Archgen Labs launches Newton, an AI agent that finds and fixes chip design problems on its own

Archgen Labs, a chip design software startup founded in 2026 by Hariharan Ayappane (CEO) and Naveen Venkat (CTO), has publicly launched Newton, an AI agent for the physical design stage of making a chip. Crypto Briefing reported that the company's launch activities ran from 6 to 9 November 2026. Archgen is part of Y Combinator's Fall 2026 batch, and Inc42 picked it as one of its startups to watch for October 2026.

Physical design is the slow, manual part of chip making where engineers decide where every block sits, route the wires and fix the errors that each software run throws up. Those errors include timing violations, where a signal arrives too late to keep the chip in step, and design rule failures, where a layout breaks a foundry's manufacturing rules. According to Crypto Briefing, Newton reads these run reports, picks out the critical problems, applies fixes and then reruns the flow to check whether the fix held, carrying what it learned into later runs.

The company's bigger bet is on 3D chips, where logic, memory and other parts are stacked in layers like floors of a building. Stacking shortens the distance data has to travel, but it makes heat much harder to get out. Archgen's LayerTherm tool predicts a chip's temperature map, and the company says it runs about 5,000 times faster than conventional simulation in its own benchmarks. Its GPU based placement engine, it says, ranked first in a macro placement challenge run by Hudson River Trading and Partcl in June 2026.

Archgen says it has signed pilot projects and is working with GlobalFoundries to build 3D chips. Crypto Briefing put the target for a first tapeout, the point where a finished design is sent for manufacturing, at April 2027, while the company's Y Combinator profile says mid 2027.

Inc42 notes that the electronic design automation market the startup is entering is projected to reach $24.07 Bn by 2030, and that its real test will be fitting into the established workflows of chip teams and proving its AI can handle production grade designs. The speed figures so far are the company's own benchmarks, not independent tests.