Thomas Reardon, who created Internet Explorer and sold neural interface startup CTRL-labs to Meta for up to $1 billion, is raising approximately $500 million for Flourish at a $2.5 billion valuation. The startup uses connectomics and neuroscience to design AI architectures that are radically more energy-efficient, targeting the architectural layer above silicon rather than building new chips. Lux Capital and GV, who backed CTRL-labs, are expected to lead the round.TL;DR
Thomas Reardon has a pattern. In 1994, he created the project that became Internet Explorer, the browser that turned Microsoft from a software company into an internet company and triggered the most consequential antitrust case in technology history. In 2015, he co-founded CTRL-labs, a neural interface startup that built a wristband capable of translating electrical signals from the brain into computer commands, and sold it to Meta for somewhere between $500 million and $1 billion. In between, he earned a PhD in neuroscience from Columbia University, served as a founding board member of the World Wide Web Consortium, and delivered the first implementation of CSS in a web browser. Now Reardon is raising approximately $500 million for Flourish, a startup that wants to make artificial intelligence radically less power-hungry by borrowing design principles from the one system that already runs intelligence efficiently: the human brain. Lux Capital and GV, the same investors who led CTRL-labs’ Series A nearly a decade ago, are expected to lead the round at a $2.5 billion valuation, according to Bloomberg. The company has no commercial product. What it has is a thesis, a team of neuroscientists, and a founder whose career suggests he is worth betting on before the product exists.
The AI industry’s power consumption is approaching the scale of a national economy. In 2022, data centres, AI workloads, and related infrastructure consumed approximately 460 terawatt-hours of electricity globally. The International Energy Agency projects that figure will nearly double by the end of 2026, and that data centres alone could account for 3 per cent of global electricity consumption by 2030, roughly twice Germany’s current total usage. In the United States, data centres consumed more than 4 per cent of the country’s total electricity in 2023, a figure projected to reach 9 per cent by 2030. The technology companies driving this demand are responding with supply-side solutions: startups are racing to curb data centre energy use, Meta is exploring space-based solar panels, Microsoft has signed nuclear power agreements, and Amazon is buying entire wind farms. The collective capital expenditure on AI infrastructure from Alphabet, Amazon, Meta, and Microsoft alone exceeds $650 billion in 2026 guidance. The industry’s answer to the question of how to power AI has been, overwhelmingly, to build more power.
Reardon’s argument is that the answer is wrong, or at least incomplete. The human brain performs tasks that no AI system can match, including generalisation, abstraction, and continuous learning from minimal data, while consuming approximately 20 watts of power, roughly the same as a couple of LED light bulbs. A single training run of a frontier AI model can consume megawatts over weeks or months. The gap between biological and artificial intelligence in energy efficiency is not a factor of two or ten. It is a factor of thousands. Flourish’s thesis is that this gap exists not because of hardware limitations but because of architectural ones: the way AI models are structured, the way they process information, and the way they learn bear little resemblance to how biological neural networks operate. The IEA has suggested that AI itself could help reduce energy consumption through grid optimisation and efficiency improvements, but Flourish is pursuing something more fundamental: changing the design of the AI systems themselves.
Flourish brings together AI researchers and neuroscientists to study connectomics, the mapping and analysis of neural connections in biological brains. The field has produced increasingly detailed maps of how neurons connect, communicate, and process information in organisms from fruit flies to mice, and those maps reveal architectural principles that differ sharply from the transformer-based models that dominate modern AI. Biological neural networks are sparse, meaning most neurons are not connected to most other neurons, and they process information asynchronously, firing only when a threshold is reached rather than performing computation on every input at every step. They are also extraordinarily efficient at routing information, using local processing and hierarchical abstraction to avoid the brute-force computation that characterises current AI architectures.
