Comment by kragen | Hacker News Reader

Comment by kragen | Hacker News Reader

Parent Full thread

it's an interesting question, and hopefully after a few decades of repealing intellectual property laws, we can find out. 40 years ago it wasn't obvious that anything like nvidia could exist in the first place. possibilities include:

- no company like nvidia could exist, and for chip design and fabrication we'd be stuck with companies like intel, digital, micron, samsung, and texas instruments; but many other kinds of companies could exist that can't exist currently

- fabs like tsmc would hire design firms like nvidia to produce designs to fabricate; the division of labor would be the same as at present, but banks and investors would send their money to tsmc to pay to nvidia, rather than to nvidia to pay to tsmc

- fabs like tsmc would provide open-source pdks like the skywater pdk to anyone who was interested in designing chips. different open-source gpu designs would proliferate, and jen-hsun huang would be the head of a nonprofit foundation in oregon, spending his days coordinating the contributions of a worldwide network of volunteer electrical engineers and raising his children

- microelectronics fabrication machinery research would be focused on small job-shop equipment using electron beams rather than multibillion-dollar euv fabs, so you could get the chips of your choice fabricated in any downtown with five-day turnaround, much like printed-circuit boards. as before, different open-source gpu designs would proliferate, and huang would be the head of a nonprofit foundation in oregon

- gpu development and fabrication would be internally funded by companies that wanted to use large numbers of gpus, such as amazon and the nsa

of course, companies like nvidia don't depend on the particular 'intellectual property' law being weaponized against anna's archive, and things like anna's archive benefit companies like nvidia rather than threatening them

fabs like tsmc would hire design firms like nvidia to produce designs to fabricate; the division of labor would be the same as at present, but banks and investors would send their money to tsmc to pay to nvidia, rather than to nvidia to pay to tsmc

There’s a bit of a snag with this. Chip designs are rarely ever done from scratch. Instead they’re iterated over many years, similar to how browsers, operating systems, other critical software is developed. It took NVIDIA decades to get where they are now. If anyone else can just take their designs as a starting point then NVIDIA’s whole investment (billions of dollars in R&D over decades) ceases to be a competitive advantage.

I think what would actually end up happening is that chip design as NVIDIA is doing would cease to exist as a business. Perhaps we’d end up with something more akin to an open source model like Linux. But then cost of manufacturing (paying for the masks and order startup costs) would still run into the millions, and TSMC would hold all the cards.

The reason I brought all this up though may have been missed by all the commenters to my original post: the U.S. government and their strategic interests. Having American companies like NVIDIA (and Apple as well, really) lose power and marketshare is not in the interest of the government. The last thing the US wants to see is for China to close the technology gap on this stuff.

in the case you mention, the geopolitical considerations run strongly counter to what you suggest

chinese policymakers can loosen domestic restrictions on innovation such as copyright and patent laws; then the laws in the us will only restrict us companies like nvidia. in large part this has happened, which is a major reason chinese companies (in both prc and roc) have become the leading organizations in a wide variety of high-tech fields, including solar panels, cell phones, electric cars, nuclear power, and microelectronics

your nvidia analogy predicts that gcc engineers and linux kernel engineers would have terrible job security, since anyone who needs a gcc backend or device driver written can hire literally any programmer; there are no legal restrictions. but in fact this seems to make the barriers to entry higher rather than lower. they're just in the form of 'human capital', knowhow, rather than in the form of the assets of a company

also, you may not be aware of this, but tsmc is a chinese company, and it's already left the us behind. sentences like 'The last thing the US wants to see is for China to close the technology gap on this stuff.' reflect wishful thinking that the technology gap is the other way around from how it actually is

In sentences like "The last thing the US wants to see is for China to close the technology gap on this stuff", China is referring to the PRC. In sentences like "you may not be aware of this, but tsmc is a chinese company", China is referring to Taiwan. It is disingenuous to conflate these.

conflating them has been the official policy of both prc and taiwan since the prc was founded, as well as of the un, and it shows little sign of changing. in day-to-day life, there is an enormous flow of money, technical information, hardware, and skilled workers back and forth between taipei and shenzhen. taipei is a 20-minute flight away from fuzhou. the idea of maintaining a 'technology gap' between the prc and the roc is more wishful thinking, like the idea of maintaining a technology gap between california and washington, or between france and germany. i mean, at least france and germany speak different languages

Being ambiguous as to what "China" means is both vitally important for international relations and also generally unhelpful for the purpose of clear communication. Corporate (and other) espionage notwithstanding.

there was no ambiguity in my comment; i said 'chinese companies (in both prc and roc)' to ensure that i was communicating clearly

Deliberately conflating the two, especially in discussion of US strategic interests, is not just unhelpful but makes your entire argument nonsensical.

sigh

Consider not spewing bs instead of sighing.

disagreeing with you is not the same thing as spewing bs. in fifteen years you'll see i was right

looking at your comment history, the most likely explanation for the disagreement is that you're out of your depth discussing geopolitics, the history of innovation, and international trade, so you're limited to repeating the ideas you're surrounded by, and even the best-founded counterarguments to them appear to you as 'bullshit' because you aren't familiar with the background knowledge they're based on

> tsmc is a chinese company

This part is wrong. TSMC is a Taiwanese company and as much as China's bullying behavior in the UN means most other contries do not officially recognize Taiwan, it still does not change the reality of the situation that Taiwan is independent in every way you can think of and the chines government has no more control over or benefit from TSMC than it does for an american company.

please see https://news.ycombinator.com/item?id=40915409

> The last thing the US wants to see is for China to close the technology gap on this stuff.

China doesn't have to care about US IP laws and largely already doesn't.

I assure you that 40 years ago chip companies were suing each other over IP just like today, if not worse.

> using electron beams rather than multibillion-dollar euv fabs

If this were possible it would be happening now.

> If this were possible it would be happening now.

that's an odd thing to say. if it's possible now it should have been happening 10 years ago? when does the infinite regress stop?

At the point that all the other necessary IC manufacturing technologies became available?

Focused electron beams are far older than microchips.

Reply on news.ycombinator.com