In an industry where AI infrastructure secrets can determine who dominates the next decade of technology, the United States has sent the same message Taiwan delivered with its TSMC prosecution: stealing cutting-edge tech secrets will cost you decades of your life. A San Francisco federal jury just convicted former Google engineer Linwei Ding on 14 counts of economic espionage and trade secret theft, with the defendant facing up to 70 years in prison for allegedly stealing Google’s most sensitive AI data center designs and transferring them to benefit Chinese interests.
This marks the first major criminal conviction specifically tied to AI-related economic espionage—a watershed moment that mirrors Taiwan’s recent 14-year sentence for TSMC chip theft. Just as Taiwan classified semiconductor technology as “national core technology” requiring enhanced legal protection, the U.S. Justice Department is now treating AI infrastructure the same way. The message is clear: AI capabilities have crossed from corporate competitive advantage into national security territory.
⚡
WireUnwired • Fast Take
- First AI espionage conviction—Linwei Ding faces up to 70 years for stealing Google secrets
- Stolen: TPU chip specs, GPU cluster designs, proprietary SmartNIC networking technology
- Used Apple Notes to evade detection while copying thousands of files over 11 months
- Mirrors Taiwan’s TSMC case—nations treat AI/chip tech as national security, not corporate IP

What Was Stolen and How
The stolen material covered seven categories of trade secrets describing how Google designs, builds, and operates its AI data centers—the infrastructure powering Gemini and enabling Google to compete with OpenAI. This included custom TPU chip specifications, system architectures for linking thousands of processors, GPU cluster configurations, and proprietary SmartNIC networking hardware that keeps training clusters synchronized. For competitors or nation-states, these operational details represent tens of billions in R&D shortcuts.
Ding’s method: copying information from Google’s internal systems into Apple Notes on his company MacBook, then converting notes to PDFs and uploading thousands to personal storage. Over 11 months, this evaded detection by mimicking normal employee behavior—until the systematic volume correlated with his pursuit of roles at Chinese AI companies and venture funding for his startup, Rongshu.
A Pattern of Strategic Technology Transfer
This isn’t isolated. Recent Justice Department prosecutions include four individuals charged with smuggling Nvidia GPUs to China, Intel employees allegedly stealing classified files, and $160 million in smuggled chips. Multiple TSMC engineers faced charges for transferring 5nm and 10nm process technology to Chinese companies. The pattern: systematic efforts to acquire Western AI and semiconductor capabilities as U.S. export controls tighten on hardware sales.
U.S. Assistant Attorney General John A. Eisenberg called it “a calculated breach of trust involving some of the most advanced AI technology in the world at a critical moment in AI development.” Together with Taiwan’s TSMC prosecution, these cases signal that nations competing for AI and semiconductor supremacy are using criminal law as a weapon, with decades-long prison sentences as the deterrent.
What This Changes
The conviction establishes that AI infrastructure qualifies for economic espionage charges—the same legal protections as military technology. Future defendants can’t claim AI is just commercial software. For tech companies, expect TSMC-style zero-tolerance policies: stricter access controls, enhanced monitoring, and aggressive prosecution of suspected violations.
For employees with connections to competing nations, this creates professional complications. Companies must balance legitimate security against discrimination—implementing controls based on job necessity, not ethnicity, while requiring disclosure of external affiliations. The risk: security paranoia leading to ethnic profiling and hostile work environments.
Ding faces sentencing soon, likely 15-25 years of a 70-year maximum. The precedent is set: steal AI secrets, face decades in prison—whether you’re in San Francisco or Taipei.
For analysis of technology espionage and semiconductor security, join our WhatsApp community.
Related: Taiwan’s 14-Year Sentence Exposes Chip Espionage Crisis
Discover more from WireUnwired Research
Subscribe to get the latest posts sent to your email.


