VanEck's Q2 2026 Market Outlook: Time To Bargain Hunt | Jan van Eck

And frankly, all of these, and you know, I'm not usually bullish on everything. All of these are things that I think should be considered for your portfolios in the second quarter because I think they have reached bottoms or are bottoming.

It's great to be with you again, Adam. For those of you who haven't seen us do this before, just a quick intro. I sit with the investment team. It's Friday afternoon in New York, so you're not going to see a lot of people walking around, but I get a lot of ideas from my investment colleagues. And what we've done is pick the most interesting charts to illustrate what we think is maybe missed in the media but very important to your portfolios. One of the things that strikes me is how much changes in a three-month period. I know it probably feels like a three-minute period. Things can change a lot in the markets, but still profound things can change. So this quarterly cadence has become a really fun routine. And Adam, I really appreciate you having me on. This quarter I have a colleague for the first time. I've invited Jonathan Wang to join. Jonathan is our CTO. He has an academic background in AI before it was cool, and he engages with researchers, academics, and clients because a lot of people are trying to figure out how AI is being used. I'm going to share some slides about VANC's usage of tokens in the first section, and he's here to correct me when I misspeak. So let us get to the slides. And for those who haven't heard before, Adam makes these available on his distribution afterwards, so you don't have to take screenshots or anything.

Okay, so here's what we're going to cover really quickly today. Believe it or not, more big surprises from the technology known as AI. I always call it the second most important technology after the railroad. We've got, as we talked about three months ago, Adam, the emergence of real value in the private credit space due to some bad headlines and concerns. Gold is another thing to look at for your portfolios, and Indian Bitcoin. Frankly, all of these, and you know, I'm not usually bullish on everything. All of these are things that I think should be considered for your portfolios in the second quarter because I think they have reached bottoms or are bottoming in corrections and bull markets. So let's talk about our investment philosophy really briefly. One could say, why talk about themes at all? The market is efficient, just buy the S&P. And the premise of this presentation and a lot of our macro thinking is that's true, but the market is shaped by non-market forces. Technology, government policy, and often these forces are more powerful than market participants will model into their future earnings. And it's because of the behavioral weakness of recency bias. We look backwards more easily than we look forward. So anyway, that's the premise for thematic investing. So what happened in the first quarter in AI? Some of this is not going to be news flash to anybody. The demand for tokens, which are the compute units for large language models, has continued to grow at a dramatic pace. We predicted this. People were still surprised by the capital spending that the hyperscalers announced in the first quarter. I'm not going to talk about the incredible destruction happening to software company stocks, but that was a shock too in the first quarter. The real insight that I want to focus on is the path to monetization for AI language models. So that's really the thesis here. So just quickly, here's the explosion in capex this year. The big argument and the reason we like semiconductors here is that there's this big skepticism in the market that these bars are going to go up, and what we've been saying, Adam, is that it's locked in. You're going to have big capex for 26 and 27 for sure, and I think the reason Nvidia is so cheap is that people are like, yeah, but then it's going to roll over in 28, and we just think this trend is so big. We just think that's very implausible. So we think these lines continue to go up. The market is pricing it as if they don't, and that's where we see the market.

Yeah, great. I love that question. So two things. First of all, with railroads, there was a bad market structure and a bad regulatory environment. Without going into a ton of details, there were way too many railroad companies that were set up because they really didn't envision the full buildout of a nationwide network. So you'd have a railroad from one city to another. And that combined with people favoring strong regulatory action over railroads because they felt so dependent on them. There's one railroad that comes to your town, no one's ever happy paying the price of whatever the ticket was. And so they translated that into political power to regulate the prices of railroads, which actually caused the pressure and bankruptcies of a lot of railroad companies because when the airline came along and when trucking came along, trucks competed with railroads but they were subsidized because the government paid for the roads. So they got the roads for free and they were capping the prices of what railroads could charge. So I would say it was a market structure. I would also say there are a lot of industries where there wasn't overbuilding of capacity. You know, we always think of the tech bubble, but think of electricity in the United States. There wasn't an overbuilding of electricity in the 1920s and 30s. That was built for what was needed. So listen, I'm not saying it can't happen. It's clearly a risk and investors should price in risks. But it's just not my base case. Does that make sense?

