Do you believe the US government is willing to commit the resources necessary for Intel to catch up?
Intel's R+D spending (13 billion U.S. dollars) is already larger than the entire annual NSF budget (10 billion).
Intel's R+D spending (13 billion U.S. dollars) is already larger than the entire annual NSF budget (10 billion).
How much did we spend on the Iraq War, despite our own country being a major producer of oil and gas? Our economic incentive to entangle ourselves in Taiwan is even greater, but the stakes of military confrontation with China would be more dangerous. How much would you pay to avert such a conflict -- or worse, to avert the consequences of acquiescing to China's expansionism? Ironically, dependency on Taiwanese semiconductor fabrication reduces our ability to defend the island. Until the time that a fab interruption there would not halt our economy, Taiwan remains a US pressure point and China will continue to press it.
I think $200 billion is a cheap price tag for a path forward that involves neither warfare nor perpetual paranoia about every perturbation in the South China Sea. Taiwan established itself as a semiconductor powerhouse largely because its government backed TSMC. South Korea's government bends over backwards for Samsung, which makes up about a fifth of their GDP. The way forward for the US may be similar in mentality: throw the government behind the industry. US fabs need resilient and high-quality infrastructure for power, water, and logistics. We can use the collimated power and authority of the government to yield a vastly more effective long-term solution than piecemeal tax credits on lithography scanners.
[0] https://www.cnbc.com/2019/06/04/china-ramps-up-own-semicondu...
>> How much did we spend on the Iraq War, despite our own country being a major producer of oil and gas?
With all due respect, from 2004 - 2007, all we were talking about is Peak Oil and US was not such a major producer then as we are today. So some recency bias here.
>>The way forward for the US may be similar in mentality: throw the government behind the industry.
Yes, US will become mercantilist again, and I will rejoice but the goddamn world will burn or not.
> How much would you pay to avert such a conflict -- or worse, to avert the consequences of acquiescing to China's expansionism?
These are questions easier to answer than what most people think.
There are countless unclassified defence studies with very good estimates of how much would USA loose in case of an all out war with a nuclear state for worst, medium, best scenarios. I'll encourage readers to check them themselves.
Washington defence planners are most well aware of these things.
The later question is much more important, and with more real world relevance.
> consequences of acquiescing to China's expansionism?
There is no limit on on how much "acquiescing" you can do before such states will come up with more claims.
Remember "Finlandisation" precluding the WW2, and how it ended.
Giving the most minimal concessions to such opponent will instantly lead to them wanting more, and more, until you will have to retaliate, but now against much stronger enemy.
There is no "strategy," no political trick that works against opponent planning to attack you from the start no matter what. All treaties, political manoeuvres, compromise moves, even overt tribute payments will be only delaying the inevitable encounter in cul de sac.
In Russia, people gave a simple way to characterise behaviour of such people: "A guy who wouldn't stop until punched in the face."
Finlandization means strong country affecting a small neighbour, not applicable to pre-WW2 appeasement policies.
It's big power saying "let them have it" basically, and to disastrous results every time.
I'd be curious to see this quantified in some way. TSMC has a lead at the moment for bleeding edge fabrication, but Intel isnt that far behind and is massive in scale (revenue is nearly double TSMC, though its not an apples to apples comparison). A lot of the most impactful shortages this past year were in non-sexy parts like microcontrollers and other lower end chips that can be built in many different countries, since they are fairly simple (28nm, 40nm and even larger processes are commonly used).
The world is intertangled enough that losing Taiwan for some reason would be pretty bad, but I cant think of another country better equipped to handle semiconductor production than the US (Korea and Japan do well also, but at smaller scale than the US).
Pull up a table of companies with cutting edge foundries by year. The list perpetually grows shorter as the cost to remain competitive grows higher.
Sure, the semi shortage is a big deal to the auto industry, but the autos are almost rounding error to the semi industry.
Strong disagree. What kind of application would not be able to run at all on Intel's latest processors? I can't think of anything aside from maybe some future development of VR.
Most HNers deal with software engineering, but the rest of the world simply does not have to have the latest and greatest. Assuming Intel stays in the back and the West has to use only Intel, applications will be a bit slower, processors will use a bit more power, there will be more focus on software efficiency and parallelism, and more cloud streaming. Hardly "entirely dependent".
Modern IT infrastructure needs lots of different kinds of chips, not just CPUs. Most of them aren't manufactured in the US. If chips stopped shipping from Taiwan tomorrow, then sure, people would scramble to switch to other manufacturing, and presumably some of the gap could be filled from existing products or of Korea, but getting a design up on a new process can easily take a year, often more. Not to mention that building up raw fab capacity takes time as well.
The impact on supply chains in the short term would be devastating even if the long term can be gotten under control. If the short term includes a war with China, then, well... it certainly complicates any sort of military scenario.
That's only relevant to companies that have to build all their own CPUs.
The world is not HackerNews. 99.999999% of companies can do with the radical idea of using an existing older, slower, perhaps more power hungry processor. [It can be argued this already happened - all the Spectre/Meltdown mitigations costing performance, and the world kept trucking on]. Apple's bottom line may be temporarily hurt, and AMD will be in trouble, but that's hardly an economy-shaking threat.
>Not to mention that building up raw fab capacity takes time as well.
TSMC is actually rather small in the raw picture.
>The impact on supply chains in the short term would be devastating even if the long term can be gotten under control.
We were just having a huge pandemic. We can handle merely another chip shortage.
>If the short term includes a war with China, then, well...
Which is a different matter entirely.
Now, does that mean intel will actually be effective? I have my doubts, baring some sort of shotgun approach where multiple experimental fabs are done simultaneously.
They have the resources and history to recover. I have heard references to massive managerial layer problems and treating contractors like crap and the brain drain of the old guard, but those aren't intractable problems.
Intel has responded in the past. AMD looked completely dead in the Hector Ruiz waning days, and look at them now. Empower some good engineers and watch what happens.
Almost guarantee what that is: management building a huge pyramid reporting structure and then enriching themselves. To avoid reducing profits (which would threaten the bigwig C-level strike price options), they simply steal from the "proles", in this case the engineers that made the company.
Good news is: it's easy to clean house on shitty managers. Just drop half the layers or more, and then hire or acquire some good.
The other thing Intel had going for it back in the day: there were at least a couple teams working on x86 execution, so when Netburst crapped out they at least had the Israeli team working on the mobile architecture to switch to.
Intel's big enough that if you need two microarchitecture approaches, or even need to start an ARM project, well, they have the revenue if they want the talent.
I mean, look at what Apple did, they practically built the top-end chip in mobile within, what, four or five years? I guess I don't know exactly when they started working on an in-house A-series processors vs when they took leadership. But it didn't take long it seems, certainly relative to the four-year wishywashy timeframe being suggested now.