Introducing Architect Labs

We are in the largest infrastructure buildout in human history. Intelligence is scaling and fragmenting across gigawatt-scale data centers, new model families, and the physical world of robotics, autonomous systems, defense, industrial automation, and personalized, local AI. Every layer of the stack is forced to be rebuilt — compute, memory, networking, storage, interconnects, sensors, and packaging.

At the same time, the global impact of AI’s promise is still in its nascent phases. Humanoids have yet to set foot in a single household. Only 0.001% of daily global miles are driven autonomously. We still walk around with phones in hand, not smart glasses and wearables. Sub-quadratic and post-transformer models have yet to prove themselves. “Personal” intelligence is still owned in the cloud, not by the user.

Despite how early we are, GPT-3 showed us how quickly AI demand scales when a new technology transitions from a research preview to worldwide feasibility. All of these workloads will reach their own nonlinear inflection points — and when they do, they will demand an unimaginable amount of compute.

Today, we see frontier AI labs, hyperscalers, OEMs, and fabless chip companies race to spin up new silicon programs. However, custom silicon is one of the slowest, most expensive, and highest-risk bets a company can make. The time and cost it takes to build leading silicon is rising. Supply side constraints force new architectures every year. Teams are forced to predict workloads years in advance, exactly when models, software, and product requirements are changing the fastest.

It is increasingly clear that the world will continue to need more chips than we can design today.

Architect Labs is a foundational AI lab for computing systems. We are starting with chip design, where our AI enables true HW/ML co-design, to evolve models, software, and hardware together. By aggressively specializing and increasing intelligence-per-watt, we’ll allow every workload to scale exponentially, towards a path of global superintelligence.

A generation ago, the fabless revolution in the semiconductor industry separated chip design from manufacturing, and unleashed a wave of "fabless" companies, like NVIDIA, Broadcom, and Apple, that built silicon without ever owning the foundry.

We're doing it again, but one layer up. There will be an exponential increase in the amount of silicon in the next years, and so we're designing AI systems that can design and verify that hardware end-to-end. With our technology, we:

Our mission is to enable the era of the "designless" semiconductor industry, by being a new-age design partner for both hardware and software companies. In doing so, we aim to democratize intelligence by making it faster, cheaper, and drastically more power efficient. In the end state, we'll help grow superintelligence and compute of all forms at a global scale — from data center accelerators and networking, to widespread humanoid robotics and self-driving, to personal devices and on-demand AI.

Every chip company operates on a rigid and proprietary methodology, called a "flow", that has been set in stone for decades. As a result, the industry is fragmented, and the current approach is to bandage broken workflows with a patchwork of tools and agents. Unsurprisingly, this hasn't worked well, and while models have scaled in capability exponentially, the same is not true for tapeout timelines and costs.

At Architect Labs, we are taking a differentiated approach. We are not trying to build AI agents for a fragmented process. Rather, we are crafting a completely new flow from scratch, where AI is a first-class actor, and building a rich ecosystem of models, harnesses, and tools around it.

We are not in the business of selling EDA tools or agents to chip companies. Instead, we work closely with hardware and software companies as a new-age design partner. Our innovation at the methodology level is what enables us to deliver real step-functions of improvement in design throughput, latency, and economics.

We are backed by $24M in seed funding led by Kindred Ventures, with participation from Race Capital, TQ Ventures, and Together Fund. And we’re grateful for the support of our angels and advisors, including Srinivas Narayanan, Kunle Olukotun, Lukasz Kaiser, Aravind Srinivas, Trevor Blackwell, Dr. Alex Wissner-Gross, Siddhartha Nath, Professor Thierry Tambe, Arash Ferdowsi, and engineering leaders from OpenAI, NVIDIA, Google DeepMind, Synopsys, Keysight, and more.

We're already deployed with semiconductor companies, accelerating their chip programs, with our AI-generated designs intended to tape out on leading-edge foundry nodes later this year.

Our founding team has collectively taped out 80+ production chips, led $10B data center product lines, been core contributors to Meta’s AI silicon, architected and designed one of the first neuromorphic chips out of Intel, led research teams at Anthropic, xAI, and Google DeepMind, and contributed to fundamental AI work across nearly every frontier lab.