TechnicalOpinion

Naveen Rao: 4D Computing, AI's Energy Wall & Beating Biology

Naveen Rao, CEO of Unconventional AI, argues that AI's exponential growth will hit an energy wall within three years, and proposes a revolutionary approach to computing based on biological principles. His company has built the first physical dynamical computer that achieves orders of magnitude better energy efficiency than GPUs by eliminating the traditional von Neumann architecture's separation of memory and compute.

Summary

Naveen Rao presents his journey from childhood programmer to serial entrepreneur, highlighting his founding of Nirvana Systems (the first AI chip company in 2014, sold to Intel) and subsequent work platformizing GPUs at Databricks before starting Unconventional AI. He establishes the urgency of the energy problem by demonstrating that Google alone consumes 12 gigawatts monthly for AI services—roughly 12% of total global data center energy—and projects that at current growth rates, energy availability will become the limiting factor within three years.

Rao argues that the fundamental inefficiency in modern computers stems from the von Neumann architecture's design, which moves information between separate memory and compute units. He contrasts this with biological systems: the human brain operates on 20 watts, a monkey's brain on 1 watt, and a squirrel's brain on 8 milliwatts, yet performs complex intelligent tasks with remarkable precision. Modern GPUs move 30 trillion bits per second in and out of memory, compared to the human cortex's 16 billion bits per second.

The solution Rao proposes is based on dynamical systems theory—using oscillators and physical principles similar to synchronized metronomes to perform computation. His company released UNO, an image generation model built on oscillators, and has now taped out the first physical dynamical computer chip, which generates images at 500 nanojoules per image compared to millijoules for GPUs—a difference of several orders of magnitude. This represents a proof-of-concept that computation and memory can be unified in a single physical element.

Rao introduces the concept of "4D computing," which uses time as a computational dimension alongside the three physical dimensions of die stacking. He also discusses the discovery of sparsity benefits: removing connections in the system can paradoxically improve trainability and performance while reducing energy consumption. He projects reaching thermodynamic limits within 3.5 years and aims to eventually beat biological efficiency, enabling distributed computing and robotic applications. The team spans theorists with PhDs in mathematics and neuroscience, chip architects, and circuit designers, with Python-based libraries enabling communication between these diverse disciplines.

