Hacker News new | more | comments | ask | show | jobs | submit login

Full disclosure, I work for Intel.

We have a fabrication plant in Chengdu, it's public knowledge that this fab is helping to manufacture products built on the latest process technology. As I've come to learn more about China's tactics when dealing with foreign companies it's become of great concern to me what this plant means for our future. I don't think it would be far-fetched to assume that some very protected and valuable IP has leeched through our doors and into China's hands. In all honestly I really can't fathom how the American government let this deal occur.

EDIT: To add more information [which is all public knowledge so if you're reading this Intel folks don't track me down! :)] its a packaging / assembly plant working on CoffeLake, which is 14nm++, CPUs are NOT fabbed there as Congress forbids it. My concern is more about the ability for the Chinese to potentially reverse engineer these products at assembly and derive IP. Also.. our packaging technology is pretty advanced so I'm concerned with even having an assembly plant there as well.

From people I've talked to who do manufacturing in China...if you aren't literally watching every single employee, every single moment, your IP is being stolen.

I’ve considered leveraging this in electronic manufacturing. Send design off to be cheaply fabricated, wait for it to appear on aliexpress as a clone with all BOM cost optimisation done free. Buy bulk lots and resell in local country. Bingo!

I agree. Doing this now for 5G "mesh" based BTLE devices which I hope will drop dramatically in price. Working out low-cost PMIC, crystal and passive component compatibilities for US based chipsets is a bit of a regulatory issue, but the "clones" iterate so fast I expect them to have complete designs available by mid-2019 and hope to make the money on the software / data.

Interesting. I suppose your value add could be EMC, packaging, support and infrastructure on that.

That works even better if your product has no electronics at all: https://news.ycombinator.com/item?id=18210950

Then hire another factory to clone that BOM optimized part for v2 :p

Just wait :)

If you look at the really cheap multimeters based on the DT830 design root they were originally derived from early Fluke DMMs circa 1985. Fluke hired Intersil to make the ASIC, Intersil second sourced it, the Chinese cloned it and cost optimised it, and then more Chinese guys cloned the target devices (panel meters and DMMs) and built mass production lines. Optimally they even optimised out the DIP packages as they were expensive and bonded the die straight to the PCB.

Down from $500 a pop to $2.

From my personal experience, they seed foreign factories with Military Hackers. It happen to us back in 2004 and we ended up closing the factory since it became a huge headache. They were constantly trying to get into our financials, HR and R&D network. It was comical when we would send a team to confront the employee. They would basically stand up and walk out without saying a word. There was nothing we could do. I still don't know how a company thinks it can remain secure having operations in China.

Samsung announced today that the Chinese stole their new bendable screen technology.


I also expect the productive work output of a foreign spy to be pretty low.

Also I think people are focusing too much on the problem just being IN China. I guarantee you there’s at least one Chinese national working at Intel IN THE US that’s a spy. That goes for just about any large company.

The US government? Isn't it Intel's problem in the first place? Don't the leaders of Intel care if their latest processor technology is copied and then used by competitors?

Yes absolutely, and look -- this is not to say that Intel does not have strict guidelines and processes that protect information. Certain things are definitely need to know around here. However, Intel is a business and if they've calculated that the cost of doing business is worth the risk they will take the risk. The government has put restrictions on semiconductors doing certain types of manufacturing in China but I wonder if these regulations are sufficient enough.

For instance, you can only manufacture CPUs in certain countries if the process is a couple of generations old. As Moore's law slow down the improvement from generation to generation will become less, are our current regulations sufficient enough to address this coming future? That's my concern.

Dont forget it's often as simple as executives bonuses could be easily gained by offshoring (motive), everyone was doing (opportunity), and they did it purely for selfish gain in short term. What happens in long term is next executives' problem, something to hand off to PR folks, and so on.

Happens all the time.

How exactly can one calculate such risk? At best somebody put an Exel file with a model that was wrong but that looked impressive in a presentation. But I doubt that even that was done. Somebody just put vague words that everybody was doing that.

There are things that one should not do no matter how tempting it looks. Rusdian roulette can be very attractive as 5 out 6 persons who tried claimed that it was a harmless.

