
Naveen Rao
Key Views & Dialogues
20VC: Why Robotics Will be 100x Bigger than Coding Agents | Inference Will Be a $10TRN Market: What No One Sees | Why Software Investors are F***** Today with Naveen Rao, Unconventional
Rao sees US spare power becoming scarce within years, tying AI growth to construction. Unconventional co-designs models and hardware for efficiency as falling token prices pressure inference margins, despite a $10 trillion market forecast within 10 years. He sees robotics 100 times larger than agentic coding, with prototypes in two to three years and a market in five to seven.
View summary (7)
- Rao expects spare US power capacity to become scarce within a few years, forcing AI growth toward the slower pace of infrastructure construction. He cites roughly $50 billion per gigawatt: building 10 gigawatts annually would already mean half a trillion dollars, while claims of 40 gigawatts are implausible today. He regards current financing concerns as “a blip,” with his confidence resting on continued demand for AI services.
- The hardware opportunity is to redesign the model and machine together around energy efficiency. Rao argues that recent gains largely came from smaller data types, while physical efficiency improvements stalled. Unconventional’s premise is that “we don’t treat the model as sacrosanct”: changing algorithms, architecture and physical implementation together can unlock gains unavailable through software optimization alone.
- Rao forecasts a $10 trillion inference market within 10 years, while expecting pressure on inference operators’ margins. Falling token prices squeeze the layer mapping models onto existing hardware; improving energy-to-intelligence conversion offers more room for differentiation. In three years, he expects open models to handle 80% of tokens, but spending to split roughly 50/50 between open and frontier models.
- AI’s immediate bottleneck is shifting from producing artifacts to judging whether they work. Rao describes reducing a research question from roughly two weeks to an hour, then warns: “I think we’ve moved the problem from producing to verification.” Specifications and experienced human judgment remain essential, which is why he expects development output to rise faster than spending shifts from people to tools.
- Rao’s attack on SaaS narrows considerably when Stebbings brings up businesses delivering outcomes through software and people. He endorses that model, including AI-assisted legal work, and says he would send a provider 80% of his legal workloads if he could gain confidence in its results. The unresolved question is whether existing firms retain distribution or vendors eventually sell those outcomes directly.
- Rao thinks robotics will be 100 times larger than agentic coding. His timetable is working prototypes for some tasks in two to three years and an emerging market in five to seven; Stebbings remains wary of application timing, particularly humanoids. Rao currently favors the technology stack and expects model developers to become integrated robot businesses: “You’re not gonna have a separation.”
- Rao now prioritizes technical progress over market validation when demand for the solved problem appears assured. For trillion-dollar markets, he calls price “almost irrelevant,” while Stebbings presses on fund concentration, LP expectations and limited diligence time. Rao’s underwriting principle is to back “someone to drive the bus,” because the original technical narrative will be wrong and the team must navigate the changes.
Naveen Rao: 4D Computing, AI’s Energy Wall & Beating Biology
Unconventional AI taped out a physical dynamical computer five months after starting in earnest, generating images at 500 nanojoules each versus millijoules on a GPU. Its 4D architecture co-locates compute and memory, targeting 1,000× power efficiency as AI demand approaches an energy wall in about 3 years. A rack-scale product is promised within two years, with adoption hinging on model-layer porting and a new Python-based software stack.
View summary (8)
- Naveen Rao publicly unveiled what he called the “first physical dynamical computer ever built”: Unconventional AI taped out the chip June 1, five months after the company “started in earnest in January,” and it is already generating images in the lab. The headline number: ~500 nanojoules per image versus millijoules on a GPU — “many orders of magnitude more efficient than a standard computer, and it’s because it just doesn’t move information around.”
- Rao’s prior track record frames the bet: he founded Nervana Systems in 2014, sold it to Intel (saying he sold it “way too early”), ran Intel’s AI group, then built GPU-scaling infrastructure with his team before joining forces with Databricks in 2023.
- The company’s goal is 1,000× power efficiency versus existing hardware, and Rao has pulled the timeline in from five years to three and a half — “we’ve actually solved very deep scientific problems more quickly because of AI, interestingly enough.” The stated endgame is bolder still: “the overarching goal of this company is to beat biology.”
- The energy math driving the thesis: Google alone crosses 3.2 quadrillion tokens per month; at 10 joules/token that’s 12 GW — against ~40 GW of total US data-center power and under 100 GW worldwide. Rao’s estimate: “we’re going to run out of energy pretty fast, in about 3 years or so,” and since ~50% of the cost of serving a token is energy, the business case is “we’re going to monetize that 1,000× better than existing hardware.”
