对谈VanEck CEO：存储芯片股是供需错配泡沫，多数加密项目五年后消亡

Compiled & Translated by: Odaily TechFlow

Guest: Jan van Eck (CEO of VanEck)

Host: Wilfred Frost

Original Title: Memory Is A Bubble, But Nvidia Protected – Jan Van Eck On Semis Surge

Podcast Source: The Master Investor Podcast with Wilfred Frost

Air Date: May 27, 2026

Editor's Note

In this episode, we welcome Jan van Eck, CEO of VanEck. His core thesis is that Nvidia has transformed from a single GPU maker into the "mainframe" of AI infrastructure, boasting a moat built on software ecosystem, scale, and power efficiency. However, the surge in memory chip stocks looks more like a bubble driven by periodic supply-demand imbalances.

The head of VanEck, which manages approximately $225 billion in assets and was among the first to push for a Bitcoin ETF, distills the main themes of the next decade into three things: AI compute infrastructure buildout, the rise of India, and excessive fiscal debt in developed economies like the US, UK, and Japan.

More striking is his characterization of 2026 as the "Year of the Corporate Controlled Chain." He believes Wall Street will absorb the best aspects of blockchain, stablecoins, and programmable money, but most crypto projects and software will become irrelevant in 5 to 10 years; Bitcoin, stablecoins, and blockchain will remain, while many token ecosystems will disappear.

Key Quotes

AI, Semiconductors & Memory Chip Stocks

• "From an AI perspective, the equation is simple. The demand for compute power is here, but the supply is below. Semiconductors are clearly at the core of this structure."
• "Nvidia is no longer just a GPU manufacturer; it's more like the mainframe for AI. The old cyclical and highly competitive nature of a single-chip maker is not the whole story for it anymore."
• "Nvidia's advantage comes not only from its manufacturing scale but also from producing chips with higher efficiency per dollar of electricity. With a forward earnings multiple in the low 20s, I still think it's a solid asset in the portfolio."
• "The profit explosion in memory chip stocks isn't primarily because they sold more products, but because prices increased. This means companies using memory chips will start looking for ways to use less."
• "I don't like calling tops easily, but I'm cautious on memory chip stocks because, in the medium to long term, they don't have as deep a competitive moat as Nvidia."

ETFs, Active Management & Asset Allocation

• "VanEck's investment philosophy is to look back from a decade in the future: By 2036, which major themes will have truly changed the world and financial markets?"
• "ETFs are a scale game. The larger the assets, the better you can serve a broader client base. Much active management, especially private equity and hedge funds, can be subject to diseconomies of scale."
• "Even if the ETF vehicle itself is passive, deciding which ETFs you own, how to allocate, and when to increase or decrease positions is inherently a very active decision."

Macro Debt, Gold & Hard Assets

• "If the market truly loses confidence in the US government's ability to meet its obligations, I don't know where to hide. Even if gold is a medium-term hedge, it could be sold off in the short term along with everything else."
• "I believe gold is in the process of becoming the world's number one currency again. If not the US dollar, I don't think China or India will become the international reserve currency."
• "The government bond market is one of the strangest, most inefficient markets in the world. It can get locked into a certain mindset and become disconnected from reality."
• "The nuclear energy ETF grew from under $20 million to $47 billion, reflecting a very dramatic policy shift. Both US political parties, as well as countries like Japan, have re-embraced nuclear energy."

Crypto, Stablecoins & the Corporate Controlled Chain

• "I call 2026 the 'Year of the Corporate Controlled Chain'. Banks, trading firms, and financial institutions want to absorb the best parts of blockchain but still control their own ecosystem."
• "I think we're experiencing a crypto winter, and it's not coming back. Many projects and software will no longer be interesting or alive in five to ten years."
• "The concept of blockchain will remain, stablecoins will remain, Bitcoin will remain. But many other parts of the ecosystem, in my view, will disappear."
• "The stablecoin bill, for the first time, gives tech companies the ability to compete with the banking system. But banks have survived competition before, like from money market funds in the past."

