Bearing AI

The shipping industry is enormous, ancient, and - to an outsider with a machine-learning background - improbably analog. Roughly 90% of everything that gets traded across the planet spends part of its life on a boat, and yet the question of how much fuel any given boat will burn on any given voyage has historically been answered with rules of thumb, sea-trial curves from decades ago, and a captain's feel for the weather. The industry-standard prediction, if you can call it that, is right about 80% of the time. That 20% gap is fuel, money, and carbon, poured into the ocean.

Bearing AI was founded in 2019 by Dylan Keil and David Liu, both alumni of Stanford's AI Lab, on a fairly direct thesis: the gap is a data problem, and data problems have solutions. Keil had done this dance before - his previous company, Chronos, built a contextual-awareness engine for mobile devices out of noisy real-world sensor data and was acquired by Cisco. A ship, viewed a certain way, is just a very large, very expensive sensor platform floating through weather.

So Bearing built models that eat the things a vessel actually experiences - onboard sensor readings, satellite positioning, and weather data - and learn how that specific hull, in that specific condition, from a flat calm to a storm, converts fuel into distance. The company claims per-voyage fuel-consumption accuracy above 98%, against the industry's roughly 80%. It is an unglamorous number that happens to be the entire business, because once you can predict fuel that precisely, you can predict emissions, cost, and the smartest way to deploy an entire fleet.

The company came out of stealth in early 2021 with a genuinely load-bearing customer: Kawasaki Kisen Kaisha, better known as 'K' Line, which agreed to put Bearing's Analysis Dashboard on more than 300 vessels after concluding its numbers were simply more accurate than the tool it already had. That is the kind of validation that money can't buy - a conservative Japanese carrier telling its fleet to trust a startup's math.

Then the regulators arrived, which for Bearing was less a threat than a tailwind. The International Maritime Organization's Carbon Intensity Indicator, or CII, began rating ships on their emissions efficiency. Every operator suddenly needed to know, ship by ship, how a rule would grade its fleet - and most had no rigorous way to find out. In 2022 Bearing shipped a CII Optimization product for exactly this, and raised $7 million from its existing backers, AI Fund and Mitsui & Co., bringing its total to $10 million. AI Fund is the venture studio of Andrew Ng, which is a sentence that explains a lot: one of the most recognizable figures in artificial intelligence decided a shipping-software startup was worth a board seat.

The product line kept climbing the value chain. In 2023 and 2024 the company released a Fleet Deployment Optimizer - built with in-depth feedback from Hapag-Lloyd - that simulates future emissions and instantly compares how different vessels perform across candidate schedules. It is the difference between predicting one ship's fuel and orchestrating an entire fleet against compliance deadlines, service requirements, and cost, all at once. More recently Bearing has recast itself for the moment, describing its platform as "maritime intelligence built for the era of LLMs," with an Ask-Reason-Decide interface that lets a planner pose a question in plain language and get a modeled answer back.