Flourish raised $500M at a $2.5B valuation led by Jeff Bezos (~$100M), Lux Capital, GV, and Catalio. That's the short answer. The longer answer is more interesting.
In early June 2026, a roughly five-week fundraise closed for a New York–based startup that, until last week, almost nobody outside neuroscience research had heard of. Flourish is building Cortex AI — a model architecture inspired by how the brain processes information — and its pitch is that the entire frontier AI stack is solving the wrong problem. While OpenAI, Anthropic, xAI, and Google race to scale model size and GPU footprint, Flourish is racing to shrink the architecture itself.
The investors who showed up tell you most of what you need to know. Bezos started at ~$50M and ended at ~$100M. Lux and GV co-led. Catalio, the bio/neuro specialist whose managing partners are credited as co-founders, brought the connectomics research backbone. Five weeks, $500M, $2.5B post-money, pre-product. That doesn't happen unless multiple decision-makers believe this is one of the bets that defines the next decade.
Flourish $500M Funding: Round Terms, Lead Investors, and the Brain-Inspired AI Pitch
Flourish closed $500 million at a $2.5 billion post-money valuation in early June 2026, anchored by Jeff Bezos (~$100M), with Lux Capital, GV (Alphabet's venture arm), and Catalio Capital as the other lead investors. The round funds Cortex AI — a brain-inspired model architecture that targets 20–50 watts of operating power, against the megawatt-scale footprint of frontier transformer models. Co-founders Thomas Reardon (Internet Explorer; CTRL-labs, sold to Meta for ~$1B) and Rob Williams (ex-Amazon S-team) are building from Soho with the Catalio-affiliated Vogelstein neuroscience team.
The Numbers Behind the Round
$500M
Total round size
Closed in roughly five weeks from first conversations in late April
$2.5B
Post-money valuation
Pre-product — comparable to early-stage frontier model labs
~$100M
Bezos personal check
Roughly doubled from his initial ~$50M commitment
20–50W
Cortex AI power target
Versus megawatts for today's frontier model training and inference
Why the Founder Pair Is the Real Story
If you strip out the architecture pitch and just look at the cap table, this is a bet on Thomas Reardon. That's a name most software people know without knowing they know it. He led the team that built Internet Explorer at Microsoft in the mid-1990s — the product that ended the first browser war and shipped on every Windows PC for two decades.
Then he went back to school. Reardon got a neuroscience PhD from Duke and Columbia and co-founded CTRL-labs in 2015, building a wristband that read motor neuron signals from the forearm. Meta acquired CTRL-labs in 2019 for an estimated $1 billion. The technology became the foundation of the Neural Band that Meta is now shipping with its Orion AR glasses. Reardon directed neuromotor work at Meta Reality Labs until founding Flourish.
His co-founder Rob Williams is a former Amazon S-team executive — the small group of senior leaders who report directly to the CEO. That's the network bridge to Bezos, and it's why a $50M check from Bezos became $100M when the rest of the syndicate joined.
The Catalio Vogelstein brothers (Jacob, managing partner; Joshua, venture partner — both Johns Hopkins–affiliated connectomics researchers) are credited as co-founders, which is what gives Flourish its technical depth in the part of neuroscience that actually maps neuronal connections at scale.
Why the Investor Syndicate Matters
| Investor | Role | Signal |
|---|---|---|
| Jeff Bezos (personal) | Anchor — ~$100M | Largest single check. Doubled from $50M when syndicate filled in. Bezos rarely leads pre-product AI rounds personally. |
| Lux Capital | Co-lead | Deep-tech specialist. Has been writing checks into hard-science neuro and physics-AI plays since pre-ChatGPT — pattern-match buyer for this thesis. |
| GV (Google Ventures) | Co-lead | Strategic exposure for Alphabet without DeepMind acquiring outright. GV's biology/neuro bench is among the strongest in venture. |
| Catalio Capital | Co-founder + co-lead | Unusual structure — partners are listed as company co-founders. Provides the connectomics research IP foundation tied to Johns Hopkins. |
The Power Crisis Thesis
The reason a pre-product startup gets a $2.5B mark in five weeks is that the problem is now blindingly visible. AI's electricity demand is the binding constraint on frontier AI deployment in 2026.
Microsoft signed a deal with Constellation Energy to restart Three Mile Island Unit 1 to power its data centers — first time a US reactor has been restarted from shutdown.
Amazon, Google, and Meta have signed nuclear PPAs totaling more than 25 gigawatts of contracted capacity, much of it for AI training and inference loads.
