Naveen Rao

Founder of Unconventional AI, neuroscientist, and serial entrepreneur in AI hardware.

Claims

As speaker

Mentioned in

Key positions and views

True AI is unattainable without a fundamental revolution in the underlying hardware substrate; software improvements on current digital systems are insufficient.

The primary physical constraint on scaling AI is energy consumption, rendering plans for massive GPU clusters physically impossible and making energy efficiency the most critical metric.

The Transformer architecture is a computationally inefficient, O(n^2) 'brute force' simulation of a more elegant recurrent dynamical system, which a better hardware substrate could execute more directly.

Analog computing, inspired by the human brain's 20-watt efficiency, represents the most promising path to overcoming the energy and data movement bottlenecks of digital, matrix-multiplication-based hardware.

Previous attempts at neuromorphic computing failed not because the core idea was wrong, but because they were premature and lacked a clear machine learning framework for evaluating their utility.

Podcast consensus on Rao

Points of consensus

▶Fundamental hardware innovation is the primary prerequisite for achieving true artificial intelligence, as current digital substrates based on matrix multiplication are hitting physical and energy limits.

▶Energy consumption is the most critical bottleneck for scaling AI, making massive GPU build-outs physically unsustainable and positioning energy efficiency (e.g., joules per token) as a key performance metric.May 2026

▶Unconventional AI is pursuing a high-risk, paradigm-shifting strategy by developing a novel analog computing substrate, rather than making incremental improvements to existing digital architectures.

▶A full-stack solution (hardware and software) is necessary to bring a new computing paradigm to market, as simply selling a chip would be insufficient for adoption.

Points of debate

▶Rao's strategy involves building a radical new computing substrate designed to replace inefficient models like Transformers, yet pragmatically, the company's initial hardware must support these very models to gain market traction.

▶He advocates for a long-term, open-ended research culture for the first few years, which contrasts with the immense pressure of a firm five-year timeline to deliver a manufacturable product and the need to raise approximately $1.5 billion in capital.

▶Rao is highly confident in achieving a 50x performance improvement with a path to 100,000x, a speculative claim that stands in contrast to his critique of other companies' ambitious but physically unrealizable infrastructure plans.

▶He believes his past venture, Nirvana Systems, was sold to Intel 'too early,' suggesting a tension between achieving a successful exit and fully realizing a technology's long-term potential, a dynamic he now faces again with a much more ambitious project.Apr 2026

Key themes

▶The Hardware Substrate as the Final Frontier

Rao's core thesis is that true AI cannot be achieved through algorithmic improvements alone. He argues that current digital hardware is a fundamental bottleneck and that progress requires a revolutionary new computing substrate inspired by the brain's efficiency.

This positions Unconventional AI as a bet against the entire existing AI hardware ecosystem (e.g., GPUs), suggesting that investors are backing a potential paradigm shift where the physical medium of computation is as important as the algorithms themselves.

▶The Inevitable Energy Crisis of AI

Rao consistently highlights the unsustainable energy consumption of current data centers, which he claims use 4% of the U.S. grid's output. He predicts this will lead to physical limitations like power brownouts and argues that large-scale GPU build-outs are physically impossible due to power constraints.

Rao uses the impending energy constraint as both a dire market warning and the core value proposition for his company, framing extreme energy efficiency not as a feature but as a prerequisite for the future of AI.

▶High-Risk, Full-Stack Venture Building

Unlike a traditional chip company, Unconventional AI is building a full-stack solution because its new paradigm requires it. This strategy involves a long, research-focused initial phase and a massive capital requirement estimated at $1.5 billion to reach the first product.

This full-stack approach significantly increases execution risk and capital intensity compared to a component supplier, making the venture a binary bet on Rao's team to deliver a complete, integrated, and market-ready system from the ground up.

▶Revisiting Neuromorphic Computing

As a trained neuroscientist, Rao believes past attempts at neuromorphic (brain-inspired) computing were not fundamentally flawed but were simply attempted too early, before a clear framework for their evaluation existed. He draws on the stable, efficient design of the mammalian cortex as evidence of a superior computing architecture.

By framing past failures as a matter of timing rather than concept, Rao positions his work as the logical and timely successor to these efforts, aiming to succeed where others failed by leveraging modern machine learning principles.

Source episodes

Timeline

2014

Rao notes that the prevailing sentiment among venture capitalists was that hardware was a 'solved problem' and not an attractive investment area.

Pre-2016

Founded and led Nirvana Systems, an AI hardware company.

2016 (implied)

Sold Nirvana Systems to Intel, a decision he later came to believe was 'too early'.

Post-2016 to ~2023

Served as Chief AI Officer at Databricks, where he claims to have overseen the growth of the AI business from ~$20 million to $700-$800 million in revenue.

~2023/2024

Left Databricks before his full earn-out from the MosaicML acquisition was complete to found his new company, Unconventional AI.

2024

Announced Unconventional AI with $475 million in funding, with Databricks as an investor, to pursue a new brain-inspired analog computing paradigm.

Suggested prompts

How does Rao's full-stack strategy mitigate the adoption risk of a novel computing paradigm, and what new execution risks does it introduce? &nearr;Given the five-year timeline for a manufacturable product, what are the key technical and market milestones Unconventional AI must hit to justify its projected $1.5 billion capital requirement? &nearr;Rao claims his initial hardware will support existing models like Transformers while also arguing they are inefficient. How can this dual-support strategy serve as a bridge to his new paradigm without compromising its core architectural advantages? &nearr;What external factors, such as breakthroughs in digital chip efficiency or new algorithmic discoveries, pose the greatest threat to Rao's thesis that a hardware substrate revolution is necessary? &nearr;

Key concepts

AI Hardware Innovation 15 ep Energy Consumption & Constraints 7 ep Unconventional AI Strategy 12 ep Venture Capital & Funding 8 ep Neuromorphic & Analog Computing 6 ep Transformer Architecture 4 ep Semiconductor Industry 2 ep Databricks 3 ep Data Centers 4 ep

Notable quotes

“My personal take is that we need innovation on the hardware substrate to actually build true intelligence. You're not going to get there with this sort of rudimentary thing that we have.”

Naveen Rao · Inside the $4.5B Startup Building Brain-Inspired Chips for AI

“I think we're going to hit 50 X with a very high chance. If we do that and we have a path to getting to a thousand and getting to 10,000 and 100,000, now we've created something really big.”

Naveen Rao · Inside the $4.5B Startup Building Brain-Inspired Chips for AI

“you're talking about you know a trillion dollars of gpus where the are you going to get the power for that i'm sorry right like you're just you're not going to be able to build these things physically”

Naveen Rao · Inside the $4.5B Startup Building Brain-Inspired Chips for AI

“I mean, I think when I left, like when I came there, we had about a $20 million business, and it was, you know, $700 or $800 million when I left, something like that.”

Naveen Rao · Inside the $4.5B Startup Building Brain-Inspired Chips for AI

Report last updated: Apr 21, 2026

Get started free

Back to Entities Intelligence Report