Architect is an AI lab for the compute stack, starting with chip design.

We are amidst the largest infrastructure buildout in human history. AI workloads are scaling and fragmenting across data centers, new model families, and in the physical world through automotive, robotics, and edge devices.

As this diversification accelerates, the limiting factors shift to performance per watt, latency, and cost, making specialized silicon inevitable. The signal is clear: frontier AI labs, hyperscalers, and next-generation platform companies are launching custom chip initiatives to own their hardware roadmaps.

While demand for custom silicon is skyrocketing, chip design remains the structural bottleneck. Silicon projects still require multi-year timelines and hundreds of millions in upfront costs, with architectural decisions locked in far too early. Teams are forced to predict workloads years in advance, exactly when models, software stacks, and product requirements are changing the fastest.

Architect is a foundational lab for the AI compute stack, starting with accelerating chip design. We are building AI systems that can discover new computer architectures and generate novel, verified chip designs for ML workloads. Our mission is to enable true HW/SW co-design, so AI companies can deliver ultra-low latency, energy-efficient, and affordable intelligence at scale.

We do fundamental research on ML systems, new accelerator paradigms for emerging AI model families, and applying AI for silicon, from exploration, design, and verification. In parallel, we apply our research to real-world silicon programs with industry partners, co-designing real silicon and taping out on leading-edge foundry technology (TSMC, GF, Samsung).

Our founding team blends world-class chip designers, systems engineers, and frontier AI researchers, from companies like Anthropic, Meta (MSL and AI Silicon Research), Google DeepMind, xAI, Intel, Tesla, Stanford and Infineon Technologies.

If you're interested in the intersection of AI, systems, and silicon, consider joining us.