
TLDR; We’re building the tools to out more transistors on a chip. Our foundational models run roughly 5,000× faster than current EDA software and our core placement engine ranked is #1 globally in Hudson River Trading’s macro-placement challenge.
https://www.youtube.com/watch?v=39z8-MXMiOI
Hey everyone! We’re Hariharan Ayappane and Naveen Venkat, the founders of ArchGen Labs (YC F26).
For decades, better chips came from making transistors smaller. That’s getting harder and more expensive, while AI keeps demanding more compute within the same power and cooling budgets.
Part of the problem is distance. Moving data between components takes time and energy. Stacking computation vertically, like floors in a skyscraper, can bring those components closer together.
But shorter connections come with a trade-off: more computation in the same space makes heat harder to remove.
Engineers need to optimize placement, timing, and temperature together. Every change needs feedback. When those checks take too long, teams explore fewer designs.
We accelerate that loop.
1. Ultrafast foundational models for chip design
LayerTherm uses physics-informed neural networks to predict a chip’s temperature map from its power distribution.
In our benchmarks, predictions averaged 62 milliseconds versus 314 seconds for EDA solvers - roughly 5,000× faster.
That lets engineers screen more candidate designs for hotspots before detailed thermal verification.
2. Worlds best chip design engine
Our GPU-accelerated tools help engineers decide which components belong on each layer, shorten critical connections, and evaluate timing and thermal trade-offs.
The core placement engine ranked #1 globally in Hudson River Trading’s macro-placement challenge: https://archgenlabs.com/blog/posts/how-we-ranked-first-in-the-partcl.html
3. Fully autonomous AI chip design engineer
Newton works through placement, routing, design-rule violations, and timing repairs. It tests candidate changes, rejects weak ones before expensive checks, and carries forward verified results.
We’ve signed pilot projects and are working with GlobalFoundries to build 3D chips. We’re targeting our first tapeout in mid-2027.
We’d love to meet:
If that sounds like your team, tell us where your design process gets stuck. We’d love to explore it with you.
Reach us at founders@archgen.tech