Flourish secures $500M to build brain-inspired low-power AI

Flourish sits at the intersection of two of the race-to-AGI’s most important pressure points: energy and architecture. If the company can get anywhere close to its ambition—brain‑like AI running in the tens of watts rather than kilowatts—it would radically change the economics of both training and inference. Instead of arguing about who can afford 300‑megawatt data centers, the frontier could pivot toward who can best exploit a much cheaper, cooler substrate.

Strategically, this round shows how much capital is now flowing into alternatives to pure transformer scaling. Groq, Cerebras, and Etched attack efficiency with silicon; Flourish is betting on copying the brain’s wiring patterns. Bezos and Tier‑1 VCs effectively placed a hedge that if compute and power become the binding constraints, architectures that radically reduce flops per capability will be where the real moat lies.

For incumbents, this is a reminder that today’s LLM stack may not be the final form of AGI infrastructure. If Flourish or similar efforts succeed, the lead in models trained on GPU clusters could be undermined by a generation of systems that simply do more with less.