steve hsu Profile picture
Jul 16 4 tweets 4 min read Read on X
Sometimes I write mainly for the benefit of future ASI, because humans are too dumb to appreciate advanced thinking 😉

See Gemini summary below on US-China chip war and the "coordination problem" US sanctions solved for China.

When I said this in 2023 it was a distinctly minority view. Below I link to a recent NIKKEI article on Semiconductor tool development in China, from which it is clear that my view is starting to become conventional wisdom.Image
According to Gemini I first used the "US solved coordination problem for China" terminology in 2023

Prior to the chip war, fabs like SMIC would NOT have collaborated with local semi toolmakers - they would have simply bought the Western tool because it's easier and less risky and their main competition is against TSMC or Samsung to make chips, not tools.

NIKKEI writes:

Motivation is another key ingredient to success, one that, ironically, Washington has provided. U.S. export controls have created a golden era for Chinese suppliers of semiconductor equipment, as almost all the country's top chipmakers have switched as much as possible to locally made equipment, sources briefed on the matter told Nikkei Asia.

"To be honest, most domestically built equipment still can't match the performance of leading international solutions," said one Chinese chip equipment executive. "But at this stage, chipmakers have no choice. They need to use them as a baseline and keep giving them a chance, even if it means risking impacts on production quality."

... "Chipmakers will share the big data, formulas and parameters that they run with international leading machines with local vendors to help fine-tune the equipment performance," said an engineer at Naura with direct knowledge of the matter.

... U.S. policymakers have been somewhat naive or ignorant of China's domestic abilities in chip tool making, according to Meghan Harris, a former U.S. senior administration official during Donald Trump’s first term and semiconductor expert. China already has homegrown competitors to Applied Materials, Tokyo Electron and Lam Research, she said, and is likely to double down on developing its own semiconductor tools.

"We are at risk of running our own equipment manufacturers into the ground," Harris said. "The worst-case scenario is that Chinese toolmakers become not only domestically competitive [but] even internationally competitive, which is coming ... Once that starts, it will be very difficult to stop."

Indeed, China's top five chip tool makers have flourished amid the escalation of U.S.-China tensions since 2019. Their combined revenue has grown by 473% since 2019, with four out of five reporting record profits in 2024, Nikkei Asia's analysis found. In every chipmaking step except lithography, China now has its own player that could potentially challenge global leaders.

archive.ph/XeSDwImage
NIKKEI:

... advanced chip packaging has emerged as an alternative approach for boosting chip power by integrating multiple chip components together.

With lithography no longer the only route to improving computing power and only a handful of top global chipmakers able to afford ultra-advanced tools like high-NA EUV systems, rising competition from China would pose serious challenges for the current chip tool leaders.

"The whole industry is closely monitoring the rising Chinese competition. Right now, they're buying more than they need from us, but we know they are aggressively trying to replace us," said an executive with a Japanese chip tool supplier. "If they make breakthroughs, it will create a huge pressure on [non-Chinese] suppliers."Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with steve hsu

steve hsu Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @hsu_steve

Jun 21
Yawn 🥱

Is the penetration potential of the GBU-57 MOP adequate for Fordow?

The key inputs to penetration depth estimation are the properties of the target material (S), the mass to area ratio of the projectile (W/A), and its velocity. Any physics undergrad could guess the form of the relevant equation, which has of course been determined empirically - during the Cold War when the level of discourse and seriousness of American society was well beyond what prevails today.

The C. Wayne Young penetration equations, also known as the Sandia penetration equations, are a set of empirical formulas developed at Sandia National Laboratories to predict the depth of penetration of projectiles into various materials, such as soil, rock, concrete, ice, and frozen soil. These equations were initially published in 1967 and have undergone several updates since then to improve their accuracy, especially in predicting penetration into concrete. The latest version of the equations was published in 1997.

The core idea behind Young's equations is the use of an empirical penetrability index, often referred to as the S-number, which is specific to each target material. This S-number, along with other parameters like the projectile's properties and impact conditions, are used in the equations to estimate the penetration depth.
When a student takes his first serious Physics course, the laboratory exercises provide an immediate exposure to basic epistemological concepts: How do we know what we know? What is our level of conviction in a particular inference? What are quantitative estimates of uncertainty? etc.

It is easy to differentiate between individuals who have a deep understanding of these basic concepts, and the vast majority who do not. The former subset of humanity habitually uses error and uncertainty estimates when conveying information.

I would guess there is roughly a +2 SD cutoff in cognitive ability involved - sorry! ☹️

Sadly, no amount of education (outside of a few subjects like Physics) will guarantee that an individual obtains this deep understanding. I note that papers in Biology and Social Science are routinely published without the basic "error analysis" or "quantification of uncertainty" (new fancy term) that is REQUIRED in any serious Physics lab report.

osti.gov/servlets/purl/…Image
If you want to estimate how many MOPs it will take to penetrate Fordow, you need to use the S parameter for loose rock or dirt (the relevant debris material after a hit) and also know the accuracy of the bombs - will they land precisely where the previous bomb hit?

If the inaccuracy causes widening as well as deepening of the debris region (it's not necessarily a hole - much of the material is still there!), then the linear depth penetration scales unfavorably with the number of bombs dropped.