Reardon’s approach is to study these biological principles and translate them into new AI algorithms and architectures, not into new chips. This is a meaningful distinction. The neuromorphic computing industry, led by Intel’s Loihi programme and a growing ecosystem of startups, has focused primarily on building specialised hardware that mimics neural behaviour in silicon. Flourish is working at a different layer: the software and mathematical structures that define how an AI model learns and reasons. The company’s bet is that architecture improvements guided by neuroscience can deliver efficiency gains that hardware improvements alone cannot, and that those gains can be applied to existing and future chip platforms rather than requiring entirely new fabrication processes. Anthropic has explored building its own custom AI chips to optimise performance and cost, and other frontier labs are similarly investing in hardware. Flourish is arguing that the problem sits one layer above the silicon.
The $2.5 billion valuation for a company with no product is a bet on Reardon’s track record and the conviction of the investors who have backed him before. Lux Capital, the New York-based venture firm that invests in what it calls “sci-fi” technologies, has worked with Reardon since CTRL-labs. Josh Wolfe, Lux’s co-founder, led CTRL-labs’ funding and has been a public advocate for Reardon’s ability to operate at the intersection of neuroscience and commercial technology. GV, Alphabet’s venture arm, has similarly maintained its relationship with Reardon across multiple ventures. The fact that both firms are returning for a round of this size suggests that Flourish is not a speculative bet on a new market but a specific bet on a founder they believe has demonstrated the ability to take esoteric neuroscience research and turn it into technology that large companies will pay to acquire or license.
Reardon’s biography supports that reading. His career is unusual not because of any single accomplishment but because of the range across which he has operated. He created Internet Explorer at 25, which is a product management and engineering achievement. He then left Microsoft, completed a PhD in computational neuroscience at Columbia, and co-founded a company that built a working brain-computer interface that read neural signals from the wrist, a scientific and engineering achievement of a different order. Meta’s acquisition of CTRL-labs was driven by the technology’s potential to become the input device for augmented reality glasses, and Reardon spent several years at Meta leading the Neural Band programme within Reality Labs before departing to become a venture partner at Lux and eventually launch Flourish. The throughline is a person who has repeatedly identified a hard technical problem, assembled a team capable of solving it, and delivered a result that a major technology company considered worth acquiring. Whether that pattern will hold for an even harder problem, making AI itself more efficient, is what $500 million is about to test.
Flourish enters a market in which the demand for AI efficiency is no longer theoretical. Every major cloud provider is constrained by power availability: Amazon, Google, Meta, and Microsoft have all disclosed that the pace of their AI infrastructure buildouts is limited by access to electricity, not by access to capital or chips. Proposals as extreme as building data centres in space have gained serious attention precisely because terrestrial power grids cannot scale fast enough to meet projected demand. The US utilities sector is planning to spend $1.4 trillion by 2030 on infrastructure upgrades driven largely by data centre demand. Any technology that can meaningfully reduce the power consumption of AI workloads would be worth billions not because of what it does in isolation but because of the infrastructure spending it displaces. If Flourish’s brain-inspired architectures can deliver even a 50 per cent reduction in the energy required for inference, the downstream savings for a hyperscaler running millions of GPU-hours per day would be measured in hundreds of millions of dollars annually.
The risk is that Reardon is selling a thesis at a frontier-lab valuation. Two and a half billion dollars for a company that has not demonstrated a working model, a benchmark, or a product places Flourish in the same valuation territory as companies with thousands of enterprise customers and billions in revenue. The investors are pricing the founder, the team, and the addressable problem rather than any demonstrated result. That is not unusual in the current AI funding environment, where startups tackling AI’s energy demands are attracting capital at a pace that reflects the urgency of the problem more than the maturity of the solutions. But it means Flourish will need to produce results that justify a valuation built entirely on conviction. Reardon has spent his career moving between fields that do not normally speak to each other: browser engineering, web standards, neuroscience, brain-computer interfaces, and now AI architecture. The bet is that someone who has operated successfully at each of those intersections can operate at the newest one. The human brain does run on 20 watts. Whether that fact can be translated into software that runs on anything close to it is the question that $500 million and a career’s worth of credibility are now staked on.
Get the most important tech news in your inbox each week.