Architect Labs

We build new silicon end-to-end using AI.

At Architect Labs, we are on a mission to create different shapes of silicon for the abundance of intelligence. We believe the next frontier of AI will depend on fundamentally better chips, co-designed for increasingly heterogeneous and creative workloads. Achieving this requires us to not only build radically competitive silicon products, but also reimagine how they are discovered, designed and produced at scale.

Exploded view of the Redwood accelerator: lid, compute mesh, frame, pin grid, and board

Every workload that matters deserves its own chip.

AI is splitting compute into specialized silicon. Data centers are moving to workload-specific accelerators. Robots, drones, autonomous systems, personal devices, and edge hardware each need their own architectures, and each is growing into a family of designs instead of a single chip. The world needs far more chip designs than it can produce today, and the gap widens every year.

Rethink the chip design flow, formally

The industry reports large AI speedups on individual tasks: 3x in design space exploration, 10x in verification, 400x in UVM environment generation. Program results have gone the other way. First-silicon success fell from roughly 30% in 2020 to 14% in 2024, a 20-year low, and 75% of programs now run behind schedule. Speeding up single steps of a sequential flow has not moved the two numbers that matter: time from specification to silicon, and first-tapeout success. We built an AI system that treats a high-level specification as the single source of truth. From it, the system co-designs and optimizes the architecture, RTL, verification environments, formal proofs, firmware, and software together. Human architects set the intent and direction.

We work in three ways:

  1. With semiconductor companies, to deliver their most ambitious silicon programs faster.
  2. With workload companies (software companies, AI labs, neoclouds, and physical AI teams), to co-design production-ready custom silicon.
  3. Bring radically new and competitive silicon products to market across different stacks of intelligence.

The Founding Team

Our team brings together researchers and engineers from frontier AI, systems, and advanced semiconductor design, including a former Senior Director of Intel's Data Center AI group, the former Head of Trust and Safety ML at Anthropic, and veterans of Apple, Tesla, Meta, Synopsys, Qualcomm, and Samsung. Collectively, we have taped out over 100 production chips, led $20B product lines in major silicon programs, and advanced AI research at nearly every frontier AI lab.

Advisors & Angels

We are grateful for our early supporters who help shape our research directions and company.

  • Lukasz Kaiser

    Co-author of "Attention Is All You Need", OpenAI

  • Srinivas Narayanan

    Former CTO of B2B, OpenAI

  • Sunil Shenoy

    Former Senior Vice President at Intel

  • Trevor Blackwell

    OpenAI founding team member

  • Aravind Srinivas

    Co-founder & CEO, Perplexity AI

  • Kunle Olukotun

    Founder, SambaNova

  • Alex Wissner-Gross

    Computer scientist, MIT and Harvard

  • Prof. Thierry Tambe

    Stanford

  • Siddhartha Nath

    AI for hardware, GDM

  • Arash Ferdowsi

    Co-founder, Dropbox

Backed by