Analog Computing: Solving AI's Unsustainable Energy Demands

Key Quotes

"The brain's 20 watts of energy and the kind of computations that can happen inside of a brain and neural systems I was just blown away then and I'm still blown away by it and I think we haven't really scratched the surface of how we can get close to that biology is exquisitely efficient it's very fast it right sizes itself to the application at hand when you're chilling out you don't use much energy but you're so aware of other threats and things and once a threat happens like everything turns on it's very dynamic and we really haven't built systems like this."

Naveen Rao highlights the remarkable efficiency and dynamic nature of biological systems, specifically the brain, as a model for computation. He argues that current computing systems have not yet replicated this biological efficiency, suggesting a significant area for advancement in artificial intelligence hardware. Rao emphasizes that biology's ability to scale energy usage based on the task at hand is a key characteristic that current systems lack.

"Very simply digital computers implement numerics and numerics with some sort of estimation right I mean in a digital computer a number is represented by a fixed number of bits and that has some precision error and things like this it's just the way we implement the system if you make it enough bits like 64 bits you can largely say that maybe the error is small you don't have to think about it and so the digital computer is capable of simulating anything that you can express as numbers and arithmetic."

Rao explains the fundamental difference between digital and analog computing by defining digital systems as those that use a fixed number of bits to represent numbers, inherently involving some estimation and potential precision error. He notes that while digital computers are general-purpose machines capable of simulating any process expressible through numbers and arithmetic, this approach comes with inherent limitations in precision.

"Analog computers the first computers and they worked really well they're very efficient but they couldn't be scaled up because of manufacturing variability so someone said okay you know what I can't actually say that they can make a vacuum tube behave as a high or low very reliably I can't characterize the in between very well but I can say it's high or low and so that was kind of where we went to digital abstraction and then we could scale up."

Rao discusses the historical reasons for the shift towards digital computing, explaining that early analog computers were efficient but faced challenges with manufacturing variability. He states that the inability to reliably control analog components led to the adoption of digital abstraction, which allowed for greater scalability and became the dominant paradigm.

"So we believe we can find the right isomorphism in electrical circuits that can subserve intelligence that's a pretty wild idea isn't it maybe unpack it one level deeper because I totally agree with you computers for decades have been sort of the complement to human intelligence right it's like hey my brain isn't really great at computing an orbital trajectory sorry and I don't want to burn up on re entry so like a computer can help us with this incredible degree of precision we're now kind of going the opposite direction right where we're actually trying to encode more sort of fuzziness into computer systems."

Rao proposes that by identifying suitable electrical circuit designs, it might be possible to create hardware that directly supports intelligence, a concept he acknowledges as unconventional. He contrasts this with the traditional role of computers as tools for precise calculations that complement human intelligence, noting the current trend toward incorporating more "fuzziness" or probabilistic reasoning into computing systems for AI.

"The US is about 50% of the world's data center capacity and today we put about 4% of the energy grid the US energy grid into those data centers and this this past year 2025 was the first time we started to see news articles about brownouts in the southwest during the summer and you know just imagine what happens when this goes to 8 10% of the energy grid it's not going to be a good place that we're in."

Rao points to the significant and growing energy demands of data centers, particularly for AI workloads, as a critical global challenge. He highlights the current consumption of 4% of the US energy grid by data centers and projects a severe strain on infrastructure, evidenced by recent brownout concerns, if this demand continues to escalate. Rao suggests this energy constraint necessitates a fundamental rethinking of computing architectures.

"I think there are certain types of workloads that are amenable to these analog approaches especially the ones that are that can be expressed as um dynamical systems dynamics mean time they have time associated with them in the real world every physical process has time and in the computing world like the numeric computing world we actually don't have that concept you simulate time with numbers actually simulating time is very useful in certain uh certain problems."

Rao explains that analog computing approaches are particularly well-suited for workloads that can be modeled as dynamical systems, which inherently involve time. He contrasts this with traditional numeric computing, where time is simulated rather than being an intrinsic concept, suggesting that analog systems can leverage the natural temporal dynamics of physical processes for computation.