华尔街资管公司CEO公开持仓：押注纳指和BTC，「中间地带」才是最差选择

Compiled & Translated by: Odaily TechFlow

Guest: Anthony Pompliano, Founder & CEO of Professional Capital Management

Host: John Pompliano, Associate at Pomp Investments

Podcast Source: Anthony Pompliano

Original Title: I Just Revealed My Current Portfolio…

Release Date: June 24, 2026

Key Takeaways

Anthony Pompliano and John Pompliano are back together for this episode, diving deep into the brutal sell-off of the Magnificent 7, the market-wide panic over AI capital expenditure (CapEx), and why the underlying inflation data is actually much safer and more controlled than sensational headlines suggest.

Today's guest is Anthony Pompliano, Founder & CEO of Professional Capital Management, a long-time active figure in the crypto investment space whose podcast is one of the most listened-to financial podcasts in North America.

Furthermore, we will fully review Anthony's current investment portfolio, exclusively analyze the real political chess game behind new Fed Chair Kevin Warsh's interest rate decisions, and deconstruct why, in the long-term journey toward the next decade, Bitcoin's "high volatility" is not a risk but its core ultimate advantage.

Key Insights Summary

The Real Logic Behind the MAG 7 Pullback

• "The core reason MAG 7 is being sold off is that they are essentially long-duration assets, making them very sensitive to inflation."
• "For the MAG 7, these AI companies aren't going to suddenly implode or go to zero."
• "If inflation falls in Q3 and Q4, the valuation multiples of the MAG 7 are very likely to re-expand."
• "Google is now cheaper than Apple in terms of valuation but growing faster. Why is that? The core reason is market concerns about AI CapEx. This actually presents a good opportunity to buy some of the world's best companies at lower valuations."
• "Many value investors have been underperforming for years because they are too fixated on applying past frameworks to today's businesses. You should respect history, but you can't just point to old data and say 'these companies are too expensive' for 15 years straight."
• "A likely scenario is a company first says they'll spend $10 billion on CapEx over the next two years, but actual spending ends up at $6 billion. $6 billion is still a huge number, but it's 40% less than the market feared. The market is overly fixated on inflation direction right now, and I think this is mispriced."

Repricing AI CapEx and ROI

• "The real problem now is token cost. Talk to a bunch of CEOs, and everyone says the same thing: we're spending too much on tokens, but we have no idea where the ROI is."
• "Everyone is pivoting to the same goal: getting the same output with fewer tokens."
• "What are the bottlenecks limiting the adoption of this technology? Electricity, data centers, chips."
• "I saw a news item the other day about a company shifting employee work hours to 1 AM to 10 AM. One reason was they believed calling models during this time was cheaper, system load was lower, and answers were more accurate."

Investment Method: Avoid the Middle Ground

• "If you summarize investment approaches, I think there are roughly two paths. The first is buying broad indices; investing in large-cap indices is a great strategy. The second path is investing in highly asymmetric opportunities. I think the worst place is the middle ground. You shouldn't allocate capital to mid-sized companies with only moderate growth and mediocre return profiles."
• "The two most interesting technologies in the world today are Bitcoin and AI."
• "I own Tesla. I believe if Elon creates a monopoly in humanoid robots and autonomous driving, it will be incredibly valuable."
• "Once he can give hardware these capabilities using AI, machine learning, and computer vision, he essentially captures the embodied entry point for 'Physical AI,' and I think that will be extremely valuable."
• "Where do I want my exposure? I'll hold some cash long-term, and I'll hold Bitcoin."
• "Anduril's model is to let startups first achieve technological breakthroughs. Once the tech is validated, it steps in to acquire these drone companies. It's very good at M&A and has a strong BD team to quickly commercialize acquired technology. In other words, it's a platform that 'buys technology and then commercializes it.'"

