Eight Years Building in Stealth. Travis Kalanick Just Raised $1.7B for the AI the Market Missed.

Physical AI — the kind that routes supply chains, optimizes mining operations, and controls food production lines — attracted approximately zero percent of the capital that went into language models over the last three years. Atoms, which has been operating in stealth since 2018 across mining, food robotics, and logistics in 110+ cities, just raised $1.7 billion. The round suggests the market is starting to recalibrate toward the problem that AI has been ignoring.

Travis Kalanick founded Atoms in 2018, less than a year after resigning from Uber. The company operated without external attention in exactly the sectors language models can't easily reach: physical industries where value lives in real-time sensor data, robotics integration, and operational coordination across thousands of physical sites. Mining. Food processing. Logistics. The company had operations in 110+ cities before it took in a dollar of outside capital. Eight years of proprietary operational data accumulated before the fundraise, not after.

The round includes Bain Capital, Fifth Wall, K5 Global, and Uber, alongside debt facilities with Bank of America, Goldman Sachs, Wells Fargo, JPMorgan, and Barclays. Uber's participation is the data point worth holding. Kalanick was forced out as Uber's CEO in 2017. The fact that Uber invested in his next company isn't just a human-interest story — it's a signal that what Atoms built is compelling enough to override institutional history. Physical AI infrastructure for logistics and supply chains is exactly where Uber's competitive interests sit.

Physical AI has different economics than digital AI. In digital AI, the cost of intelligence is essentially compute and training data — both of which can be purchased, commoditized, and shared across companies. In physical AI, you also need sensors, robotics, and — most critically — the operational data that only comes from being physically embedded in industrial environments. Atoms has eight years of that data across three industries. The moat isn't the model. It's the 110+ cities of real-world operations that no competitor can replicate in 12 months regardless of how much capital they deploy.

The sectors Atoms targets process staggering value. The global mining industry handles roughly $2 trillion in annual revenue. Food production is a $10 trillion market. Logistics is another $10 trillion. These are industries where a 2% operational efficiency gain is worth more in absolute dollars than building another large language model API. They're also industries that have received almost no attention from AI venture capital, because they're not addressable from a desk — they require physical presence, industry-specific engineering, and patience that the typical startup funding cycle doesn't support.

For investors in Latin America, the Atoms round has a specific implication. Brazil is among the world's largest agricultural producers and has significant mining operations in iron ore, oil, and bauxite. Colombia, Chile, and Peru are significant mining economies. The application of physical AI to those sectors — precision agriculture, mine safety and efficiency, logistics network optimization — is an opportunity that has barely been identified as an asset class in LatAm. Atoms is validating that physical AI at industrial scale attracts frontier capital. The question is whether the physical economy of Latin America gets its own version of this bet.

Kalanick built Uber on the insight that physical logistics were fundamentally software problems waiting for the right platform. He spent eight years and $1.7 billion quietly proving the thesis again — this time without a ride-hailing logo and without announcing what he was building until the data moat was already in place.