Flourish Raises $500M to Build Brain-Inspired AI With Bezos and Google Backing

--- headline: Flourish Raises 00M to Build Brain-Inspired AI With Bezos and Google Backing slug: flourish-500m-brain-inspired-ai-bezos category: business story_number: 07 date: 2026-06-09 ---

# Flourish Raises 00M to Build Brain-Inspired AI With Bezos and Google Backing

A secretive New York startup wants to reverse-engineer the human brain and run the result on 50 watts. Half a billion dollars says it might not be crazy.

Flourish Inc. emerged from stealth in early June 2026 with roughly 00 million in committed funding and a .5 billion valuation, according to reporting by WIRED, SiliconANGLE, and Hoodline. The backers form a who’s-who of the current bet-big-on-the-next-paradigm crowd: Jeff Bezos contributed close to 00 million — initially 0 million, then nearly doubled his stake as other names piled in — alongside Alphabet’s venture arm GV, Lux Capital, and healthcare-focused fund Catalio Capital. Catalio co-founder George Petrocheilos confirmed in a statement to citybiz that the firm is also listed as a co-founder and co-lead investor.

The company is co-founded by Thomas Reardon and Rob Williams. Reardon is best known as the engineer who shipped Internet Explorer at Microsoft in the mid-1990s, then pivoted hard into neuroscience — he later co-founded CTRL-labs, the brain-computer interface startup Meta acquired in 2019 for a reported figure in the vicinity of billion; its technology now underpins the Meta Neural Band wristband. Williams is a former Amazon S-team executive. Catalio managing partner Dr. R. Jacob Vogelstein and venture partner Dr. Joshua Vogelstein are also named founders, giving Flourish direct roots in Johns Hopkins’ neuroscience networks.

The product on the drawing board is called Cortex AI. The stated goal, as Reardon described it to citybiz, is building “a synthetic artificial intelligence brain that runs on 50 watts or less.” For context: the human brain operates at roughly 20 watts during active reasoning; a single server-grade AI training chip consumes more than 30 times that. Current large-model training clusters scale that gap into the megawatt range. Flourish is not trying to build a better GPU — it is trying to make the GPU irrelevant.

To get there, the company is building an in-house neuroscience wet lab, reportedly inside a West SoHo building in Manhattan that includes a built-in data center. WIRED, which visited the facility, reported that Flourish had hired roughly two dozen neuroscientists and AI researchers by the end of March. The lab will be equipped with electron microscopes — instruments that cost millions of dollars apiece and can image structures far smaller than optical microscopes allow — to pursue connectomics research, the mapping of neurons and their connections at high resolution. Flourish’s specific focus area is cortical columns: the repeating neuron structures that many researchers believe are responsible for most of the brain’s information-processing work.

The goal is to identify what Reardon calls the brain’s “core algorithm” and replicate it in silicon. WIRED reports the company expects a breakthrough within five years. In the nearer term, Flourish says it plans to release “a model that can learn continuously” for consumer devices and is in active talks with an unnamed chipmaker about co-designing a processor to run it. It is also building an AI memory management system intended to reduce the data volume required for model training.

Why it matters

Flourish’s raise lands in the middle of a distinct new wave of foundational AI bets that are skeptical of pure transformer scaling. On the same week Flourish news broke, Generalist AI closed a 00 million round at a billion valuation to pursue general intelligence for robotics; and PhysicsX, the London-based physics simulation startup, announced a 00 million Series C at a .4 billion valuation. All three are predicated on a similar thesis: that the next decade of AI progress will not come from piling on more GPU compute and data, but from fundamentally different architectures rooted in physical reality — whether that is the physics of industrial engineering, the mechanics of embodied robotic systems, or the electrochemical dynamics of the human brain.

For investors, the energy efficiency argument is increasingly not abstract. AI infrastructure power demand is already straining electrical grids in the United States, Europe, and Asia. Any architecture that can match or exceed transformer-class performance while running at a fraction of the wattage is not just a research curiosity — it is a potential wedge into every deployment context where power is constrained: edge devices, mobile hardware, data centers under regulatory pressure. Bezos and GV are not betting on connectomics because it is interesting science. They are betting that whoever cracks brain-level efficiency first will own a substantial share of whatever AI looks like in 2035.

There are substantial reasons for skepticism. Connectomics is a painstaking, expensive discipline — mapping even a small volume of brain tissue at electron-microscope resolution can take years and produce petabytes of data that are still far from fully understood. No connectomics-derived AI system has yet demonstrated capabilities that challenge frontier models. And Flourish is, at this writing, pre-product: no published models, no peer-reviewed connectomics output from the new lab, and a roadmap measured in five-year horizons. The .5 billion valuation is therefore a bet on team pedigree and thesis, not demonstrated output — which is precisely the kind of bet that can look either visionary or reckless depending on what the lab publishes over the next 24 months.

What to watch

Three signals will tell you whether Flourish is tracking toward its ambitions or burning through capital on hard science. First, watch for peer-reviewed or preprint publications from the SoHo lab — original connectomics or neuroscience findings that show the research is generating usable insights, not just expensive microscope hours. Second, watch for the chip partnership announcement; a named semiconductor company willing to co-design a processor around Cortex AI’s architecture would be strong validation of the technical roadmap. Third, watch for any early model release on consumer hardware — even a narrow demonstration of continuous learning on a device would substantiate the core efficiency claims in a way that investor backing alone cannot.

The funding is real. The problem Flourish is attacking is real. The gap between those two facts and a working brain-inspired AI system remains, for now, the largest unknown in the story.

Reporting based on WIRED, SiliconANGLE, Hoodline/citybiz, and Crunchbase News coverage from June 3–8, 2026.