- Biology is the existence proof: a human brain runs on 20 watts, a monkey brain on one watt — the same as your phone — and a squirrel’s on 8 milliwatts. The mechanism gap is data movement: the human cortex moves ~16 billion bits/second while a GPU moves ~30 trillion bits in and out of memory, “10–100× more than that” inside the chip.
- The architecture moves beyond conventional von Neumann designs — “compute and memory in one thing, we don’t have a memory interface” — branded “4D computing”: three physical dimensions via die stacking plus time as the fourth. A supporting breakthrough on sparsity turned n-squared connection scaling into a “holy grail” result: throwing away connections made systems more efficient, more scalable, and more trainable at once.
- Under Chamath’s questioning, Rao said a full product is “within two years”: a rack-scale data-center system, “tokens in, tokens out through a network cable, but the inner guts are completely different.” Existing models will work — porting happens “at the model layer,” not the operations layer — but matmul as such doesn’t exist in the hardware, and the software stack is Python libraries, not CUDA.
- The investor kicker is Jevons’s paradox: Rao says halving the price can lead to more than 2× consumption, while making something 1/1,000th the price can lead to consuming more than 1/1,000th as much. He thinks a 1,000× disruption of a roughly $1 trillion 2030 AI market — “maybe it’s bigger than that” — will create “the largest market that humanity’s ever seen,” with the topology shifting from gigawatt data centers to “many small data centers all over the place” and eventually “billions of robots.”
The Chip That Could Unlock AGI.
Unconventional is pursuing physical learning circuits, arguing that analog dynamics could complement digital computing for stochastic, time-dependent workloads rather than simply extend matrix multiplication. The thesis targets an energy wall: U.S. data centers use roughly 4% of the national grid, while some estimates call for 400 additional gigawatts over the next decade. Commercial significance depends on finding a manufacturable paradigm within five years, with TSMC partnership, a potentially very large analog prototype and unresolved AGI claims creating major execution risk.
View summary (6)
Naveen Rao’s thesis is that AI’s energy wall requires changing the computer, not merely scaling today’s architecture. U.S. data centers already consume roughly 4% of the national grid, he says, while some estimates call for 400 additional gigawatts over the next decade. The bottleneck is becoming physical: “The current paradigm, as good as it is and as far as it’s taken us, is not going to take us to that level.”
Unconventional AI is starting with a theory of physical learning, not a conventional chip roadmap. Rao wants to use circuits whose own dynamics perform useful intelligence rather than relying on layers of digital abstraction: Bornstein summarizes the idea as “Intelligence is the physics,” and Rao agrees. Human brains operate around 20 watts, while a squirrel or cat may use roughly a tenth of a watt—an efficiency gap large enough to justify revisiting first principles.
Analog is not meant to replace digital computing; it targets stochastic, time-dependent workloads that may benefit from physical dynamics. Rao’s candidates include diffusion, flow, and energy-based models because their dynamics can be written as ordinary differential equations and potentially mapped onto physical systems. His goal is an “intelligent substrate” alongside conventional computation, preserving digital precision for problems that require it.
Rao thinks dynamics might supply the causality missing from current AI, but he repeatedly labels the AGI argument “hand wavy.” His intuition is that systems built from elements with real time evolution will provide a better basis for understanding cause and effect than systems whose basis does not include such dynamics. Today’s models contain intelligence and are highly useful, he says, but remain “nowhere close to AGI” because they still make elementary errors and do not feel like working with a person.
The commercialization test is whether Unconventional can find an intelligence-like physical paradigm within five years and make it manufacturable at massive scale. Rao sees TSMC as a necessary partner for prototyping and eventual scale; he characterizes Google as having everything internally and pursuing lower-risk, continual TPU improvements, while Nvidia has built the dominant programming platform. Unconventional is trying to build “a better substrate than matrix multiply,” though Rao leaves open collaboration rather than assuming direct conflict.
Execution risk is extreme, but Rao sees several partial proofs rather than a blind leap. The brain provides an existence proof, more than 40 years of academic work supplies evidence of promise, and dynamical-systems and neuroscience theory offer pieces that engineers can combine and “sand down.” The first prototype may be one of the larger, possibly the largest, analog chips ever built, making talent across theory, systems, algorithms, analog circuits, and digital circuits central to the thesis.