His Current Portfolio

• "In the private market, I hold many software AI companies like Replit, Lovable, and Micro One, which are very strong in their respective verticals. In the public market, I'm more concentrated on physical AI and robotics."
• "Whether public or private, software or hardware, I have a relatively complete set of AI exposure."
• "I own Tesla not for the cars, but because I believe if Elon creates a monopoly in humanoid robots and autonomous driving, it will be incredibly valuable."
• "He will eventually merge SpaceX and Tesla, likely before 2030. Once that happens, he'll consolidate his most important projects into a giant conglomerate."

Bitcoin Enters a New Phase

• "Unhappiness is essentially the gap between expectations and reality. So you can't set unrealistically high expectations for Bitcoin. If you set your expectation at 25% to 30%, and it performs better, you'll be pleasantly surprised."
• "Governments will continue printing money. Bitcoin's core investment thesis hasn't been broken. The change is simply that it has entered a different stage of the game."
• "Bitcoin used to be like playing high school basketball, where you were one of the few standout players. Now it's more like college basketball. All players are stronger, the gap isn't as huge, but the quality of the game is higher."
• "Retail is typically more emotional, institutions are typically less emotional. I know the internet sentiment is bad, but from a market structure perspective, this is somewhat a common part of the bottoming process."

MAG 7 Sell-off and AI CapEx

Anthony Pompliano:

Many people worry just because Bitcoin is volatile. I think that's completely the wrong way to think. I actually like it when the things I own go through phases of being unloved by the market. I don't want to hold assets that are always hot, because if an asset is perpetually the market's darling, it means it's too popular, too crowded, and the returns have likely been arbitraged away. Conversely, assets that periodically fall out of favor often have stronger asymmetry and higher potential future returns. So volatility itself is important.

In today's episode, John interviews me. We'll talk about Kevin Warsh, the Fed, inflation expectations, interest rates, the AI CapEx debate, MAG 7, the S&P 500, Bitcoin, SpaceX, and many other topics. I'll also lay out my current investment framework for both public and private markets, and specifically mention some assets I've already allocated into my portfolio.

John Pompliano: Let's start with the first topic. There's been a recent sell-off in the MAG 7. Do you think this is a valuation repricing, or is the market starting to rethink the entire AI trade?

Anthony Pompliano:

I think the core reason MAG 7 is being sold off is that they are essentially long-duration assets, making them very sensitive to inflation. The market was recently worried about the Iran war pushing up energy prices and inflation trending higher, so these assets were sold first.

I've said two things repeatedly throughout 2025. First, tariffs will not cause sustained inflation. Second, as long as the Iran war isn't a multi-year protracted conflict, it's just a short-term price shock. We've indeed seen that short-term spike on the energy side, but it won't evolve into a long-term inflation problem. The bigger picture is that there are still many structural deflationary forces at work in the US economy.

Of course, there will always be people who see inflation ticking back above 3% and scream "inflation is out of control again." I don't see it that way. Many people's understanding of inflation has been biased by the extreme 9%+ episode a few years ago, making it seem like high inflation will keep recurring. But if you look at history, inflation spiking above 9% is an extremely rare event in one's lifetime. If I recall correctly, the last time US inflation exceeded 9% was in the 1970s, before I was even born. In my entire life, it has only happened once.

That period of high inflation was certainly severe, but also easy to understand. The massive money printing and zero interest rates in the US in 2020 made high inflation almost inevitable. However, that kind of artificial manipulation is fundamentally different from the price disturbances caused by the Iran war and tariffs.

If you look at the MAG 7, the market currently has two layers of worry. The first layer is that if inflation continues to rise, these long-duration, rate-sensitive assets will see valuation compression, causing stock prices to pull back. That logic makes sense. The second layer is the fear that AI CapEx is too high, leading to decreased free cash flow, less cash to return to shareholders, and thus lower valuations.

But I believe two things will happen next. First, inflation won't be as bad as the market fears. We're already seeing signs. In the latest inflation data, 60% of the increase came from energy alone. Once energy prices fall, inflation will naturally follow. Oil has already dropped below $80. If it goes back to $60, what do you think happens to inflation? I'm not 100% ready to call "peak inflation," but I think we are very close. Whether it peaked already or will peak in the May/June data, my judgment is that inflation in Q3 and Q4 will be lower than in Q2. If that's correct, the valuation multiples of the MAG 7 will re-expand.