India & the SpaceX IPO

• "You can't fight demographic trends. India, under Modi, continues to push pro-business reforms. There's no reason why such a country shouldn't grow at a faster rate."
• "SpaceX is massive. As an ETF issuer, we are very happy to see it entering the public market. The liquidity flowing into the economic system will be on the order of hundreds of billions of dollars."

The Frenzy in Memory Chip Stocks

Wilfred Frost: Our guest today is Jan van Eck, President and CEO of VanEck and its affiliates. VanEck, an asset management firm founded by his father, has become a significant player in the ETF industry, managing about $225 billion in assets. Jan is a frequent podcast guest known for his direct and clear opinions, which is why we warmly welcome him. Jan, welcome to the show.

Jan van Eck: Wilfred, it's great to be doing this episode with you for the first time.

Wilfred Frost: I want to start with one ETF. Fair to say it's driven a lot of your performance in recent years and sits at the center of the current market: SMH, the VanEck Semiconductor ETF. It's had an incredible run lately. I understand it's now around $65 billion in AUM, correct?

Jan van Eck: It's roughly that size.

Wilfred Frost: It's become the primary vehicle for investors wanting semiconductor exposure. Up 58% year-to-date, up 135% over the past 12 months. More astonishingly, an annualized return of about 29% since its inception.

Jan van Eck: It's crazy, isn't it?

Wilfred Frost: Truly incredible. Doing that on a compounding basis is very difficult. You could retire now.

Jan van Eck: Yes, should probably call it a day.

Wilfred Frost: But I assume you won't, which is why you're here. Reflecting on SMH's growth to $65 billion over the past year or so, how much of that was price appreciation versus inflows?

Jan van Eck: A huge part was price performance. I'd be hard-pressed to imagine inflows being more than 10% to 20% of that increase over the last 12 months.

Wilfred Frost: That's interesting. I would have guessed a higher proportion of inflows. What do you think is driving it? Maybe a simple question, but is it purely the AI theme?

Jan van Eck: Yes. VanEck's investment philosophy is to try to look at things from a big-picture, macro perspective. I call it '10-year macro.' It means asking, by 2036, which themes will we say have most profoundly affected the world and, consequently, financial markets? This perspective aims to filter out a lot of noise.

I think at least three things will remain: AI, the rise of India, and the excessive debt being led by the US, UK, and Japan. From an AI standpoint, the logic is simple: demand for compute power is very high, and supply can't keep up. Semiconductors are clearly at the center.

Drilling down, you get to Nvidia. The reason our ETF has outperformed other semiconductor ETFs is partly that it only focuses on the top 25 stocks and allows the largest holding to go up to 20%. So, to a large extent, it hitched a ride on Nvidia.

Nvidia itself could be a whole show. Are we still comfortable with semiconductors, with Nvidia today? My own answer is yes. No one can guarantee a company won't lose its competitive moat, but I think Nvidia will certainly still be one of the leaders ten years from now. Partly because it's become like the mainframe for AI, no longer just that single-chip or GPU manufacturer of the past. That old business was not only cyclical but also highly competitive.

Now Nvidia has software, cost advantages, production scale, and higher power efficiency. In other words, more efficient chips per dollar or pound of electricity. Its forward earnings multiple is only in the low 20s. So, even though Nvidia hasn't been the hottest stock in SMH over the last nine months, I still think it's a very solid part of the portfolio.

Wilfred Frost: According to your latest disclosures, Nvidia is about 17% of SMH, TSMC around 9%. I want to dive deeper into them later. You mentioned the advantage of having significant Nvidia exposure; it's also interesting that, at least this year or as you said the past nine months, performance hasn't been driven solely by mega-caps like Nvidia. Many semiconductor companies were actually left behind by the AI theme until recently catching up.

Jan van Eck: Absolutely. In SMH's methodology, there's some thought and some luck. When you only pick the top 25 names during this particular investment era of the last 15-20 years, what happened was that large-cap stocks truly led the market. There are certainly over 100 semiconductor companies, and filtering out the bottom ones in more intensely competitive spaces essentially removes the drag.

Of course, this doesn't apply to all investment phases. But during this period, it amplified the impact of these big winners.