Okay. So that increased capex, no matter how rich these hyperscalers are, was beginning to hit the free cash flow because these are cash gushing machines, right? We've talked about this before, what great stocks, Google, Amazon. But now because of their buildout of their capital expenditures, their free cash flow from existing businesses less what they have to spend on these big data centers is becoming a bigger and bigger percentage of their sales, as you can see in this chart. And that's hit their stocks in Q1. So that's a little bit for the negativism, but I think that's filtered over into the semiconductor market. Again, this isn't new news to the market. I'm getting to the punchline here. Nor is this new news, right? Token demand continues to grow industrywide. This is China and non-China. A lot of token usage is still closed models. So this is the big insight of Q1: who's going to pay for this big AI buildout? It's not going to be the consumer. It's going to be corporate America. So this is the Anthropic run rate revenue which started really jumping in February and March of this year. And why not? Because chat was a very consumer-oriented application. Hundreds of millions of people adopted chat when it came out in 23 in the US, India, globally because it's free. Who doesn't want to use free products? But consumers are notoriously cheap and they don't want to pay for things. But Anthropic's increase in revenue really showed that corporate America was willing to pay for AI. And I'm one CEO and I'll tell you I'm very happy to pay my AI bills. So you can just see this is a snippet of data that the number of companies spending over a million dollars annually just on Anthropic is doubling, and I would expect it to continue to grow. And this is just another stat on companies that are happy to spend more on technology and more on AI. So the people with growing IT budgets, like Jonathan my colleague, has a growing IT budget even though we've saved a lot of money in a lot of areas. AI is really becoming part of his annual budget, and that's true for a lot of companies and that's going to continue to grow. There's no way corporate America doesn't spend a lot on AI because the payback, the return, is huge. The time saved for VANC employees using agents is dramatic. We can't quantify headcount per se, but anecdotally many people are saying I've got AI to do in minutes what it used to take me days to do. So whether it's better quality assurance, reconciliation of data, creating things, coding, there's a lot of time being saved.

Yeah. And that's why I thought it'd be interesting to show VANC data just as a use point so you could actually see it.

So here's the last market slide and then we'll get to Jonathan's world. Nvidia has been a dead stock for nine months at least here. I was very high conviction. I have been bearish if you recall in the summer of 24. I thought it was way overvalued when it was up here. Oh maybe it was 23 when it was at these nosebleed valuations. At what point it was 50 times sales and I was like that's crazy. But anyway, very comfortable buying it, still very comfortable, enthusiastic because of this huge continued compute spend that we anticipate. And why are people spending money? Because they're getting productivity out of it. So let's look at this. VANC in 2024 got an enterprise license for chat. And so you can see it went from basically zero to a bunch of usage. And then in February of this year we got an enterprise license for Claude for Anthropic. So we started with OpenAI and then we got an enterprise license for Claude. What enterprise license means is that anyone at VANC can use this software, sign up, start saving their work, start sharing agents, developing, using agents, giving the agents more skills. So that's what happened in February. You can see immediately a decline in our estimated token use of chat because we're now using Anthropic. All right. Well, you would think that our estimated combined would continue to go up and to the right. That's what every AI investor is talking about. And I was really surprised to see this chart and that's why I'm really excited to talk about it. Like why is our estimated token use not continuing to go up? Because I know a lot of people have developed agents that they were using regularly. So I asked Jonathan, I said Jonathan what is going on with our agent use? And here is a chart of our API calls, which is basically AI software asking for data, which means it's working. It's a metric for judging how much data these AI agents are crunching, and surprisingly it's kind of flat. Right. I mean, isn't that shocking?

