Jeff Bezos backs Thomas Reardon's Flourish in a $500 million neuro AI bet

Why it matters

Flourish is one of the clearest counter-bets to brute-force AI scaling: a heavily funded founder team trying to turn neuroscience into lower-power, continuously learning models.

Thomas Reardon and Rob Williams have assembled $500 million for Flourish, a New York neuro AI startup backed by Jeff Bezos (@JeffBezos), Lux Capital and Google Ventures, according to Wired, in a wager that the next AI leap will come from studying neurons rather than simply adding chips.

The bet is personal to both men. Williams left Amazon last fall after serving on Jeff Bezos' S-team and overseeing software products including Alexa, Wired reported. Reardon, a neuroscientist and repeat founder, previously built CTRL-Labs, the brain-computer interface company Wired profiled in 2017 and later sold to Meta (@Meta), where he worked for six years.

Wired reports that Williams pitched Bezos in December 2025 using the Amazon-style practice of writing a press release for a product before it exists. The pitch described Flourish as a "neuro AI company" building "Cortex AI," a synthetic intelligence system meant to match the human brain's computational capacity, learning efficiency and power budget.

Bezos initially put in $50 million after reading the two-page pitch, then "almost doubled" that stake, Wired reported. Flourish now has a $500 million war chest and a reported $2.5 billion valuation, per Wired. The story does not specify round names, ownership, closing dates or the source of the valuation figure.

The founder bet

Reardon is not pitching incremental model optimization. He told Wired the target is "a synthetic artificial intelligence brain that runs on 50 watts or less." His critique is that frontier AI has grown dependent on enormous training runs, vast datasets and power-hungry infrastructure, while the human brain processes information at roughly 20 watts.

"There's something fundamentally wrong with saying, 'I need to basically read every book ever written 20 times over in order to learn English,'" Reardon told Wired. "A human baby does it with a couple hundred thousand utterances."

That is the opening Flourish is trying to exploit: not a better chatbot wrapper, but a different theory of computation. Wired says Flourish's scientists are focusing on cortical columns, which one Flourish scientist described as "the canonical computational unit" of the brain. The company is also tied to connectomics work through Jacob Vogelstein, an investor and adviser, and Joshua T. Vogelstein, whom Wired identifies as a cofounder.

By the end of March, Reardon had hired around two dozen neuroscientists and AI researchers, according to Wired. The recruiting list includes Greg Wayne, a longtime DeepMind researcher who has led work on Project Astra, Google's AI assistant initiative. Wired reported that Wayne kept his Google job while committing 20 percent of his time to Flourish.

The hard part is still unproven

Flourish's advantage, if it has one, is that it is treating neuroscience as a product input rather than a metaphor. Wired describes a West SoHo office with a built-in data center and plans for wet-lab experimentation using advanced equipment, including electron microscopes.

But the important caveat is in Wired's own reporting: Reardon and Williams have not yet figured out how to build a system that matches the brain. Flourish plans to release nearer-term models while pursuing the larger Cortex AI ambition, but Wired does not identify launch dates, customers, revenue, benchmarks or a product beyond the Cortex AI name.

The idea also has history. IBM and Intel have both worked on neuromorphic chips, and UC Berkeley computer scientist Ben Recht, a Flourish adviser, told Wired that brain-inspired software research predates the current large language model boom. His critique of today's neural networks is blunt: "They call those neural nets, but there's nothing brain-like happening there."

That makes Flourish a test of both science and capital allocation. Bezos and the other backers are effectively funding the search for a missing abstraction: whether the brain contains design principles that can make AI learn continuously and run cheaply, or whether today's scaling laws remain the most reliable path despite their cost.