Your whole premise is that China can't do it on their own, which is proved false by the supercomputer chip ban. Now Intel not only lost their market, China actually gained the capability and revenue to support future R&D. Your leaders at Intel appear to know more about business than you do: that it is higher risk over the long term to lose market share, because that gives your competitors revenue to support R&D and opportunities to work with and leverage a customer base to develop a better product. China has no problem making a CPU on their own. They just can't make a commercially competitive CPU, yet.

I've read through your argument multiple times and it still doesn't make sense to me.

> Your leaders at Intel appear to know more about business than you do: that it is higher risk over the long term to lose market share, because that gives your competitors revenue to support R&D and opportunities to work with and leverage a customer base to develop a better product.

Why would Intel be losing market share by not manufacturing chips in China? They may still manufacture the same number of chips but with less profit.

> China has no problem making a CPU on their own. They just can't make a commercially competitive CPU, yet.

Pretty sure parent comments meant processor fab tech in addition to processor tech. That is, tech that makes it possible to manufacture a competitive CPU at competitive costs.

Fabs are so fabulously expensive nowadays that you need serious volume to make them viable, and the volume today comes from mobile phones, IoT and telecoms equipment.

That's why TSMC and Samsung are spanking Intel in fab technology, which was originally Intel's competitive advantage, because Intel dropped the ball on ARM architecture chips due to its inwards-looking obsession with profitable x86. Intel is now boxed into the same low-volume corner as, say IBM PowerPC and ultimately condemned to irrelevance.

The last thing you want is for the Chinese to start investing in their own fabs and have a captive source of demand for those fabs' output, e.g. x86/x64 chips licensed from AMD. Quite frankly, they're going to steal the technology from TSMC and Samsung, not a has-been like Intel that's been incapable of moving to 10nm.

I'd argue a company like Huawei is probably able to make that kind of fab investment (they also design their own ARM chips), and they've shown in the past they are willing to take the long view by out-investing in R&D their Western competitors managed by the typical short-termist bean counter MBAs. That's why Huawei's CFO (and daughter of the founder) was arrested in Canada, to pressure them not to cooperate with the Chinese government's "Made in China 2025" plan to wean itself off dependence on US-controlled microelectronics. The fact John Bolton admitted he was aware of the move shows it was made on geopolitical grounds.

As I said earlier it is not just about profit at the margin to put a plant in China, it is about showing good faith with the government. Intel's technology lead is relative. It faces competitors like Loongson (MIPS), ARM licensees, AMD licensee and even Alpha licensees. It can't afford to take an antagonistic stance against a very large customer as any lost sales would go to supporting a competitor, and more importantly give the competition a customer base to sustain a positive feedback cycle of product improvements.

You say Intel decided the risk is worth the reward. Are you suggesting that the trade-off they chose is not the right choice for the US as a country? Why? Wouldn't Intel be the one hurt the most by any IP theft?

I think it's not unusual for companies to be forced to accept such risks if their competitors have already done so and that's enabling the competitors to win.

Governments can intervene to keep the playing field level, while protecting the national interests.

If it's true that the US has a dominant position in this space, then it's true that the US government has an opportunity to prevent the US companies from shifting manufacturing to where it's likely to result in IP theft in the process of competing with eachother in the manufacturing costs deparment.

If all the companies in the space are forced to ignore the cheaper manufacturing option, that self-destructive avenue of competition is off the table on equal terms. However there's still the threat of foreign competition undercutting on manufacturing, if they catch up in the IP department (which IP theft certainly accelerates).

This makes sense... So IIUC, it's not clear whether government restrictions are a net positive or negative for the US economy: it helps against one source of IP theft, but hurts in the sense of lower global competitiveness due to higher manufacturing costs.

A very difficult trade-off I guess.

IANAL but I believe the US government considers access to superior Silicon a matter of national security and regulates it as such.

It made a crucial difference in the 1980s but now? Who cares if your CPU is a couple of generations old?

It's more expensive to manufacture the same amount of compute power using older processes. That makes old silicon not competitive. From the viewpoint of the consumer, indeed, price per compute power is relatively constant, but from the viewpoint of a silicon manufacturer, newer processes are cheaper per compute power.

This doesn’t answer the parent point. Compute is more expensive with older tech, so what? They’ll just pay more. Consider North Korea. Their ICBM technology is pretty bad compared to the state of an art, however it is still great security concern.

