We're witnessing the new age of corporate thefts, and the biggest thefts can be in a PDF file on a laptop that just happens not to be returned. Intel's engineers downloading 18,000 top secret files before layoffs. Silicon Valley engineers stealing more than 2,000 pages of confidential AI trade secrets documents and saving them to their Google Drive. And of course, Apple suing OpenAI and accusing its former employees of walking out the door with Apple's trade secrets. Everyone talks about the chip shortage and how only a handful of people really know how to build them. But the real shortage is time. And in this race, that stolen year of progress is what everybody is fighting over, whether they do it legally or not. What happens when old-school trade secrets collide with hypermobile tech talent? This is a story of how secrets really leave tech companies. How top secret information worth billions of dollars can slip past every safeguard and how fragile the whole idea of national tech security really is once you see how it all connects underneath the surface. Let's dive in.

Imagine a situation. You work in tech. You're senior or very senior, and you get a message on LinkedIn from a recruiter offering you a job. Much better pay, same job as before, and they want you exactly because of what you already know. And you decide to go for it. But where is the line between being hired for your skills and being hired to walk out the door with the documents and the knowledge of the trade secrets of your previous place? Let's map it out from start to finish. How does a trade secret theft happen? We're going to start with the demand side. The demand side appears when somebody, a company, an individual, or a country needs a skill or a piece of hardware that it doesn't have, and it doesn't have the years or cannot afford to spend years that it would normally take to build it. And to speed things up, they go shopping for someone who already has it in their head. One of the best documented trade secret thefts in modern tech history came out of Micron. And what makes it fascinating is the fact that it wasn't just a straight-up dirty theft. It was a government-orchestrated plan. Here's what happened. In 2015, China launched a plan called Made in China 2025. The goal behind this plan was to transform China from being the world's factory for cheap goods into a global high-tech superpower. One of the priorities of the plan was to create a stable production of RAM memory. As many of you know, we are going through a severe shortage of memory right now, which is why Apple products are getting more expensive and laptops are getting more expensive. Basic SSD cards went up in price several hundred dollars. If you want to understand the real reasons and economic consequences of the memory shortage, watch our previous episode. So, becoming the hub for RAM memory is a very logical move. But the two companies that were supposed to be frontier RAM producers under the Chinese plan, the UMC and Fujian Jinhua, did not have any RAM production up and running at the time. A logical question here would be, why would you pick someone to be the face of RAM production if they're not producing it? To that, I'll say, "Good question." China needed three major components to start producing RAM, a physical factory, experience in semiconductor manufacturing, and someone who knows the production of RAM inside and out. And so, Fujian Jinhua had the capital and the physical factory, UMC had the semiconductor manufacturing experience, and the only element they were missing was talent. UMC recruited a man who knew the ins and outs of the DRAM production. The man happened to be the former general manager of Micron's Taiwanese division, Steven Chen. If you think about it, seems like a very normal arrangement. We've all looked for jobs. We have all been called to interview exactly because of our past experience and because we know how to do something. And it is exactly the current experience that a future employer wants from us. So far so good. Then, Chen reaches out to two of his former subordinates. Both were still at Micron at the time when he reached out. And the two also happened to be the ones to hold the technical design files. He convinced them to jump to UMC. The two agreed, and upon leaving Micron, they download Micron's confidential RAM designs and manufacturing files. Later on, the US prosecutors revealed that one of them, whose last name is Wang, alone took over 900 confidential files and put them on USB drives and personal cloud. And once they jump to UMC, they used a file containing Micron's architecture to set up UMC's own DRAM design rules. When later on investigators finally raided UMC's files, the employees tried to hide the evidence, but it was too late. At the same time, the man who orchestrated the theft, Mr. Chen, became president of Jinhua's new factory. UMC pleaded guilty in the US and paid tens of millions in fines, but by then the transfer had already been done.

If you study more cases of corporate trade secret thefts, the mechanism usually works like this. Create a middleman, a new company, a subsidiary, or a partnership. In Micron's case, it was the Fujian Jinhua. Then they find a very senior, often an executive-level insider with access to the core subject of interest, whether it's a product or a design or a piece of hardware, you name it. It is important that it is not a person who makes the thing. It is always someone with authority. That insider recruits others, usually former subordinates, who hold the actual technical files and the skills. The files get copied before or shortly after the departure, often in a stupidly simple way, like a USB drive or a personal cloud account, or in the Google's case, internal notes converted to a PDF. When the files and the poached person move over to the middleman company, the management conceals the theft. They keep the files on off-network devices, they delete logs, they do all kinds of things to hide the fact that the files were stolen. Then the stolen data gets merged into the new company's own R&D, and the insider often moves into a leadership role. If the company from which the files were stolen finds out and wants to prosecute, they can and most likely will do so, but the state will make it difficult because they will apply diplomatic pressure because that would threaten national tech security. Criminal prosecution starts and it may last for years, and the suing side will probably win, but the problem is that by then the technology transfer will already have been completed. China runs hundreds of official talent plans that recruit scientists and engineers abroad. They offer salary, equity, lab funding in exchange for a contractual obligation to share new breakthroughs exclusively with Chinese sponsors. Sometimes, even while the person still holds their US job. China is of course not the only one. The US is doing very similar things and the Europeans are doing the same.

Now, let's reverse this and look at the supply side. The modern semiconductor fabrication is so complex that you can't just copy and paste someone else's R&D simply by looking at the designs. You need the people who run it. Taiwan's National Security Bureau states that it is close to impossible to steal the secrets of a semiconductor manufacturing process without stealing the fabrication plant itself, which is why when somebody's being poached, it's not just for the pure document theft. And this is also why the poaching is camouflaged very, very well. It can be disguised as domestic or third country capital. It can be done by a recruitment agencies. It can be done by using known industry names as a cover-up. For example, a networking chip company called Cloud Next built a Taiwanese subsidiary through which they aggressively recruited engineers. The engineers were coming from Intel and Microsoft and the company was doing that while hiding its Chinese ownership behind a Singapore-based registration. If you are one of those engineers who got poached, when you got the job offer, it might have looked perfectly normal, a perfectly normal offer from a normal-sounding company. Now, at this point, it would be fair to ask, do people who get poached always know? Do they know what they're being poached for? The answer is yes and no. Many high-risk engineers don't know the true ownership behind an employer that may reach out to them. I mean, put yourself in their shoes. You get contacted by a recruiter who offers you a fat salary to do the exact same thing that you're doing right now. You look up the company online, it looks fine, and you decide that you're interested. You get a job offer at a supposedly Taiwanese company as an example, but then they re-register to be Singaporean. Something that in corporate English sounds as circumventing regulatory scrutiny. Can you blame a candidate for taking the job when it is natural that a normal candidate doing basic diligence wouldn't think much of it? They see a Taiwanese employer and they think that it is a Taiwanese employer. Taiwan's investigator found that there is a whole market behind Chinese chip companies running structured poaching operations using front companies disguised as Taiwanese firms to legally recruit engineers. So, no, there isn't always foul play involved. As somebody who gets the job offer, you're just getting a better job. At the same time, it is important to note that the mechanism of corporate thefts is very deliberate and it often repeats. So, when you're being asked to bring the designs from the previous employer, you might as well start suspecting something. As a rule of thumb, the more senior the individual, the more impossible it becomes to claim ignorance. If you're being asked to use a VPN to sync work to servers, or if you're receiving salary transfers not from the country when you're supposedly employed, yeah, you are part of the fraud.

Now, everything we just talked about, from a recruiter's call to a state-backed subsidiary, is one example of a country solving the problem of "I don't have a technology I need and I don't have 10 years to build it." But this problem is not specific to China or Taiwan. It very much exists in the US, too, because the same pressure that pushed China towards poaching engineers and stealing files is pushing America's tech companies toward buying their way out of Nvidia's dependence. Let me explain. Meta will begin mass production of its in-house AI chips, code-named Iris, in September 2026. Who's providing the designs? Broadcom. Who's making the chips? TSMC, naturally. Now, why am I talking about Meta's chips in this video? And what does it have to do with Chinese operations? Meta is entering an already established business of major hyperscalers building their own domestic chips. Examples: Google is making its TPUs. Amazon pumps out Trainium and Inferentia. Microsoft makes Maya. Yes, they're all very different chips and they're made for different purposes and they're technically not direct competitors, but you can see how each of them tries to occupy and dominate its own segment of the chip market. Microsoft, for example, positions its Maya chips very narrowly. Their selling point is that they're very efficient at specific type of inference and if you buy their chips specifically for this purpose, you will save a lot of money. All in all, every hyperscaler wants to reduce reliance on silicon vendors such as Intel, AMD, and Nvidia. But to build those chips, you need the talent. So, the competition for the talent makes Intel's, AMD's, and Nvidia's talent pools a natural target. It is the exact same goal that China pursues when they built Fujian Jinhua, just pointed at Nvidia instead of Micron. The economics behind these moves are also awfully similar. The reason a hyperscaler needs chips is to bypass Nvidia's margins. Do you see the similarity? China needs to bypass sanctions. A hyperscaler needs to bypass Nvidia's dominance. Nvidia's AI chip business posted a gross margin of 75% and its data center revenue alone hit $51 billion in a single quarter. Those are huge margins and they exist exactly because nobody can replicate Nvidia's chip design. Trade secret thefts have the same goal that they achieve legally or illegally.

What I'm trying to get at and what all of this means is that AI creates a market situation where dominance doesn't make you safe anymore. If anything, it makes you the target. In traditional tech economy, it took decades to chip away at a monopoly. And of course, not every monopoly is falling just because AI race showed up. Perplexity has been around for years, yet we're still living a world where Google is used nine out of 10 times when somebody goes to search something on the internet. Nevertheless, the timeline to challenge a monopoly has compressed from decades to years and perhaps even months. You can see the exact same thing happening with other companies. The most obvious example is OpenAI. They went from being seemingly untouchable to facing very serious competition from Anthropic, Google, and open-source communities in under a year. And it is not because the models got worse, but because their competitors found ways to compress their time for learning curves. What I'm trying to get at is that the core difference in this market is time. By time, I'm talking about the speed at which a too big to touch player can be challenged. In AI economy, the real power isn't just being ahead. It's how fast you can stay ahead once the rest of the world decides that it won't wait for 10 years to catch up, and it's going to find ways to get to you, legally or not. If you'd like to understand the hidden links behind tech and AI economy, and see how one event on one side of the globe connects to a myriad of consequences on the other, follow us and never miss an episode. We hope this was helpful, and we'll see you in the next video. Bye.