Frontier model training runs now consume hundreds of megawatts per cluster. Inference at hyperscale is the new bigger bill — and growing faster than training.
Industry analysts now project AI will account for roughly 8–12% of total US electricity consumption by 2030, up from under 2% in 2023.
The frontier labs are solving this from the supply side — more reactors, more transmission, more cooling. Flourish's pitch is the opposite: solve it from the demand side by changing the math at the architecture level. The brain does multi-modal reasoning, generalization, abstraction, and continuous learning from minimal examples — all on roughly 20 watts. If you can model that, you don't need a gigawatt to run a frontier system. You need a wall outlet. For more on the supply-side response, see our deep dive on AI data center power demand.
How This Compares to Other Pre-Product AI Bets
| Company | Round at pre-product / pre-revenue | Lead investors |
|---|---|---|
| Flourish (2026) | $500M at $2.5B post | Bezos, Lux, GV, Catalio |
| Safe Superintelligence (2024) | $1B at $5B | a16z, Sequoia, DST, SV Angel |
| xAI (2023 seed) | $135M at $24B mark | Musk-led, founder syndicate |
| Anthropic (2021 Series A) | $124M at ~$1B | Skype founder Jaan Tallinn, Dustin Moskovitz |
| OpenAI (2019 Microsoft round) | $1B investment at undisclosed mark | Microsoft |
Compared to its peer set, Flourish is priced rationally. The valuation is a fraction of SSI's $5B founding mark, well under xAI's seed, and roughly in line with where Anthropic was priced at Series A. For the level of founder pedigree, technical IP, and the size of the addressable problem, $2.5B is the price of admission, not the ceiling. For context on where the rest of frontier AI is trading, see our breakdown of AI startup valuation multiples.
What Could Go Wrong
The same things that have historically gone wrong in this space. Brain-inspired AI is one of the oldest unfulfilled promises in computer science — neural networks themselves were named after it. Neuromorphic computing has been a research line at Intel (Loihi), IBM (TrueNorth), and Qualcomm (Zeroth) for more than a decade with no breakout commercial product.
Connectomics — the mapping of biological neural connections at scale — is also still very early. The largest fully mapped brain connectome to date is a fruit fly's, completed in 2024. A full human cortical map is decades away, and Flourish's thesis depends on extracting the right algorithmic abstractions from a partial map before that work is finished.
The other risk is timing. Even if the architecture works, transformers don't stand still. NVIDIA's Blackwell Ultra and AMD's MI400 line are each cutting power-per-token by roughly 40–60% per generation. By the time Cortex AI is production-ready, the efficiency gap Flourish is trying to close may already be much smaller — see our analysis of AMD vs NVIDIA for AI training for where that race stands today.
What to Watch Next
First public benchmark
Flourish hasn't shown a model. The first leaked benchmark — even a narrow domain comparison against a small transformer at matched parameter count — will be the moment the architecture either gets validated or written off.
Hardware partner choice
A brain-inspired architecture can't run efficiently on standard GPU tensor cores. Watch for a partnership announcement with Cerebras, Groq, SambaNova, or a custom chip play. The choice signals how far from standard the architecture really is.
Talent gravity
$500M in NYC with Reardon's reputation should pull senior researchers out of DeepMind, Meta FAIR, and the IBM neuro group. The first three named hires will tell you whether the team is real or aspirational.
Follow-on syndicate
Bezos doubling in mid-round is a tell. If the Series B follows quickly with a16z, Sequoia, or Founders Fund coming in at a step-up valuation, the architecture story is being underwritten by the consensus mid-market.
The Wider Read on June 2026
Flourish closing is part of a pattern. The first week of June 2026 saw four separate $500M+ rounds: Ramp's $750M Series F at $44B, Supabase's $500M Series F at $10.5B, Impulse Space's $500M Series D, and Flourish. Helion's $465M Series G at $15.5B and AlphaSense's $350M at $7.5B came in the same window.
What links them: every one of these companies sells into the same problem from a different angle. Ramp is the cost-control layer for AI spend. Supabase is the database layer for AI-generated apps. Impulse is the propulsion stack for space-based compute and sensing. Helion is the fusion power that AI eventually needs. And Flourish is the architectural rewrite that would make the rest of it unnecessary. The capital is flowing to anyone with a credible answer to the same underlying question — how do we keep paying for AI when the inputs are scaling faster than the economics support?
For more on what's driving AI infrastructure valuations and the power constraint, see The Hidden Cost of AI, AI Data Center Power Demand, and the AI Valuations dashboard at Value Add VC.