They won't sell, and not enough to fund competing research. I'm pretty sure the security concern isn't directly related to defense, but to the security of US industries and economy.

That is assuming that a manager at Intel cares more today about another company stealing his companies technology and using it in future as opposed to getting the lowest cost for a new plant today. If the answer isn't obvious, here is a hint: which thing do you think the manager gets a bonus for?

You could say the same about outsourcing in general. It was a giant trend that everyone hopped on, and now it's affecting the entire US economy and transferring know-how to locals in other countries that are now starting their own competitors with government backing. The markets and businesses can be very short-sided.

The CEO of Intel is 58 years old. What does he care if China steals all of Intel's IP? He'll be retired by then.

Not knowing anything about the CEO of Intel, there are certainly many other driving motives for folks besides money and retirement. There's legacy, loyalty to family, country, environmentalism, pride, general ethics, etc. Certainly some business managers are driven by pure greed, but you'd certainly have to make a case more than point out someone's age to indicate that.

That's the wonder of the stocks insane liquidity. People just want the maximum possible return this quarter, so they can sell it and buy some other stock next quarter. Who cares how long the company survives?

And the best part is, if you force all the companies to act this way, and get some odd government protection for the large ones, the companies won't even go out of business. It's just society that gets poorer.

GP observes that Intel is behaving in a way that indicates it does not value future cash flows. I reply with one hypothesis for why Intel is discounting future earnings. (Very high discount rate on future cash flows, perhaps) You respond with reasons its discount rate could be lower.

Yes, I agree it would be nice for the United States if Intel's discount rate was lower, but where's the evidence? Does opening a chip packaging plant in China show "legacy, loyalty to family, country, environmentalism, pride, general ethics, etc"?

Although I think it's unfair to call out the CEO for his age, the notion of people behaving according to the laws of the system they live in, despite trying (and occasionally succeeding) in raising themselves above them, was noticed as far back as Ricardo and and only really put into perspective by Marx in the 1867 preface to the German edition of Capital:

>To prevent possible misunderstanding, a word. I paint the capitalist and the landlord in no sense couleur de rose [i.e., seen through rose-tinted glasses]. But here individuals are dealt with only in so far as they are the personifications of economic categories, embodiments of particular class-relations and class-interests. My standpoint, from which the evolution of the economic formation of society is viewed as a process of natural history, can less than any other make the individual responsible for relations whose creature he socially remains, however much he may subjectively raise himself above them.

The preface ends on a wonderful little note that hopefully everyone can agree on:

>Every opinion based on scientific criticism I welcome. As to prejudices of so-called public opinion, to which I have never made concessions, now as aforetime the maxim of the great Florentine is mine: “Segui il tuo corso, e lascia dir le genti.” [Follow your own course, and let people talk – paraphrased from Dante] Karl Marx, London. July 25, 1867.

I need to read Marx.

He is famously (or infamously) rhetorical in most of his work, but even his harshest critics (at least, those in academic circles) praise the fecundity and depth of his work. This possibility for interpretation was criticized by Pareto ("words are like bats - you are never sure if you are looking at a mouse or a bird" OWTTA) but has also provided the ability to use Marx's critical analysis on himself, and not only in the most narrow philological senses as one might imagine philosophers and political economists quibble over newly discovered German manuscripts.

One could spend their whole life studying Marx and not get through half of the work in Marxology (and two lifetimes if we include mathematical Marxism), value theory and classical economics - but to me that makes it all the better.

You seem well read in him, and philosophy/philology also. Can you recommend works in each you found worthwhile?

Most basically I'd recommend Marx, but maybe not without some guide to what it is exactly you're reading, why things are peculiar and the tradition which Marx inherited, which as his collaborator Engels admitted, was moribund even by the time Marx published Capital[0]. Although you can't avoid overlaying some interpretation, I'd recommend reading Capital alongside something like Johan Fornas' "Capitalism" (Routledge, 2014) who tries to mention various interpretative traditions, and for something more opinionated, Michael Heinrich's "An Introduction to the Three Volumes of Karl Marx's Capital". Value is a big topic in Marx, and while neoclassical economics dismisses it (what I believe to be out of hand) I.I. Rubin, who was persecuted by the USSR under false accusations, wrote a fantastic set of essays on his theory of value, named just as a collection of I.I. Rubin's essays.

