Ricursive Intelligence raises $35M from Sequoia to use AI for chip design and close the hardware-model co-optimization gap
Dec 19, 2025 with Anna Goldie
Key Points
- Ricursive Intelligence raises $35M from Sequoia to build AI tools that co-optimize chip architecture and model design, targeting a claimed 10x improvement in total cost of ownership.
- Current AI infrastructure relies on general-purpose silicon misaligned with specific model computational patterns, leaving what Ricursive sees as the largest untapped efficiency lever in the AI stack.
- Sequoia's lead investment signals institutional conviction that hardware-model co-design is moving from academic curiosity to commercial viability as inference costs become a primary competitive variable.
Summary
Ricursive Intelligence has raised $35 million led by Sequoia to pursue a thesis that sits at the intersection of chip design and AI model development. The company's core argument is that hardware and models are currently optimized in isolation, and that even rudimentary co-optimization, tuning chip architecture alongside the model it runs, can yield close to a 10x improvement in total cost of ownership.
The pitch is directionally simple but technically demanding. Most AI infrastructure today runs on general-purpose silicon that was not designed with a specific model's computational patterns in mind. Ricursive's position is that closing this gap represents one of the largest untapped efficiency levers in the AI stack.
The $35M round from Sequoia signals institutional conviction that hardware-model co-design is moving from academic interest to a fundable commercial category. The bet is that as model architectures stabilize and inference costs become a primary competitive variable, purpose-optimized silicon will matter significantly more than it does today.