Wilfred Frost: Looking short-term, up 58% year-to-date, the rally has clearly broadened significantly. Memory chip stocks have seen a massive surge. Is this sustainable?

Jan van Eck: I doubt that kind of performance is sustainable. We just saw historic performance in May, so I don't think it will continue at this pace. But I don't necessarily think the market pricing is irrational. Going back to the super macro view, if demand is high and supply is low, capital markets are essentially telling entrepreneurs and corporates: come here, we need your capital, we are willing to value your capital, because we need to build AI compute centers. That's not surprising.

I think this ten-year perspective is useful because humans naturally tend to look backward. When a major trend appears, whether it's the rise of a country or a significant technology, we can't just look at company earnings from the last few quarters or the past uses of that technology to understand its scale of buildout.

Of course, not all tech trends pan out. There are many fake waves, fake technologies. But AI is clearly grabbing global markets by the neck and shaking them.

Wilfred Frost: Another short-term question. The KOSPI hit another all-time high today. It's tripled in the last 18 months, astonishing for a country index, driven mainly by Samsung and SK Hynix. The Korean index rose 12% in one day last week. Does this remind you of the other side? Like late 2021 with meme stocks surging, followed by a big correction in 2022. I know these memory chip stocks, especially those two, have phenomenal EPS estimates, so it's different from meme stock mania. But are there any similarities that raise a red flag?

Jan van Eck: Within the AI ecosystem, I would say there are indeed some bubbles. Going back to the end of last year, the question was the financial sustainability of the OpenAI ecosystem. Is OpenAI, one of the leading model companies with ChatGPT, facing challenges from Claude? Within what I call the 'OpenAI ecosystem,' companies like Oracle, leveraged by building compute for OpenAI, and CoreWeave both fell 50%.

So even within the larger AI trend, you find local bubbles, or company-specific bubbles. Getting back to your question, I do think the memory segment is a periodic moment. One doesn't want to call a top at times like this, but I personally remain cautious on memory chip stocks because, in the medium to long term, they don't have the competitive moat I believe Nvidia has.

There will be new entrants in this space. There's definitely a shortage right now, which gives them pricing power. The main reason for their profit explosion isn't a huge increase in sales volume, as they have capacity constraints; it's because they raised prices. This also means that companies using memory chips will start looking for ways to use less.

So I agree with your feeling that it feels very bubble-like. In our active funds, we are reducing exposure to the memory segment.

Wilfred Frost: Nvidia is about 17% of SMH, followed by TSMC, then US giants like Intel, Broadcom, AMD, Micron, Texas Instruments, Qualcomm, each around 6% or 7%. Does TSMC also have a defensible moat similar to Nvidia's? Different in type, but is the defensiveness comparable?

Jan van Eck: I think so. TSMC not only has manufacturing capability but also the capital capacity to build extremely expensive chip fabrication facilities. I suspect one common advantage for both Nvidia and TSMC is that they work with a wide range of ecosystem participants and can see almost all their clients. Therefore, they can see where technology is going and how customer demands will evolve. Most people would say TSMC will still be there in ten years; it will be a survivor.

Wilfred Frost: You mentioned earlier that Oracle or CoreWeave saw significant drawdowns from their late October high to the 'Iran war low' in March, with Oracle nearly halving, significant for its size. I heard you say on another podcast not to worry too much about an overall AI bubble because it already popped to some extent. The question is, in these moments, how confident do you feel re-entering the right companies, especially when a large part of the discussion involves companies not yet public, so investors can only play with proxies?

Jan van Eck: This sounds very much like an ETF issuer's answer, but from a company perspective, a diversified approach is certainly more reasonable. In terms of timing, if you're in the midst of such a trend, it's better to buy on pullbacks rather than chasing it right now. We talked about SMH's flows earlier; I think a lot of the assets in this fund came from investors who bought years ago and let the appreciation happen naturally. That's healthy in a way because there isn't a lot of fast money chasing it.

Of course, money is chasing memory chip stocks and definitely chasing the hottest parts of the ecosystem. But overall, we are still overweight semiconductors in our broad model portfolios, though we're now thinking