Were you asking about philosophy in general? In that case, I admit I'm not very well read at all, even on Marx's Hegelian background - but there is a rather famous 4chan /lit/ board guide to philosophy from the start, though in my experience, it need not be read sequentially.

By philology I was referring to the study of the interpretation of Marx specifically - in my opinion it has produced some fascinating results (see Heinrich above, and importantly The Mismeasure of Wealth, which is a a set of essays by contemporary Marxian philosopher Patrick Murray[0]).

Also of interest is the discussion on Marx's more qualitative points besides his theory of value, specifically his theory of history, alienation, technology etc. for which Gerry Cohen and Sean Sayers have plenty, and Peter Hudis[2] claims from a philological point of view, it is easy to see how the USSR's style of socialism would probably come under fire from Marx himself.

On the whole, I'd just like to conclude with the fact that Marxism is very much not a unified school of thought, and in my modest and relatively uninformed opinion, it has only increased fracturing and has been marred by unfortunate internal hostilities after the 70s in which various philosophers tried to salvage what remained, reinvent, reinterpret or deny the grounds on which famous economists like Paul Samuelson and Piero Sraffa dismissed him. Marxism is so fractured that specifically regional schools have emerged, for instance the Uno School in Japan, AM in the Anglo-Saxon world, the Frankfurt School in 20th c. America, Neue Marx-Lektur in the German world etc.

Despite what it may seem, you asked for what I found worthwhile, but I've replied by citing almost all the authors who I've read at least one work of - I've read so little that I literally can't select what's worthwhile and what's not. This means I'm providing a limited picture, so take what I've said with a Capital-sized grain of salt!

[0] The "dialectical presentation" has come under fire from some Marxists as being insufficiently rigorous, outdated etc. - this is the Analytical Marxist school, which Roberto Veneziani (who you might also refer to on the matter of mathematical Marxism, along with Kliman's easy (but thick) "Reclaiming Marx's Capital") has written about: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2125671

[1] "Probing Marx's dialectical accounts of the commodity, value, money, surplus value, wage labour and capital, The Mismeasure of Wealth establishes Marx's singular relevance for critical social theory today."

[2] Peter Hudis - Marx's concept of the Alternative to Capitalism

Setting aside the questionable question of the CEO's motivations (they would happen in a young person as well, who would probably be in another company by then), the design of corporate governance is supposed to stop CEOs from making exactly that kind of terrible, self-interested decision. There is a board for a reason.

Likely he has Intel stocks. His children can have some too. Isn't it in his interest to preserve the value of Intel then?

Does intel even have a CEO now?

The American government is not a single entity as much as the Chinese government is. Also, if you are a CEO in the US and your tech gets stolen by China what is the repercussion? Nothing really. If you are one of the critical companies in China and the government finds you are letting info leak out I would think there would be major problems for everyone involved.

I mean, your business might go down the tubes if you let trade secrets out in the wild, while Chinese manufacturers undercut you on cost.

Any adverse effects would be felt many quarters in the future.

If this is a widespread issue, I wonder why do companies insist in building/keeping fabrication facilities in China?

I would guess the cost structure of chip manufacturing is heavy towards things other than personnel costs... but is that really so?

Is the waste management more "flexible"? Is it the price of electricity that is cheaper? Or what is it?

Because you can't access the market without giving in / setting something up in China.

Which is the whole Trade War argument about Free market, access and IP. I am not sure how long before the West learn they can expect whatever International "Standard" and "Values" to ever work in China.

A packaging facility is not a fab. It doesn't really give them any advantage in reverse engineering the silicon over someone that buys the end product and de-packages the die.

Nope. When you de-package the chip you damage it, especially on small features. They have access to the die bank, which means they have perfect die. Probably, with test pads still exposed. A microscope, etcher, etester, and patience is all you need to reverse engineer.

I always wondered about the latency on a successful reverse engineering of something like a modern CPU.

I'd expect there's a huge validation stage involved-- flaws in the extraction process could give you a plausible looking but ultimately faulty design to work with, and you have to check and clean that up.

