Atmospheric AI raises $10M to tell buyers what a raw material should actually cost

Your supplier says prices are up $730 a tonne. Is that true? Atmospheric AI wants to answer with a model of the plant that makes the stuff. DBL Partners led the startup's $10 million seed round, according to Unite.AI, whose article was written by an AI research agent and reviewed by its editors.

How it works

Most supply-chain tools depend on suppliers sharing their own numbers. Atmospheric AI starts from how materials are actually made instead: chemical engineering, plant economics, and the links between feedstock, energy, production cost and emissions. Its connected model covers everything from oil fields and pipelines to refineries, chemical plants, terminals and the shipping lanes between them, and the company says it runs about 50,000 facility models.

From that it estimates what a material ought to cost, whether a price increase holds up, which other plants could supply it, and what a blocked route would do to cost, carbon and resilience. In one test, the company says, it matched a producer's published carbon footprint within 4.4% without touching any of that producer's data.

The $730 question

The best example is the one at the top. During the Hormuz disruption, a manufacturer got a proposed increase of $730 per tonne. Atmospheric AI's team worked out that the underlying production cost had risen by roughly $120 per tonne. That gave the buyers evidence to bring to the conversation, instead of a choice between paying up and arguing on instinct.

Haleon has worked with the startup for almost two years on resin sourcing. Chris Low, who runs sustainability procurement there, says his team now starts with what a material should cost, its carbon footprint and its supply risk rather than just the supplier's quote.

What the money buys

So far the work is mostly plastics. The round is meant to stretch it to chemicals, paper and metals, and to deepen work with about 100 manufacturers. Torsten Lichtenau, a former Bain senior partner, joins as founding chief commercial officer, and DBL's Vinnie Caruso takes a board seat. CEO Roee Goldberg says the eventual aim is to model the entire raw-material economy computationally, with chemical engineering underneath and AI doing it at scale.

The skeptic's note: the 4.4% match and the $120 figure both come from the company, and success with resin doesn't prove the method works for metals. That is exactly what this round has to show.

It's also a useful reminder that some of the most practical AI isn't a chatbot. It's a huge simulation pointed at a number people used to guess.

Scoreboard update: AI +1. A model of 50,000 plants finds only about $120 of real cost behind a $730 price hike.

Stay undead.