“We’re Going To Get Lots Of Booms And Busts” In AI Stocks - Forbes
High confidence: full text extraction produced 5080 characters.
Platform stocks such as Nvidia, hardware makers such as TSMC remain poised to profit, says TriOrient’s Dan Nystedt
Volatility has been a big theme among AI-related stocks among companies profiting from the AI rollout. This week, Meta lost 9% on an earnings miss, but climbed 3% on Friday; Apple fell 7% on Friday alone even though it beat forecasts. Chipmakers were hammered earlier this week, but industry heavyweight TSMC soared 7% overnight in the U.S. on Thursday after tumbling to a three-month low a day earlier.
What should investors make of all of this? I spoke in Taipei last week to long-time successful Taiwan tech stock-picker, Dan Nystedt, vice president at family office investment firm TriOrient Investments. Taiwan’s been a center of the semiconductor and infrastructure hardware boom powering the Nasdaq higher of late. The tech-heavy main index of the Taiwan Stock Exchange is up by more than 83% in the past year.
“We’re going to get lots of booms and lots of busts” in stock prices, yet the underlying growth trend in the industry will Nystedt believes “the AI trend is going to go on for a while.” Winners will include platform companies such Nvidia and Amazon, as well as Taiwan hardware suppliers such as TSMC and Wistron. Excerpts follow.
Flannery: Where does the AI boom go from here?
Nystedt: You can look at the boom in terms of stock market versus separately from companies and their earnings. The hype cycle that goes around a trend like AI is big. We're going to get lots of booms and busts.
We'll get more since this AI trend is going to go on for a while. I wouldn't even look at it in terms of, “Did we miss it? Is it done?” I don't think so at all.
When investors get excited and start putting money in just because they hear about big earnings numbers, smarter investors will sit back and say, “Well, are they really making that just because they're giving these forecasts? I'm not sure that that money is actually going to go where they say it's going to go.”
So there's a period where you have people trying to look at what's actually happening beyond the forecast, and that that's going to lead to different opinions, which is what makes a market.
Flannery: Who’s poised to last through this?
Nystedt: That’s the best question of all. I would say the platform companies are going to last. And then for a long time, it's going to be hardware companies.
What I mean by platform is companies that don’t just make chips. Like Nvidia, they make the entire server and the entire data center. They're also figuring out ways to hook up millions of GPUs at the same time to act as one massive brain for AI. They’re not just chip makers anymore.
You're talking about system makers and an entire platform, because they also have their own software. Google is another platform company because they make TPU (tensor processing unit) chips. They also create the ecosystem -- they have the servers, TPU servers and TPU data centers. And they have their own software. Nvidia also has its own software. Amazon creates one as well with its Trainium and Inferentia chips, and also have their own software environment as well.
And probably Amazon got the biggest boost from Anthropic being so successful because Anthropic trained on Trainium chips and in that environment – which makes them look good. To me, those are platform companies.
Flannery: How does Taiwan fit into this going forward?
Nystedt: Those platform companies all have Taiwan manufacturing their goods, and not just the chips. Wiwynn makes cloud servers and AI infrastructure, and its business is mostly non-Nvidia – only around 10-20%. They're very good at design, and they're very good at creating systems that work around new kinds of chips - AI accelerators. They can work with companies like Google and Amazon to get the production process done. The smoother the mass production process, the lower the cost.
A lot of companies (in the AI area) are doing that in Taiwan now, though Nvidia, came first. Foxconn (Hon Hai) works closely with Nvidia, but also with Google and other hyperscalers; the next two biggest AI server makers are Quanta and Wistron. Most of their volume is for Nvidia. Google is coming on strong.
Taiwan will benefit because more companies are becoming platform companies and putting a lot more services together as well. Anthropic is now going to make its own chip; OpenAI is going to make its own chip, Meta makes its own chip, and Microsoft makes its own chip. Google and Amazon have a head start, but these companies can move fast with partners like Broadcom. Amazon started early by developing their own chips and servers to deal with the workload of their cloud business. Same for Google.
That's the key right there -- they got into it early enough, they have their own design teams, and they have their (supply) partners already set up to do what they want. They understand how to make the software and the chip work very well together, which is the key to a successful platform.
And they all partner with Taiwan suppliers so no matter who wins, Taiwan benefits.