What are your key takeaways from recent investment trips?

Video Transcript

I've been on the road for the past three weeks all over Asia, Korea, Taiwan, Japan, in California and in Germany, meeting with over 50 different companies, all throughout the supply chain. And there's really four key takeaways that I've learned from this trip.

Number one is that GPU spend and getting ready for the GPU build out still remains robust. AMD has cited that they think that the AI chip market will go from US$45 billion in 2023 to US$400 billion by 2027. That's a 70% CAGR. If you meet with all the supply chain companies like we just did, they're getting ready for robust growth. If I take a step back, that US$450 billion market doesn't sound that scary in the top in the context of the entire global semiconductor industry. This year, global semiconductors will be roughly US$500 billion, and I think we'll be at a trillion dollars within a decade. US$450 billion of that trillion being spent on AI chips makes logical sense to me.

The second thing that we've learned is that this may be the best memory setup I've ever seen in my career. Memory and semiconductors is a commodity. What does that mean? It doesn't mean that it's not hard to make. This is some of the hardest stuff to make in the whole world. All it means is that you could swap the vendors out and have effectively the same performance, similar to a barrel of oil coming from Saudi Arabia or the USA. Memory chip from Korea or the US usually has roughly similar characteristics. Post COVID, we under invested in the memory market because of weakness across PCs, smartphone, telco, etcetera. But now we have an under invested supply market with a surplus and surging demand that's in the form of AI. AI requires 6X the memory consumption in an AI server versus a traditional server for DRAM memory chips and I think that the memory setup looks strong from here.

The third thing we learned is that there's going to be a renaissance in the Japanese semiconductor market. Why does it make sense to manufacture semiconductors in Japan? Number 1, there's plenty of semiconductor talent in Japan. You have great companies who’ve built up a strong ecosystem in the entire industry today. The second reason is ease of travel throughout Asia. It's a lot easier to move to Japan than say, move all the way to the United States. The third reason that we think is it's got geopolitically secure supply. We've learned throughout COVID how important it is to have geopolitically secure supply. And now we're learning through AI how important it is to have geopolitically secure supply. So with all of that, I think we're going to see a renaissance in the Japanese semiconductor market.

And finally, I think that the Chinese semiconductor market is investing heavily in analog and lagging edge semiconductors for the auto and industrial industry. What does that mean? We're probably bringing on a little too much supply in that industry today. So we'll see how the incumbents react to that supply, but we got to watch it closely as supply demand dynamics are the most important thing when navigating a semiconductor cycle. That leads us to our belief that the digital semiconductor ecosystem tied to AI is more attractive than the analog system today, which is more tied to the auto and industrial world and has these potential supply chains issues bubbling up underneath the service.