The archive · Hardware & Devices · Strategic decision · 2025–2026
Ricursive bets AI can design its own chips: $335M raised, $4B valuation in 4 months
AlphaChip's creators left Google to bet AI can compress multi-year chip design; a $35M seed, then $300M at $4B in months.
Ricursive Intelligence
What the business is
Ricursive is a frontier AI lab building software that automates semiconductor design — from component placement through verification — so chip makers can produce custom silicon in a fraction of the time.
Starting capital：$35M seed led by Sequoia at a $750M valuation (Dec 2025); $300M Series A led by Lightspeed at $4B post-money (Jan 2026).
How it started
Anna Goldie and Azalia Mirhoseini co-created AlphaChip at Google Brain — generating chip layouts in hours that human designers take a year or more to produce — and watched it help design generations of Google's TPUs. They left Google in late 2025 and launched Ricursive on Dec 1, 2025 with a $35M seed led by Sequoia at a $750M valuation.
What happened
On Jan 26, 2026 — under two months after launch — Ricursive announced a $300M Series A at a $4B post-money valuation led by Lightspeed, joined by DST Global, NVentures (Nvidia), Felicis, 49 Palms, Radical AI and Sequoia. TechCrunch reported in February 2026 that the company had already heard from every major chip maker, was building full-stack AI design tools, and claimed roughly 10x performance per total cost of ownership for model-specific architectures.
No ending yet — it is still running.
Background
Ricursive is a frontier AI lab founded by Anna Goldie and Azalia Mirhoseini, the Google Brain researchers whose AlphaChip project generated chip layouts in hours — work that took human designers a year or more — and helped design generations of Google's Tensor Processing Units. The pair left Google in late 2025 and launched the company on Dec 1, 2025 with a $35M seed led by Sequoia at a $750M valuation.
The bet is that the multi-year, capital-intensive chip design process has become the bottleneck on AI progress, and that AI can compress it into a recursive loop: Ricursive's models design the next generation of chips, and those chips train more capable models. The stated endgame, which the founders call the 'designless' era, is AI designing its own silicon substrate, enabling a Cambrian explosion of custom chips.
On Jan 26, 2026, less than two months after launch, Ricursive announced a $300M Series A at a $4B post-money valuation led by Lightspeed, with DST Global, NVentures, Felicis, 49 Palms, Radical AI and Sequoia participating. TechCrunch reported in February 2026 that the startup was building AI tools that design chips — not the chips themselves — making chip makers such as Nvidia, AMD and Intel its target customers rather than rivals, and that early development partners had not yet been named.
What has to be true
- The bet is sharply defined: AI accelerates chip design, and chips accelerate AI, in a loop the founders say ends with AI designing its own silicon.
- The funding trajectory is unusually legible: $35M seed at $750M in December 2025, then $300M at $4B in January 2026, before any product shipped.
- The founders' AlphaChip work provides a concrete, verifiable proof point: layouts in hours and adoption across generations of Google TPU.
- It deliberately avoids the crowded 'AI chip maker' category by selling design tools to chip incumbents, with Nvidia itself as an investor.
- It is distinct from every existing library entry — no other case covers AI-driven semiconductor design automation.
What can be applied
Scarce technical pedigree lets capital price the endgame before the product: a two-month-old lab raised $335M on 'AI designs its own silicon,' with no chip shipped.
Aftermath
As of September 2, 2026, Ricursive has not announced a product release, named customers, or revenue; it says it is scaling its research and engineering team and compute infrastructure to build a platform spanning the full semiconductor design stack. Public evidence remains the founders' AlphaChip lineage and a pitch that AI-hardware co-evolution could deliver roughly 10x performance per total cost of ownership for model-specific architectures.
Sources
- How Ricursive Intelligence raised $335M at a $4B valuation in 4 months
- Ricursive Intelligence Raises $300 Million Series A at $4 Billion Valuation to Accelerate AI-Driven Semiconductor Design
- Ricursive Intelligence Launches Frontier AI Lab
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card