CADDi raises $114M to write down what factory veterans never did
Two funding stories about AI leaving the screen: a manufacturing data platform that reads engineering drawings, and a drone autonomy company chasing a $20 billion price for pilots that never sit in the cockpit.
CADDi raised $114 million in a Series D that values the Tokyo- and Chicago-headquartered company at $1.2 billion, more than double the $470 million it carried in March 2025. The round brings total funding to $234 million and was led in part by Woven Capital, Toyota's growth-stage arm, alongside Moore Strategic Ventures, Coreline Ventures and Recruit Holdings' HR Tech Fund; Atomico, Globis and the JPS Growth funds reinvested. CEO Yushiro Kato declined to give Fortune revenue or customer counts, saying only that sales are more than doubling year over year. The traction numbers he would share are the interesting ones: customers in 22 countries, more than half of Japan's 100 largest manufacturers on the platform, and headcount up to about 900 from 600 in early 2025.
The product argument is narrower than "AI for factories." CADDi's in-house models parse 2D drawings and 3D CAD files directly, while general-purpose models handle the surrounding spreadsheets, procurement documents and supply-chain records — Kato's line to Fortune is that nobody uses an LLM for design review because LLMs don't understand drawings. Its Design Review workflow flags likely errors in new blueprints by comparing them against historical part failures, and the company's stated aim is a digital "superhuman" veteran: Kato says more than 80 percent of manufacturing process knowledge, including why a part was designed a certain way or a supplier chosen, is never recorded anywhere and lives only in retiring engineers' heads. Twenty design-review cycles per vehicle is the tax CADDi wants to cut; its target is compressing a car's planning-to-delivery timeline from roughly four years to four or five months by 2035.
Why it matters: the durable moat in industrial AI is data that was never digitized, not model quality — which is also why CADDi carries more customer-success engineers than salespeople and hires forward-deployed engineers to sit inside factories. We covered the same thesis from the inspection side yesterday — EXAONE Omni Inspect checks factory quality without retraining — and the pattern holds: the labs selling to factories win on the messy proprietary corpus, not the benchmark.
Shield AI is in talks for a funding round that would value it at $20 billion or more, according to The Information — up from $12.7 billion five months ago. Neither the round size, the investors, nor the final terms have been reported, and the company has not commented, so treat the number as a target rather than a closed deal. The San Diego-based firm sells military drones and the Hivemind autonomy stack that lets aircraft fly without a pilot or GPS, which is exactly the kind of product defense buyers now fund at software multiples — the same dynamic that had the Pentagon discussing a $5 billion loan to an AI cloud startup earlier this month. Valuations here price operational trust, and operational trust is the slowest thing in the industry to earn.
What to watch: whether CADDi turns its 22-country footprint into North American revenue as reshoring pulls factory software budgets, and whether Shield AI's round actually closes at $20 billion or the number drifts.
Factory knowledge walks out the door with every retirement — if your company's expertise lives in three veterans' heads, is an AI capture system an infrastructure cost or a bet on losing them? Tell us in the comments.
Sources: Fortune · NILE1 · Dealroom · The Information · AI Weekly · GuruFocus
