technologies and S3 Graphics.

The round was co-led by Samsung Catalyst Fund and Fundomo. Existing investors including MediaTek, Pegatron Venture Capital, CDIB-TEN, Darwin Ventures, Morgan Creek Digital, AM Intelligence Labs and Intel Capital also joined the round. With this funding, OXMIQ’s total capital raised has reached about $60 million.

The big idea is simple but ambitious. OXMIQ wants to help companies, cloud providers and governments build custom AI chips without starting from zero. Instead of making its own consumer GPU like Nvidia GeForce or AMD Radeon, the company wants to license GPU intellectual property and software tools.

What OXMIQ Labs does

OXMIQ Labs is building a licensable GPU platform for AI and advanced computing. Its main product is called OxCore.

A GPU, or graphics processing unit, is a chip that can handle many calculations at the same time. GPUs were first known for gaming and graphics, but today they are central to AI training, AI inference, data centres, scientific computing and large language models.

OXMIQ’s approach is different from selling finished chips. It wants to provide the core design and software layer so customers can create their own AI processors.

A simple example may help. If a company wants to build an AI chip for a data centre, it usually needs a large chip design team, expensive tools, years of engineering and deep software support. OXMIQ wants to reduce that effort by offering a ready architecture that customers can license and customize.

Founder and founding year

OXMIQ Labs was founded in 2024 by Raja Koduri. The company is headquartered in Campbell, California, and has engineering teams in Hyderabad and Bengaluru.

Raja Koduri is a major name in the GPU world. He has worked on graphics and compute technology across companies such as AMD, Intel, Apple and ATI Technologies. At Intel, he was known for leading graphics and architecture work. At AMD, he was connected with Radeon graphics leadership.

That experience matters because GPU architecture is not a small software problem. It needs deep knowledge of hardware, software, developer tools, memory, power use and manufacturing constraints.

Why the $35 million funding matters

The Series A funding will help OXMIQ complete OxCore by the end of 2026 or early 2027, before customer deployments and commercial silicon tape-outs.

A tape-out is the stage where a chip design is sent for manufacturing. In simple words, it is the point where the design moves closer to becoming a real physical chip.

This funding is important because the global AI market is facing a serious compute challenge. AI models need a huge amount of processing power. Today, much of that market is led by Nvidia. Many companies and governments want more options because AI infrastructure is becoming expensive and strategically important.

OXMIQ is trying to offer one such option through licensing.

What is the aim of OXMIQ Labs

OXMIQ’s main aim is to make AI chip development more accessible. It wants companies and governments to build custom AI processors without spending billions of dollars and many years creating everything from scratch.

This is especially relevant for countries and regions that want sovereign AI. Sovereign AI means having more control over AI infrastructure, data, chips and software instead of depending fully on foreign technology providers.

For India, Southeast Asia and other emerging markets, cost is a major issue. AI compute is expensive. If custom chips can be built at lower cost using licensed architecture, more companies may be able to run AI workloads locally.

Raja Koduri has argued that countries like India need control over AI compute costs if they want wider AI adoption. That is the larger purpose behind OXMIQ’s work.

OxCore and the software stack

OxCore is OXMIQ’s flagship GPU architecture. It is based on RISC-V, an open instruction set architecture. In simple terms, RISC-V gives chip designers more flexibility compared with fully closed instruction systems.

Reports say OxCore combines scalar, vector and tensor compute engines. These are different types of processing units used for general tasks, parallel workloads and AI calculations.

But hardware is only one part of the story. AI chips also need strong software. Developers do not want to rewrite everything for every new chip.

This is where OXMIQ’s software tools become important. Its software stack includes OXCapsule and OXPython. OXPython is designed to help Python-based CUDA workloads run on non-Nvidia hardware without major changes.

This is a big point because CUDA, Nvidia’s software platform, has become a major reason behind Nvidia’s strength in AI. If OXMIQ can make software movement easier, it may reduce one of the biggest barriers for alternative AI hardware.

Who OXMIQ is targeting

OXMIQ is not mainly targeting normal laptop or gaming buyers. Its focus is on semiconductor companies, AI infrastructure builders, neocloud providers and governments working on sovereign AI programmes.

A neocloud provider is a newer cloud company focused heavily on GPU and AI compute services. These companies need large amounts of processing power for AI customers.

Governments may also be interested because AI infrastructure is becoming important for national competitiveness. If a country wants to build its own AI capacity, it may not want to depend only on imported hardware and software.

OXMIQ has also launched a public beta of its software stack. According to reports, the stack is being used by 20 companies and 10 universities across nearly 300 GPUs.

Competitors and market context

OXMIQ is entering a tough market. Nvidia is the clear leader in AI GPUs. AMD is pushing its Instinct accelerators. Intel, Google, Amazon, Microsoft, Tenstorrent, Cerebras, Groq and several other companies are also active in AI chips and accelerators.

The difference is that OXMIQ is not trying to compete only by selling finished chips. Its model is closer to licensing chip architecture, similar to how Arm licenses CPU designs.

This can be useful for companies that want custom chips but do not want to build a full GPU architecture from the ground up.

Still, the challenge is huge. GPU architecture needs trust, performance, software support and customer adoption. Even if the design is strong, customers will ask whether the tools are stable, whether developers can use them easily, and whether the chips can be manufactured at scale.

Why this matters for India

OXMIQ has engineering teams in Hyderabad and Bengaluru, which gives India a meaningful connection to the story.

India wants to build strength in semiconductors, AI infrastructure and deeptech. Most discussions around AI focus on apps and models, but chips are the base layer. Without affordable compute, AI growth becomes expensive.

If companies like OXMIQ can help reduce AI compute costs or support custom AI hardware, it can create opportunities for Indian engineers, data centre companies, cloud providers and government-backed AI projects.

This does not mean OXMIQ will immediately change the market. Chip development takes time. But the direction is important because AI hardware is becoming one of the most valuable parts of the technology industry.

Conclusion with key takeaways

OXMIQ Labs’ $35 million Series A funding shows how serious the race for AI compute has become. The startup is trying to build a new path by licensing GPU architecture instead of selling only finished chips.

Its biggest promise is flexibility. If OxCore and its software stack work as planned, companies and governments may get a faster way to design custom AI processors.

But the test will come in execution. OXMIQ must complete OxCore, support customer tape-outs, prove performance and build developer confidence. In the AI chip world, good ideas need strong silicon and strong software to survive.

Key takeaways

1. OXMIQ Labs has raised $35 million in Series A funding.
2. The round was co-led by Samsung Catalyst Fund and Fundomo.
3. The startup was founded in 2024 by Raja Koduri.
4. OXMIQ’s main product is OxCore, a licensable AI GPU architecture.
5. The company has engineering teams in Hyderabad and Bengaluru.
6. Its aim is to help companies and governments build custom AI chips at lower cost.
7. Main competitors include Nvidia, AMD, Intel, Tenstorrent, Cerebras, Groq and cloud companies building AI chips.

Facts Input- ET, ToI, Tom’sHardware