Atoms, Bits, and Moats: Why the Next Defensible Companies Will Be Built in the Physical World

Travis Kalanick spent eight years in stealth building what he calls an "atoms-based computer." Yesterday, he raised $1.7 billion from a16z to scale it. The company is called Atoms.

The name is the thesis.

I watched his full conversation with Ben Horowitz on the I'm Back podcast and came away convinced that his framework for thinking about atoms versus bits is the clearest mental model I have seen for understanding where durable competitive advantages will come from over the next decade.

Here is the framework, why it matters, and what it means for moats.

#The Atoms-Based Computer

Kalanick draws a direct parallel between digital computing and the physical world:

Digital Computer	Function	Atoms-Based Computer
CPU	Manipulates	Manufacturing
Storage	Stores	Real Estate
Network	Moves	Transport & Logistics

A CPU manipulates bits. What manipulates atoms? Manufacturing. Storage stores bits. What stores atoms? Real estate. A network moves bits from point A to point B. What moves atoms? Transport and logistics.

These are the three core computing resources in an atoms-based computer.

Kalanick takes the analogy further. A 10,000 square foot CloudKitchens facility is a "10,000 foot semiconductor." Thirty kitchens inside it are a 30-core processor, computing atoms instead of bits. The corridors where food moves are a network bus. The processing center is an L1 cache. Cold storage for ice cream and juice holds precomputed items sitting on the edge, like Akamai servers. The courier picking up food and delivering it is a TCP packet.

He even uses TCP congestion control (specifically TCP CUBIC) to manage how much demand gets routed to specific kitchens based on capacity and reliability.

Every data structure, every algorithm, every framework from computer science applies directly to the physical world. Same principles, different substrate.

#Peter Thiel Was Right (and Wrong)

A decade ago, Peter Thiel made a famous observation: we have had enormous progress in bits, but stagnation in atoms. His essay "The End of the Future" argued that the digital world races ahead while the physical world stays frozen.

He was right about the diagnosis. Software was easier. No FDA approvals, no zoning laws, no physical supply chains. A team of five could build a product that reaches a billion people. The incentives pushed every ambitious founder toward bits.

But Thiel was wrong about the prognosis. The stagnation was not permanent. It was a timing problem. We needed three things to converge before atoms could move at the speed of bits:

1. AI mature enough to handle stochastic physical environments. The physical world has far more variables and axes of stochastic distribution than the digital one. You cannot model a kitchen with deterministic algorithms. You need deep learning.
2. A generation of entrepreneurs not scared of regulation. Kalanick's words: "The entrepreneurs coming up since the Uber time are not scared of regulation the way the bits entrepreneurs were." Uber broke the template.
3. Enough operational knowledge encoded in software. Eight years of running 500+ cloud kitchen facilities across 110+ cities generates the kind of empirical data you cannot get from a whiteboard.

The window is now open.

#Why Bits Moats Are Eroding

Here is the uncomfortable truth about software moats: bits are infinitely copyable.

A SaaS product with strong network effects can be replicated by a well-funded competitor in months. An AI model that took $100M to train today will cost $10M to replicate next year and $1M the year after that. APIs get commoditized. Features get copied. Even data moats are weakening as synthetic data and foundation models level the playing field.

The classic software moats (network effects, switching costs, data advantages) still exist, but they are compressing. The half-life of a pure software competitive advantage is getting shorter every year.

This is not a new observation. But the atoms framework gives us the language to understand why, and where to look instead.

#Why Atoms Moats Compound

Physical-world businesses built on Kalanick's framework have fundamentally different moat characteristics:

Real estate appreciates, not depreciates. In a digital computer, storage becomes obsolete over time. In the physical world, if your computer optimizes the storage, the storage becomes more valuable. A food computer multiplies the value of its underlying real estate. As Kalanick puts it: "Imagine a computer that multiplied the value of its storage."

This is the same dynamic happening with data centers. The value of a data center in terms of what it provides society is going up dramatically over time. The real estate underneath it captures that value.

Operational knowledge is empirical, not algorithmic. The physical world is stochastic. You cannot solve it with an algorithm alone. You solve it by empirically understanding how it works. That empirical knowledge, earned through years of operating at scale, is extraordinarily difficult to replicate. Nobody starts a competing cloud kitchen network with the operational insights of 500 facilities on day one.

Regulatory navigation is a earned advantage. Once bits start controlling atoms in the physical world, regulators are in the game. Kalanick notes this is how regulation has worked for thousands of years: it is fundamentally about the physical world. Companies that have already navigated regulatory frameworks city by city, country by country, have a compounding advantage that new entrants must spend years and capital to match.

Physical infrastructure has a deployment timeline. You can deploy a software update to a million users in seconds. You cannot deploy 500 kitchen facilities in seconds. The speed of atoms deployment is measured in years, and every facility operational is a unit of moat that a competitor must physically build to match.

#The Convergence: Bits Controlling Atoms

The real insight is not atoms versus bits. It is atoms plus bits.

The most defensible companies of the next decade will combine software speed with physical defensibility. They will use AI to optimize physical operations, creating a flywheel where better software makes the physical infrastructure more valuable, and more physical infrastructure generates more data to make the software better.

Kalanick's three divisions (Atoms Food, Atoms Mining, Atoms Transport) each target trillion-dollar industries where this convergence is happening. Food production, mineral extraction, logistics. These are not sexy software categories. They are the backbone of the physical economy.

Ben Horowitz frames it well in his backing announcement: "It is inevitable that within a generation, robots are going to do most of the menial work in the world of atoms: transforming atoms, moving atoms, and storing atoms are inevitably robot tasks."

#What This Means for Builders

If you are building a company and thinking about defensibility, the atoms framework offers a clear lens:

If you are pure bits: your moat window is shrinking. AI is commoditizing software faster than ever. You need to either move incredibly fast or find a way to touch atoms.

If you are bits controlling atoms: you are in the sweet spot. Software intelligence layered on physical operations creates compounding advantages that pure software competitors cannot replicate and pure physical competitors cannot match.

If you are pure atoms without bits: you are waiting to be disrupted. Every physical industry that has not been digitized is a target.

The framework is simple. CPU, storage, network. Manufacturing, real estate, transport. The companies that build computers for the physical world, industry by industry, will own the next generation of durable competitive advantages.

Kalanick spent eight years proving this with kitchens. Now he is applying it to mining, transport, and whatever comes next. The man who digitized the car is digitizing everything else that moves, stores, or transforms physical matter.

The moat is not in the bits. The moat is in the atoms the bits control.