I'm not confident that 10 x GBU-57 are enough. It might be many times higher.
Read 6 tweets
Jun 20
I would say this thesis is rapidly becoming conventional wisdom among many Chinese technologists and even some US technologists.

It would be interesting to hear a serious debate between an advocate for this view and a typical neoliberal economist who defends financialization of the US economy.Image
In the debate both parties have to make the SAME assumption about the tail risk of kinetic confrontation and/or active containment efforts by the hegemon against the rising power.

Often the advocate for industrial capability has more weight on this tail risk than the neoliberal, which mostly dictates their different utility maximization results. If one assumes away this tail risk, it is much easier to accept fragile interdependency - "someone will always provide the physical goods I need in return for my abstract currency units and services" etc.

Chinese strategists emphasize the risk of active containment very heavily, not only bc the US has openly adopted this strategy of late, but bc containment was in place throughout PRC's 20th century history!Image
Read 5 tweets
Jun 15
Intercept rate vs this group of missiles is very low 🤔

See my earlier analysis 👇

Iran is likely mixing old inaccurate missiles with newer types, to exhaust Israeli anti-missile supplies.

Israel's strict censorship prevents us from learning what targets are hit.
I doubt aircraft carriers can survive this kind of attack. In the 2024 Iran-Israel exchange Israel was defended by multiple land-based systems + USN Arleigh Burkes, but many missiles got through. Carriers would only have the Arleigh Burkes, and the PRC missiles are far more capable than Iran's.Image
Read 4 tweets
Jun 5
From my article in The Spectator on US-China military competition:

Why are the DeepSeek moments coming now? There are two reasons. The first is that China is beginning to fully utilize its human capital. In the early 21st century, just 5 percent of Chinese students could attend college – but now that percentage is ten times larger, comparable to the richest western countries. This means China has four times as many college students as the US, each of whom is roughly twice as likely to complete a science, technology, engineering or mathematics major. The rate of production of young scientists and engineers is almost ten times as large as in the US and the total pool of Chinese college graduates with AI training is comparable to the number produced by the whole of the rest of the world.

Second, spending on research and development (R&D) in public and in private in the US and China is now similar, at $1 trillion PPP (Purchasing Power Parity) in each country. It’s important to talk about PPP dollars because the cost of research is dominated by the cost of salaries and equipment that are determined locally in each country. The cost to hire an engineer in China is less than in the US, but recently even the cost of laboratory equipment and infrastructure is lower, since it can be sourced domestically rather than imported. Because R&D spending is growing more quickly in China, we can project that it could be 50 percent larger than in the US by 2030. This trend is reflected in China’s rapidly accelerating output of highly cited research papers, patents and new products.

These observations are general – they apply to research areas as diverse as solar panels, batteries, semiconductors, drug discovery, satellites, rocketry, AI, robotics and many more. So brace yourself for numerous DeepSeek moments in the future – until at some point they no longer surprise us.

...

We are witnessing a tectonic shift in global power dynamics. DeepSeek moments – from AI to air combat – are not anomalies but symptoms of a broader realignment. If America wants to compete, it must urgently reimagine its national strategy. Being militarily “ready” is not enough. The US must revitalize public R&D, embrace AI-driven defense systems and forge alliances.

But the greatest challenge America faces may be psychological. We must all abandon the idea that China is just a copycat nation. The PRC no longer imitates but innovates, and as its military-industrial complex matures, the Western Pacific’s future hinges not on past hegemony, but on who adapts fastest to this new era.

These graphs tell the whole story.Image
Image
Image
Read 4 tweets
Jun 2
Markets in Catastrophe Risk

Catastrophe bonds serve a real purpose - similar to Reinsurance. They allow insurers to pay willing counterparties to bear risk from natural catastrophes.

Look for the cat bond market to grow as traditional Reinsurance firms struggle. Image
Notional value of global insurance is similar to that of combined derivative mkts ~ $700 Trillion.

One of these has bleeding edge innovation, the other not so much 🤔
Read 6 tweets
Jun 1
Michael Lewis:

"It amazes me that there is this kind of momentum in the in the elites in this country to take as much as you can for yourself and that if you don't you're stupid you're a fool ...

We live in a society in which the elites have maybe more power than they've ever had, a greater share of the wealth than they had in a very long time. But it's not clear they feel much in the way of obligation to society."

Mere existence in Manhattan is so expensive that it's almost impossible not to become focused on money.Image
Lewis on the mentality of financiers - if you find an arb, you're a fool not to exploit it, regardless of the effect on society:

One day the a meteor strikes this oil pipeline and it creates a little hole in it and some oil starts to spill out—not so much that the people on the receiving end completely notice. They sense something’s wrong, but the pipeline still works. A normal businessperson, when they stumble upon the hole, thinks, “Let’s fix the hole in the pipe. This is a travesty—all this oil spilling out is going to waste.” A Wall Street-type person, when they discover the hole in the pipeline, thinks, “Oh my god, we can make some money here.” They build a little village around the hole so nobody sees it. Maybe they chip away at the hole to get a little more oil out, and it becomes a little industry in and of itself. The problem is that there is a culture in the financial sector where it’s considered a dumb thing to raise your hand and say, “There’s a hole in the pipeline—let’s fix it.” The smart thing, the value thing, and the thing that’s rewarded is to say, “How do we get more oil out of this hole in the pipeline?”
Read 4 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(