About this episode

<p dir="ltr">(0:00) Welcome Naveen Rao</p> <p dir="ltr">(4:45) Is energy really the problem? The cost of a token, power contracts & the gap to close</p> <p dir="ltr">(10:24) Cutting out the middleman: abstractions, dynamical systems & a new kind of machine</p> <p dir="ltr">(19:46) Chamath joins: the path to product, porting existing models & building the team</p> <p><strong> </strong></p> <p dir="ltr">Thanks to our partners for making this possible!</p> <p dir="ltr">IREN is a vertically integrated AI Cloud platform, delivering data centers, compute and software for AI training and inference. <a href="https://iren.com/">https://iren.com/</a></p> <p dir="ltr">Oracle connects the data, applications, and infrastructure that turn AI into business outcomes—with the flexibility, choice, and control to optimize as AI evolves. http://oracle.com/ai</p> <p dir="ltr">EY helps tech innovators scale from startup to exit to megacap. You build the future. We'll handle the rest. <a href="http://www.ey.com">http://www.ey.com</a></p> <p dir="ltr">Meta believes the future is for everyone. We're focused on giving every person the tools to reach their full potential and making sure the benefits of technology are distributed to all. <a href="http://www.meta.com">http://www.meta.com</a></p> <p dir="ltr">Keel Infrastructure owns the power, land, and connectivity that HPC and AI run on - backed by secured energy assets and established grid interconnections across North America. <a href="https://keelinfra.com/">https://keelinfra.com/</a></p> <p dir="ltr">Airwallex - Agentic Global Business Accounts. Open local accounts in 70+ countries to accept payments, earn yield, pay globally, and manage spend. <a href="http://airwallex.com">http://airwallex.com</a></p> <p dir="ltr">PayPal has been revolutionizing commerce globally for more than 25 years. Creating innovative experiences that make moving money, selling, and shopping simple, personalized, and secure, PayPal empowers consumers and businesses in approximately 200 markets to join and thrive in the global economy. For more information, visit <a href="https://www.paypal.com">https://www.paypal.com</a></p> <p dir="ltr">Google for Startups connects founders with the right people, products, and best practices to help startups build faster and go further. <a href="https://startup.google.com/">https://startup.google.com/</a></p> <p dir="ltr">Explore ideas, industries, and technologies worth understanding with Chamath every week on Learn with Me: <a href="https://research.socialcapital.com/allin">https://research.socialcapital.com/allin</a></p> <p><strong> </strong></p> <p dir="ltr">Follow the besties: </p> <p dir="ltr">https://x.com/chamath</p> <p dir="ltr">https://x.com/Jason</p> <p dir="ltr">https://x.com/DavidSacks</p> <p dir="ltr">https://x.com/friedberg</p> <p><strong> </strong></p> <p dir="ltr">Follow on X:</p> <p dir="ltr">https://x.com/theallinpod</p> <p><strong> </strong></p> <p dir="ltr">Follow on Instagram:</p> <p dir="ltr">https://www.instagram.com/theallinpod</p> <p><strong> </strong></p> <p dir="ltr">Follow on TikTok:</p> <p dir="ltr">https://www.tiktok.com/@allin</p> <p><strong> </strong></p> <p dir="ltr">Follow on LinkedIn: </p> <p dir="ltr">https://www.linkedin.com/company/allinpod</p> <p><strong> </strong></p> <p dir="ltr">Intro Music Credit:</p> <p dir="ltr">https://rb.gy/tppkzl</p> <p dir="ltr">https://x.com/yung_spielburg</p> <p><strong> </strong></p> <p dir="ltr">#allin #tech #news</p> <p> </p>

Key Insights

  • Rao claims that energy has become the primary constraint in data center design, shifting from floor space and networking equipment concerns to energy contracts being the first consideration, with energy representing 50% of the cost to serve a token in AI inference.
  • The speaker argues that computers since 1945 have been optimized for speed rather than efficiency, using a von Neumann architecture that inherently requires massive information movement between separated memory and compute units, which is fundamentally inefficient for intelligence tasks.
  • Rao demonstrates that biological brains achieve intelligent behavior with vastly lower energy consumption than artificial systems by using physics-based computation rather than digital abstractions, with the human cortex using 125x fewer bit movements than GPUs while performing comparable tasks.
  • The company discovered that removing connections in a dynamical system can simultaneously improve efficiency, scalability, and performance—an unexpected alignment of typically competing optimization goals that Rao describes as a 'holy grail' problem solved by reframing the computational approach.
  • Rao predicts that a 1000x efficiency improvement would trigger Jevons Paradox, creating the largest market humanity has ever seen, as demand would increase more than proportionally to the cost reduction, fundamentally shifting from centralized gigawatt-scale data centers to distributed small data centers.

Topics

Energy efficiency in AI computingVon Neumann architecture limitationsDynamical systems and biological computation4D computing and chip designAI infrastructure and data center economicsSparsity in neural networksThermodynamic limits of computation

Transcript

Naveen Rao, co-founder and CEO of Unconventional AI, which is an AI chip startup. Best known for building and selling two deep tech companies. Naveen is kind of definitionally outlier founder. When I came there, we had about a $20 million business and it was 700 or 800 million when I left. I don't think you really understand something until you can build it. Just because something is tried does not mean it's wrong. I'm the opposite of an AI doomer. I think AI is the next evolution of humanity. We need innovation on the hardware substrate to actually build true intelligence. Please welcome Naveen Rao. Hey everyone, great to be here. You know, switching gears a little bit to…

Full transcript available for MurmurCast members

Sign Up to Access

More from All-In with Chamath, Jason, Sacks & Friedberg

Get AI summaries like this delivered to your inbox daily

Get AI summaries delivered to your inbox

MurmurCast summarizes your YouTube channels, podcasts, and newsletters into one daily email digest.