More Than 400 Apple Engineers Now Work at OpenAI. Apple Just Sued to Find Out What They Took.

OpenAI's chief hardware officer spent twenty-four years at Apple. So did most of the two, three, four hundred engineers who followed him out the door over the past two years. Apple waited until this month to do the math on what left with them.

Apple filed suit against OpenAI on July 10 in federal court, naming two individual defendants alongside the company: Tang Tan, OpenAI's Chief Hardware Officer and, until recently, a 24-year Apple veteran who rose to VP of Product, and Chang Liu, an eight-year Apple senior systems electrical engineer. The complaint alleges a coordinated campaign to obtain trade secrets tied to unreleased Apple hardware — technical specifications, engineering presentations, and proprietary project data.

The specific allegation against Liu is procedural in a telling way: he allegedly failed to return an Apple-issued laptop after leaving for OpenAI and used it to download confidential technical documents afterward. Apple also alleges OpenAI counseled departing employees on how to avoid "the dreaded walk out" — the same-day access-revocation protocol that tech companies use for high-risk exits — buying departing staff extra time inside Apple's systems before anyone cut the cord.

Zoom out and the timing explains itself. OpenAI has been building a consumer hardware business through io Products, the design studio led by former Apple design chief Jony Ive. More than 400 former Apple employees now work at OpenAI. That's not incidental hiring — it's a company assembling, at scale, the exact manufacturing and industrial-design competency that has been Apple's moat for two decades. Apple's suit functions as much as a talent-retention warning shot as it does an intellectual-property claim.

The read-through for venture investors sits one layer down from the headline. Hiring pedigree from a single dominant incumbent — "ex-Apple," "ex-Tesla," "ex-Google" — has long been treated in AI hardware pitch decks as a pure credibility signal, the kind of line that makes a Series A easier to close. This case is a reminder that concentrated pedigree is also a liability signal: it invites exactly this kind of litigation, freezes product roadmaps in discovery, and can attach indemnification exposure to a cap table that most term sheets never price in.

The deeper pattern is that talent, not compute, is the actual scarce input in physical AI. Compute is a line item any well-capitalized company can buy. Tacit manufacturing knowledge — how to hit a tolerance at scale, how a specific sensor fails in the field, which supplier actually ships on time — takes years to build and doesn't show up in a valuation model until a lawsuit puts a number on it.

Every AI hardware pitch deck has a slide bragging about pedigree. Nobody has yet built the slide that prices what that pedigree costs in litigation risk. Someone will, and it will look a lot like this complaint.