There are many things that evolve every day in the chip industry, but some things remain the same.
What is changing is the advancement of AI chips that continue to train ever-larger models with more and more data.
Stagnation: Worrying vulnerabilities in the semiconductor supply chain.
Apple, Nvidia, Qualcomm, AMD and most other chip designers rely on Taiwan Semiconductor Manufacturing Company to make these critical components. TSMC is a cutting-edge manufacturer, and its location on the Chinese-claimed island of Taiwan puts the future of the tech industry and AI in a precarious geopolitical position.
Chris Miller, author of “War on Chips: The Battle for the World's Most Critical Technology,” spoke with Business Insider about what's happened since his book was published in 2022 and how we should think about the forces that threaten progress with AI.
This Q&A has been edited for clarity and length.
BI: Over the long term, what is the return on investment in AI and what GPUs can enable?
The investment level will only be sustainable if the products that all this computing produces are monetized. Personally, I'm a bit concerned about the size of the investment compared to what can be monetized today. A big tech company that's investing $10 billion per quarter in AI infrastructure will have at least another year, and probably a little longer, before investors ask, “What's the return on all this capex?” But at some point, the question will be asked in a way that forces them to answer it, or they'll be forced to slow down their capex.
So the best case scenario is that it doesn't actually matter. What do you think the worst case scenario is?
There's no question that we'll see a lot of AI use across a variety of use cases over the next decade. I think the realistic downside scenario is that we end up overinvesting a little bit and have to wait a cycle or two before building our next data center. This would be very tough for chipmakers. I don't expect this to happen, but I am concerned that the investment slope is too steep.
That brings me back to one of the key lessons from your book: how volatile the semiconductor industry is.
Yeah, it's always cyclical. So the question is, when does that cycle go up and down? Right now it's in a really steep upswing, and it always feels good when it's there. But it's a cycle.
Let's stick with the worst-case scenario: Given the rise of TSMC and Nvidia, do you think it's more or less likely now that China will invade Taiwan than it was 18 months ago?
I don't think it's unlikely. I think the international situation overall is pretty bleak with everything that's going on in Russia, Ukraine, the Middle East. If you look at what China is doing right now in the South China Sea with regards to the Philippines, it's largely because they know that the U.S. is overly nervous, so China can get away with it. And I don't see any dynamics that would change in the short term that the U.S. is overly nervous. And I think that's obvious to everybody, including the people in Beijing. This is a reason for serious concern.
What would happen to TSMC in that scenario?
The moment conflict breaks out in the Taiwan Strait, you have to assume that TSMC will be shut down immediately, regardless of what any company does, regardless of who decides to disrupt the supply chain or disrupt this or that.
Taiwan imports most of its energy, and semiconductor factories need energy. Also, many important chemicals and materials are imported into Taiwan, and these would also stop. Furthermore, if there was a gunfight, we would not be able to withdraw our ships from Taiwan. So if we can't actually sell semiconductors or get them off the island, the incentive to mass produce them would also rapidly decrease.
People like to speculate about who will blow up what. It probably doesn't matter as much as it seems, because supply chains will be much more vulnerable than many realize.
Could you tell me more about that?
Taiwan imports a lot of liquefied natural gas. Taiwan's power grid depends on it. According to data released by Taiwan, Taiwan has about 10 days' worth of LNG storage. This means that even running a few semiconductor factories would suddenly plunge Taiwan into a power crisis. And with limited power supply, what should be shut down first? Even without a gunfight, even a Chinese block tanker coming into port could trigger some very tough decisions about who can and can't get energy in Taiwan. And semiconductor manufacturers are major users of electricity.
In terms of semiconductor manufacturing materials, there are many important materials used in semiconductor manufacturing, such as photoresists, many of which are produced by Japanese suppliers. A stable supply of these materials is essential for semiconductor manufacturing, and if war breaks out, these materials will no longer be shipped.
Once people understand the true importance of TSMC, they are often quite shocked at how precarious the chip situation is. Are such reactions common?
Yes. The fact that we've got ourselves into this situation is very troubling, precisely because we can't take it lightly. The more we learn, the more we worry. The more we realize that everything hinges on maintaining peace in the Taiwan Strait, the more clear the stakes become. This is a conflict with potentially devastating effects on every aspect of the American economy. There are no silver linings. That's the reality. It's dark, and it's high stakes.
Nvidia CEO Jensen Huang said Paranoid It's about competition. Do you think a company in that position should also be worried about geopolitics?
My sense in talking to many people in the technology industry is that they believe military escalation and all-out war is less likely than is generally believed by people who spend time thinking about these issues in Washington, Tokyo and other capitals. So there is a mismatch in predictions.
If you're part of the Silicon Valley consensus and you say to yourself, “Oh, China is going to maximize their GDP. They're not going to do that because they're so focused on the economy,” then you don't have to worry as much, because you simply think that it's not going to happen.
But if that could happen, what would you do? For some companies, they have good options to diversify. But for many companies, they haven't had good options in the past. It will be interesting to watch such companies over the next few years and see whether their decisions change, especially as more high-performance chip manufacturing capacity is built in the U.S.