Well, me too. So just two hours ago when I was looking at that, I was like, call Jonathan, what's going on? Something's wrong here. So let me go forward one slide, but then I want to go back. So I said that estimated was the key word in that prior slide. What this shows, this is a very short period of time but from February when we turned Claude on, first of all the actual is like a multiple, an order of magnitude different difference of how much compute VANC is using compared to what our estimate was. Our actual usage here, daily usage is measured in the billions of tokens, Adam. We're one little firm. Billions of tokens. And it's gone up from 500,000 to 500 million to a billion range up in a matter of days, right, to two billion. So this is more the traditional narrative. And this is accurate data provided by Claude by Anthropic. What's interesting is that OpenAI does not give us the actual data according to Jonathan, which irritates him and other people in the industry. Jonathan, the mystery of our token usage of Claude is going up like Adam and I would have thought. What's going on here? Why are our API calls not going up to the same dimension?

That's to you.

Well, I think the bottom line is, I didn't put the costs of the software in this presentation although I was really tempted to, Adam. But I'll tell you that we spend about $350,000 a year on chat and about the same on Claude. And I know some companies are incentivizing people to use more compute. There's always the budgetary impulse to use less, but I can tell you the amount of time that we're saving, and I don't know about headcount, but as a single CEO, I am more than happy to spend those bills. And that's really the point. Anthropic's revenue, OpenAI's revenue, it has to increase and it will increase from corporate America. It's not going to increase from the consumer. That's the hugely big insight from Q1. Perplexity did the same thing. They pivoted towards corporate use and their revenue started going up, and I think the revenue will go up a lot more because we're happy to spend more on AI and we get hooked to it as well. No one's going to go back and start reconciling manually. There's almost an infinite amount, not infinite, but a significantly greater amount that we would spend.

Yeah, I mean that's why I thought the VANC timeframe would be interesting. Obviously some companies adopted this earlier than we did and some are adopting it later.

Yeah. I think we've been very aggressive because I look at it as R&D as well because we invest in the space. So it's like, Jonathan's talking to my investment colleague sitting near me as well. So it's maybe, but I think some companies are super aggressive too. Tech companies are obviously spending a ton on AI.

Yeah, I mean it's significant. I don't know. It's lumpy because it takes the champions to build the agents and add the skills. And then that has to proliferate across the department. But we're a growing company. We're lucky that way. And so we definitely need less headcount. Now, some of the things that we do like our summer internship program, we have 30 summer interns, we do just because we think it's good to build a pipeline of human capital. But what those people are going to be doing is going to be very different. And I think that's what's dramatic. In two or three years, it's going to be so different what people at VANC are actually spending their time on during the day. But look, that was true with the PC revolution, with word processing, email, and Excel. A lot of people spent a lot of time and they weren't talking about it in the media at that time. So yes, I can't quantify it, but yes.

The market's awesome. It's pricing a lot of it in. It's just that this inflection point where corporate America is opening up its checkbook is more visible in the private companies of Anthropic and Perplexity through what they report rather than necessarily in the stock market. But the mayhem that's being caused for traditional software companies, that's the other beauty of AI. You used to do software projects in months and years, and now it's faster. Stuff can be spun up. It's not just that it saves you time, you can do so much more faster. The component one thing I want to emphasize while Jonathan's here too is I say it's data, software, and people. The use of technology in corporations has often been the people part, the adoption, getting the subject matter expert to tell the technologist how to use it and what they need. And I'll emphasize the positive, which is data. Clean data is super important. I shared an interview with the CEO of PNC Bank who talked about how much he's invested in data over the last decade. He was frustrated because when he did acquisitions, there was so much work to do to get the systems aligned. He said, 'I'm going to get my data organized so I can do acquisitions seamlessly.' He spent $2 billion a year for a decade cleaning up his data. So that's where the IT budgets are shifting to, and the corporate budgets are shifting to as well because without data, there's no AI.

Yes, it's a loop.

AI can help, but a human still has to design the program. You have to have the vision and then yes, you can enlist Claude to help you do some of the work.

All right. I'm going to move on if that's all right, Adam.