Raja Koduri’s AI startup Oxmiq raises $35M to build AI chip architecture that cuts the cost of custom AI silicon
Former Intel architect Raja Koduri’s AI startup Oxmiq raises $35 million to reshape custom AI chip design and lower the cost of AI infrastructure.
Building a new AI chip has become one of the most expensive bets in technology. It can take years of engineering work and hundreds of millions of dollars before a single processor reaches production. Oxmiq, the AI chip startup founded by former Intel graphics chief architect and AMD executive Raja Koduri, thinks that the process can be far simpler.
On Wednesday, the company announced a $35 million funding round to develop a licensable AI chip architecture that promises to reduce both the cost of designing custom AI silicon and the expense of running AI applications.
The new financing brings Oxmiq’s total funding to $60 million. Samsung Catalyst Fund and Fudomo co-led the round, joined by investors including MediaTek and Pegatron Venture Capital.
Rather than building a complete chip from scratch for every customer, Oxmiq is developing intellectual property that semiconductor companies and AI system builders can license. The goal is to shorten development cycles and lower the barriers for companies building AI hardware at a time when demand for custom silicon is accelerating across cloud providers, enterprises, and AI infrastructure vendors.
Oxmiq plans to use the fresh capital to complete the first version of its chip architecture and software platform, then bring the technology to market. The company plans to expand its engineering team as development moves forward.
“The funding will scale OxCoreTM, OXMIQ’s licensable GPU architecture that allows semiconductor companies and AI system builders to build custom AI silicon without a full chip program,” OXMIQ said in a blog post.
Why Oxmiq wants to become the Arm of the AI era
Today’s AI hardware typically separates workloads across graphics processors, central processors, and tensor accelerators. Oxmiq wants to merge those three elements into a single licensable architecture that chipmakers can use as the foundation for custom AI processors.
Koduri believes the model could play a role similar to the one Arm established in the smartphone industry, in which companies license processor designs rather than creating every component internally.
“We would want to be the Arm of this next era,” Koduri said, referring to the U.K. company that supplies processor designs used in nearly every smartphone worldwide.
The company is pursuing a broader hardware strategy beyond processor cores. Its roadmap includes a computing fabric that combines chiplets and memory into a single package, an approach that has become increasingly attractive as chipmakers look for alternatives to building ever-larger monolithic processors.

OXMIQ Technology Stack (Courtesy: OXMIQ)
Oxmiq’s ambitions extend into the growing market for custom AI chips, where companies including Broadcom, Marvell, and MediaTek help hyperscalers and enterprise customers build processors tailored to their own AI workloads. Demand for application-specific AI silicon has surged as cloud providers seek to reduce their dependence on general-purpose GPUs and improve inference performance.
Samsung Catalyst Fund sees Oxmiq’s approach as a way to address that shift.
“We are very excited to co-lead OXMIQ’s financing round and back Raja Koduri and the strong team at OXMIQ. OXMIQ’s novel AI core and software platform enable heterogeneous compute for efficient, custom inference solutions serving large-scale agentic workloads.”
The investment reflects a larger trend across the AI infrastructure market. AI models continue to grow in size, inference workloads are climbing, and companies are looking beyond off-the-shelf processors for lower-cost, more specialized hardware. By licensing chip architecture rather than selling finished chips, Oxmiq is betting that the next wave of AI infrastructure will resemble the licensing model that helped Arm become one of the semiconductor industry’s most influential companies.