Moreover, if you don't have exactly Intel's fab technology and oral-tradition knowledge, you probably have to retool the design to be more suited to the process you have.

It might still be a boost over whatever they have to offer now, but by the time it hits the market, it would be a generation or two old.

Intel would do well to be wary. AMD already gave away all their IP to their new CPUs to the Chinese government with the Dhyana cpus.

AMD gave away the farm but the money (I think 300 million) saved the company. All of this new innovation from AMD certainly arises from that money. Now however China has easier access to CPU's to build new supercomputer's find find ways to nuke the USA in new novel ways.

> All of this new innovation from AMD certainly arises from that money.

Just to be clear:

Zen microarch design work started circa-2012. The USD$300m deal was circa-4/2016.

It's accurate to say the capital from the deal likely helped AMD finalize, produce, and distribute Zen / Epyc.

But the microarchitecture design innovation predated the cash infusion.

>It's accurate to say the capital from the deal likely helped AMD finalize, produce, and distribute Zen / Epyc.

It didn't, it just gave a pass / access to Alibaba and Tencent as well as other Government related project to use AMD CPU.

We're not talking about the same thing. You're talking about the deal, I'm talking about what AMD did with the cash from the deal.

Look at their balance sheet, and note that without the $300M infusion, the market would have been spooked by their debt levels.


Amd also gave away the GPU farm when it made a similar deal with Intel. Both are shortsighted.

AMD delivered completed physical GPU components to Intel, which Intel then stuck next to their CPUs. They didn't give Intel anything Intel couldn't get by walking over to MicroCenter.

(So far as I know anyway)

Yes, what AMD did was basically give intel a dedicated GPU die with HBM which was then glued on the same package as the CPU.

I believe only for local use, not export outside China.

I think the whole point is that the Chinese government will buy (if allowed) or steal the IP (if not allowed) without compunction or care about agreements.

They weren't allowed the Intel/AMD64 ISA in all the previous years; but didn't steal it - some IP is much harder to get away with stealing than other IP.

For now.

At all, according to the contract. If you're predicting a change, I'd love to have more supporting detail.

This should be a management problem - why is it a government problem? Basically you are saying that your capital structure incentivizes managers to value short term gain over long term costs and so you would like to short circuit that with government intervention. However, when a government intervenes, it will never really leave you alone :-)

I'm concerned that a company like Intel, makers of some of the most valuable IP in the world, decided to invest in a foreign country instead of it's own when a company like Amazon -- which has great IP but don't get me wrong the Chinese have an equivalent (Alibaba) was able to scoop up prime real estate in NYC and DC.

Now Intel has a lot manufacturing based in the US, like I said Congress passed a law to make that happen. But have we looked into why we're allowing Foxconn to get tax breaks to build in "heartland" America but not Intel? An iconic American name?

So you think Intel should've setup shop in NYC or DC?

I'm sure Intel could/does get tax subsidies for land/building. But I don't think that's major the driver of cost for what we're talking about here. It's cost of skilled & specialized labor, and externality costs of supply chain availability.

> I really can't fathom how the American government let this deal occur.

greed and cheap labor. the exploited become the exploiter.

The hype of China has been done by the different investment bankers who were going on an on about China in the 2000-2005 time frame. The governments are lobbied by these investment bankers.

Why did Intel decide to build a fab in Chengdu?

I guess nearly all laptops are made in China so it makes sense to make the CPUs nearby. Esp if cheap labor, cheap land, and low environmental standards.

Rules of Origin & tariffs at a guess? Sometimes you can be creative with attributing value to each subassembly such that the place where you finally package something can make it "Made in China" (which in this case might be an advantage).

I can wholeheartedly believe to state of affairs as a nature of economics. Nobody bothers to quantify the risk of corporate espionage

For that same reason, Intel's chips should not be trusted by other countries. It is highly likely that Intel has planted backdoor into their chips. Thus, China and other nations should foster and develop their own chip industry and abandon Intel altogether. India, Canada, Japan, Europe, Russia should all ban Intel chips.

>>so if you're reading this Intel folks don't track me down

Should have changed your name for this post that doesn't put Intel in a good light: Just by your posting history they can probably track you down. Black activist--start from there and see if you made any complaints or suggestions to intel. Plus our style of writing is more or less unique.