Then there's the AI CapEx part. In my view, this boils down to two key questions. First, does the future demand we think exists actually exist? Second, will these capital expenditures ultimately generate returns? If demand exists, returns are possible because someone will use the infrastructure. Does demand exist? Simple. Do you use AI more frequently today than a year ago? Yes, much more.

But recently, the market and media have started discussing another issue: How efficient is our current use of AI tokens? Is the computing power and energy spent by companies on calling models really worth it? I've felt this personally. Many know we're building a product called CFO Sylvia. During this process, we realized a real problem: costs were going up. Because users can generate their own queries, without a smart enough token management mechanism, expenditure is theoretically unlimited.

When we started building the product, the team's main focus was "get the product out first." Once it was running,you realize the real problem is token cost. So what do you do? Start improving token efficiency.

We made many specific adjustments. For example, on some pages, refreshing would re-query the model. That was totally unnecessary, so we stopped it, immediately reducing token consumption. Some features periodically called the model. While cool, they were major token hogs. We cut them and watched for user complaints. None came, so they were gone permanently. That saved a huge amount of tokens.

Then we adjusted some architecture and started looking more seriously at cash costs. Initially, I thought this was our problem, just ours to fix. But when I started talking to a group of CEOs, they all said the same thing: we're spending too much on tokens, but we have no idea where the ROI is.

So, everyone is pivoting to the same goal: getting the same output with fewer tokens. That's why I was tweeting that the phase of companies blindly chasing token usage and creating leaderboards for "who calls the model more" won't last. The world will eventually return to real, measurable metrics like efficiency, effectiveness, and ROI.

And this is precisely where I'm most bullish on private model companies like OpenAI, Anthropic, and Grok. You'll see customers like CFO Sylvia, or any enterprise client, demanding "I want to use fewer tokens but get the same result." Individual customers are becoming more efficient. Yet, the total addressable market is expanding, and product adoption is rising. So these model companies' total revenue could still be very high. To me, this is a sign of product-market fit. As long as open-source models haven't completely eaten their lunch, these companies are running great businesses right now.

The next question becomes: What are the bottlenecks limiting the adoption of this technology? Electricity, data centers, chips. You'll find constraints everywhere. I saw a news item the other day about a company shifting employee work hours to 1 AM to 10 AM. One reason was they believed calling models during this time was cheaper, system load was lower, and answers were more accurate. I don't think most companies will do this, but it shows people are starting to seriously think about model call efficiency.

If demand for this technology is persistent, it means we still don't have enough data centers, electricity, or chips, and bidding for these resources will continue.

So, those worried about AI CapEx returns can certainly raise a long-term question: Could there be overbuilding in the future? Yes, it's possible in any cycle. It's a historical pattern and hard to control precisely. But in the short term, I don't see this problem at all. I see strong demand for software, specialized workflows, data centers, electricity, and chips – so strong that some companies are even discussing building orbital data centers.

Therefore, I think the market's current anxiety is somewhat misplaced. Especially regarding the MAG 7, many are fixated on what companies "expect" to spend. But expectations aren't reality. Just as expected revenue isn't confirmed revenue, expected CapEx isn't what gets spent.

A likely scenario is a company first says they'll spend $10 billion on CapEx over the next two years, but actual spending ends up at $6 billion. $6 billion is still a huge number, but it's 40% less than the market feared. The market is overly fixated on inflation direction right now, and I think this is mispriced. More importantly, people also place too much faith in those AI CapEx guidance numbers. I suspect the actual money spent will likely be lower than what the market currently worries about.

So look at specific companies within the MAG 7. Google, for instance, its stock has dropped recently, which is interesting. Google is now cheaper than Apple in terms of valuation but growing faster. Why is that? The core reason is market concerns about AI CapEx. If you actually look at the data and fundamentals, I think this presents a good opportunity to buy some of the