Tractian raises $130 million at a $1.4 billion valuation to predict machine failures

The round

Tractian, an Atlanta company that puts sensors on industrial machines and uses AI to predict when they will fail, has raised a $130 million Series D at a $1.4 billion valuation. The company announced the round on 22 October 2026. It was led by Sapphire Ventures and co-led by Catalyst, the growth equity strategy of the UK investor M&G. New and existing backers N47, General Catalyst, Bosch Ventures, 53 Stations, Monashees, Endeavor Catalyst, Norte Ventures and Y Combinator also took part.

Sapphire also led Tractian's Series C. Anders Ranum, a partner at the firm, said in the announcement on Tractian's blog that the AI boom is driving a big build-out of physical infrastructure, and that the companies building and running it cannot afford equipment failures. When Sapphire led the previous round, he said, its bet was that combining hardware, data and AI would give Tractian an edge, and that thesis has only strengthened.

The valuation makes Tractian a unicorn, and according to the Atlanta tech outlet Hypepotamus, one of the first AI-focused startups in the city to reach that mark. Hypepotamus says the round was first reported by The Information, which also reported that Tractian's annual recurring revenue crossed $100 million in September. Tractian's own release does not give a revenue figure, so that number rests on The Information's sourcing.

What Tractian sells

Tractian's core product is a small wireless sensor that bolts onto motors, pumps, compressors and other rotating equipment. According to its condition monitoring page, the sensors install in minutes and send vibration, ultrasound, temperature and RPM readings over 4G/LTE, so a plant does not have to set up Wi-Fi or involve its IT department. Tractian says it is the only product on the market that combines vibration and continuous ultrasound in one device, with vibration spectrum readings up to 64 kHz.

The data goes to Tractian's software, where models map signal patterns to specific failure modes such as bearing wear, misalignment and imbalance. The company now describes this as a multimodal, agentic platform: it combines several sensor types into what it calls a single living picture of each machine, then produces a diagnosis, a likely root cause and a recommended plan for the operator. Tractian says its models are trained on one of the largest industrial failure datasets in existence, which is a claim it has not backed with a number.

Put simply, it is predictive maintenance: fixing a machine shortly before it breaks instead of on a fixed schedule or after it fails. The idea is decades old, but it used to need specialist analysts walking the floor with handheld meters. Cheap connected sensors and machine learning let a company watch thousands of machines continuously.

Who uses it

Tractian says more than 2,000 facilities around the world now rely on its technology. The customers it names include the drugmaker Eli Lilly, McKesson's distribution centres, Whirlpool's appliance plants, Coca-Cola bottling sites, Cummins' engine factories and the aircraft maker Embraer, along with unnamed data centre operators and government infrastructure.

The equipment it watches is everyday plant machinery: compressors, chillers, palletisers, pumps and backup systems. Its target industries are pharmaceuticals, food and drink, oil and gas, data centres, government and defence, energy and water, and factories that are bringing production back to the US.

Government work got easier this summer. In July Tractian said it had become available to federal agencies through a FedRAMP High authorised environment in AWS GovCloud, run through a partnership with Knox Systems. FedRAMP High is the strictest baseline in the US government's cloud security programme, and having it means agencies can adopt the product without running their own lengthy authorisation.

The pitch: downtime and missing technicians

Tractian's funding announcement leans on two problems. The first is the cost of machines stopping without warning. Citing a January 2026 Siemens report on the true cost of downtime, it says unplanned stoppages cost the world's 500 largest industrial companies about 11 percent of annual revenue, or roughly $1.4 trillion a year. It adds that industrial fixed assets now average 24 years in service, so much of the equipment in use is old.

The second is people. Tractian cites a Deloitte projection that more than 2.1 million manufacturing jobs could go unfilled by 2030, and argues that the know-how of experienced technicians, how each machine behaves before it fails, leaves with them when they retire. Its software is pitched as a way to keep that knowledge in the plant, recorded against each machine where the next technician can find it.

Those figures come from Tractian's release and the reports it cites, so they show the size of the problem the company wants to address rather than its own results. The release gives no figures on how much downtime its customers have avoided.

Founder and CEO Igor Marinelli tied the mission to his own family. He grew up watching his father work as a maintenance technician at a paper mill, he said, and saw how backwards the tools were for the people keeping those machines running. He said the money will help Tractian scale its platform for manufacturers, data centre operators, defence companies and utilities, close the gap between an alert and an action, and eliminate unplanned downtime.

From Brazil to Atlanta

Tractian is headquartered in Atlanta, with offices in São Paulo in Brazil, Monterrey in Mexico and in Poland, and has more than 700 employees. Marinelli told Hypepotamus this spring that he moved the company from the Bay Area to Atlanta to be closer to engineering talent and to customers in asset-heavy industries, and the company was preparing to move into a new Midtown Atlanta headquarters.

That makes Tractian an unusual unicorn: its biggest investors are on the West Coast, but its customers are factory and plant operators across the Americas, served from offices in Brazil and Mexico as well as the US. Its investor list mixes Silicon Valley firms with Latin American funds such as Monashees and with Bosch's corporate venture arm, a sign that one of the world's big industrial suppliers is watching the space closely.

Why investors are backing physical AI

Tractian's round is one of several this month for companies that apply AI to the physical world rather than to text or code. Gecko Robotics, which uses wall-climbing robots to inspect industrial assets, saw its US Navy contract ceiling raised to $300 million, and Kevin Weil's Warp Sciences raised $215 million to build robotic chemistry labs. Eka Robotics raised $150 million to teach robot hands to feel force.

What sets Tractian apart is that it is already selling a product to large, conservative industrial buyers, with a reported nine-figure revenue run rate, rather than building toward a future market. Its challenge is the familiar one for hardware-plus-software companies: keeping margins healthy while shipping and supporting physical sensors in thousands of plants. Its software side competes with a growing number of AI forecasting tools for sensor data, including open models such as Mistral's Timestral time-series models.

For plant managers the pitch is simple: fewer surprise breakdowns and less reliance on a shrinking pool of experts. With $130 million more and the federal market now open, Tractian has the money to test whether it can turn that pitch into a standard part of how factories, data centres and utilities run.