But I agree with others, Intel management (at a certain point) may have interests that might be against the long term interests of Intel. Say, bonuses. That's the worst, but they may also make mistakes in deciding pro-cons.

Lastly, say that tech is worth $25Billion to Intel, and after x years is essentially worthless. Intel gets $15Billion from China in tax breaks, and sells more than $10B to others meanwhile. So economically it's worth to Intel and China gets the tech. But it is worth to our NatSec?

Question, out of all places, why did Intel put plant in the middle of nowhere Chengdu?

packaging and assembly is kinda low on the value chain, where china has already developed the know how. That's probably why there's a fab there.

My concern is more about the ability for the Chinese to potentially reverse engineer these products at assembly and derive IP.

The Chinese are, along with the Russians, probably at the top of the world's best reverse-engineers. You can bet they'll have already reverse-engineered it even if it wasn't made in their country.

Is it really any cheaper to build things in China? There are enough broke-ass places in the US where you could open a plant and offer $9 an hour and get the cream of the crop

$9 / hour, the cream of what crop? What in the world are you talking about?

Nationally Walmart is at $11, Target is at $12 and Amazon is at $15. Outside of convenience stores and fast food, increasingly that's your floor now.

Your $9 would actually be $30 / hour, at the absolute minimum. And you'd never get enough employees.

$9 an hour gets you better than average wages in a lot of places.

I was getting 12 as a qualified heavy equipment operator not not long ago.

Can you get ~1000 9$ an hour employees with engineering degrees though?

In China and most of Asia, you easily can. While these people won't be MIT grads, you can expect significantly higher scientific temperament compared to $MiddleOfNowhere, Utah.

You have hit the nail on the head.

The reason that Silicon Valley has so many of the Chinese/Indian/Other nationalities is simply because of the volume of available "Good Enough" Engineering talent. In most of these cultures a Master's degree is almost the norm. With this in hand, the practical aspects can often be learnt in training/on-the-job. Given that much of Modern Engineering has been systematically streamlined and commoditized, you really don't need very much "Creative" and "Out-of-the-Box" thinking as much as people who can manage and contribute within the existing framework and systems.

I assume you can, because holy fuck, are these supposedly engineering grads that I deal with from tata and cognizent and att and others of these gawdawful out sourcing outfitz, fucking terrible. I wouldn't hire for them to mow my lawn.

I can't express my distaste such companies any more without entering explicit territory.

> I wouldn't hire for them to mow my lawn

You wouldn't but enough people are, apparently, hiring them to do quite a lot of IT work nobody else would touch. Otherwise their business model wouldn't work at all.

It's sad, but apparently that is the case. I feel badly for those companies

> My concern is more about the ability for the Chinese to potentially reverse engineer these products at assembly

Would that be substantially more difficult with the finished product?

In order to push the performance on high end chips, you have to improve thermal conduction between the wafers and the chip package. This usually requires some sort of material (sort of like thermal paste) that doesn't interfere with the electronics but still provides excellent thermal conductivity. You can't use metallic compounds because it would short the wafer so in practice the material usually has strong bonding properties like a glue.

The smaller your feature sizes, the higher the probability that opening the chip to reverse engineer the wafer will rip up part of the silicon you're trying to see. The most interesting pieces of an integrated circuit also tend to be thermal hotspots (i.e. the Intel CPU instruction pipeline vs the cache) which are more likely to be destroyed because they use a stronger bonding material.

I see the same comments every time a china topic is posted on HN. I wonder why?

> I don't think it would be far-fetched to assume that some very protected and valuable IP has leeched through our doors and into China's hands.

China has a "technology transfer policy" for foreign companies doing business in china. So if intel has a fab in china, they agreed to the technology transfer.

> In all honestly I really can't fathom how the American government let this deal occur.

Probably because it benefited US companies and the wealthy.

If china is stealing anything, it's here in the US, not from intel who agreed to trade technology for chinese market access.

When American companies cannot operate freely (in the economic sense) in China and, more recently, when Chinese companies cannot sell their products freely in America, does it really matter? They are basically isolated markets at some point.

This is hardly the reality, plus the world is bigger than just China and the US.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact