Source: Raoul Pal The Journey Man
Compiled by: Felix, PANews
Macro investor and Real Vision co-founder Raoul Pal invited Wall Street strategist Jordi Visser to his podcast. In the interview, Raoul and Jordi analyzed how AI is creating a supercycle driven by an explosive growth in computing, energy, intelligent agents, data centers, and intelligence. They also discussed how cryptocurrencies, tokenization, and the data economy can open up new markets.
PANews has compiled the highlights of the dialogue.
Host (Raoul): Today, my regular guest and good friend, Jordi Visser, is here. This isn't really an interview, but rather a discussion about what's happening and how to assess and seize opportunities. Jordi, how have you been lately? What are you thinking about?
Jordi: I'd like to start with your recent conversation with Julian. You talked about the shift from "labor and capital" to "computing power and energy," which really resonated with me. In the past, growing a business required borrowing money, hiring people, and finding office space. But in the world of "computing power and energy," the rules are completely different. I've been writing about the AI cycle, and in this new world, if we can't manufacture all the chips we need or get enough electricity, it will lead to bottlenecks and shortages due to supply and demand imbalances. These bottlenecks might slow down these companies' profitability, but not because there's no demand; rather, because the demand is simply too great.
Host (Raoul): I've built a dashboard to monitor the exponential growth of intelligent output per unit of energy. In the past it was Moore's Law, but now, thanks to GPUs and AI, it's showing exponential growth on a logarithmic chart (double exponential), which is what I call Reed's Law (the square of the exponent). When you consider that data center construction is only 30% of what was initially projected, the competition between the US and China, and the fact that no single cutting-edge AI company can dominate, you realize this is almost inevitably going to be a "supercycle." These bottlenecks will only slow it down, and to break through them, we need to build infrastructure like electricity, which will be the largest capital expenditure (Capex) cycle in human history .
Jordi: I completely agree. In surveys like the PMI (Purchasing Managers' Index), which examine the business cycle, we often see that a high PMI actually masks supply chain bottlenecks and rising prices, while the new orders index may have already fallen to 50. But let me delve deeper: how do we enhance intelligence without building enough data centers? This is because algorithms, human feedback reinforcement learning, and reasoning abilities have improved. This isn't a simple "build more oil wells to get more oil" process, nor is it "build more data centers to get more intelligence." This intelligence undergoes self-recursive learning and algorithmic improvement, resulting in a double exponential growth. This parabolic linear growth has become the norm, which confuses many people because it's rare in emerging markets. From Jensen Huang's talk about building an AI agent economy at CES in January to the entry of major giants at the Morgan Stanley TMT conference in March, people realized just how massive the numbers are. This phase is different from the purely IQ-enhancing efforts of the past three years; in the coming year, the results of recursive self-improvement will be even more astonishing .
Host (Raoul): But the market can't currently concentrate all its attention and capital on one stage indefinitely; it hasn't grasped the significance of the AI agent economy. In the past, the Total Potential Market (TAM) for business was always humanity, but now TAM is expanding infinitely. Therefore, I believe the market will see sector rotation; it won't always be just Nvidia rising. When bottlenecks like electricity are encountered, all "roadblocks" will be cleared to solve the problem of energy consumption per unit, and capital and attention will concentrate on the bottlenecks. For example, peptides and gene science technologies, which you mentioned before, will also see rotation.
Jordi: Exactly, electricity is a clear bottleneck. This explains why Nvidia and Siemens recently announced a collaboration in China on solid-state batteries, which require a lot of silver rather than lithium. People should pay attention to how much silver solid-state batteries require. If we have energy storage innovations to cope with peak electricity demand, the US power grid actually has enough electricity until 2030. Jensen Huang once divided AI into five layers: the bottom layer is energy, chips, and infrastructure, then models, and the top layer is the application layer. But at the application layer, everyone is always focusing on the old-fashioned SaaS (Software as a Service) model. Currently, the application layer is actually attracting the most funding for "human software," which is the capital that Eli Lilly (a US-based multinational pharmaceutical company) has attracted through its GLP-1 weight-loss drug. They have data centers with thousands of GPUs in their campus, and I believe the cash flow generated by GLP-1 is actually financing the next stage of human bioscience software development .
Host (Raoul): This is similar to how Musk's used car business generates cash flow to support new businesses. Speaking of overcoming bottlenecks and improving efficiency, to address the global copper shortage, Musk upgraded the Cybertruck's voltage architecture from the traditional 12V to 48V, directly reducing copper usage by 70%. When given capital and attention, with the intelligence of humans and AI, we can always find ways to circumvent obstacles. For example, when faced with oil bottlenecks, we invented shale oil extraction; when faced with data center bottlenecks, we switched to fiber optic transmission.
Jordi: People haven't realized the significance of the AI agent world yet. Imagine if someone said in January that the Earth's population would suddenly increase by 7.5 billion; we'd feel resources would immediately run out. But the AI agent world brings billions, even tens of billions, of new "thinkers" into this world, and they only consume one thing: computing power. Digital employees won't need to buy houses or send their kids to college; people will have to change their old business cycle thinking. As Musk said, if billions of agents tirelessly solve our problems, we will enter an era of abundance where humans can choose whether or not they still need to work. These massive numbers of AI agents are running countless "Manhattan Projects," ultimately solving all problems .
Host (Raoul): And this isn't a single, large model thinking. OpenAI has 1 billion users, and each user is using a different instance of this vast intelligence, which is exponential in terms of intelligence enhancement .
Jordi: To support myself, I started building my own "knowledge brain." I transcribed Huang Renxun's speeches and even uploaded and compiled transcripts of several hours of speeches by Eli Lilly CEO David Ricks. Using the professional knowledge of a group or individual as material for AI work is much more focused and profound than directly searching the entire internet .
Host (Raoul): I'm also building my "GMI Brain" using a vector database, which contains all the long-form content I've written, video transcripts, and tweets from the past 21 years. I've also built a tool called "Lens," based on my Exponential Era framework and first principles, capable of analyzing everything from the US election to the market. But the problem I'm facing now is that I'm doing everything myself, and I simply don't have enough time .
Jordi: I worked at a hedge fund for 20 years. After closing it, I decided to stop working for or managing anyone. Asking investors for money and staying up late explaining trading strategies took away the fun. I realized that 8 billion people worldwide didn't know how to navigate the AI era, so I focused on content creation, using easy-to-understand language to help people understand trends. I have no employees, only one assistant, but the business is growing very rapidly and steadily. This is thanks to using AI agents to handle everything. Building an AI business has amazing profit margins and extremely low costs; you can achieve rapid, linear, or even explosive growth.
Host (Raoul): So how do you squeeze out time to do these things? I feel like all my time is taken up by learning new tools.
Jordi: I have more time than you because I don't need to handle as many interviews, business trips, or manage company affairs. And I often let ChatGPT help me simplify things. I'll ask it, "Should I spend time on this new tool?" and it usually tells me, "Don't worry about it, it'll be easier next month," which saves me a lot of trial and error time.
Host (Raoul): That makes sense. I'm currently focusing on building the underlying database engine, what I call my "personal vault." It will store all your personal files, photos, and phone recordings, becoming your personal operating system, which could be used for personal brainpower or monetization in the future. Many people haven't yet realized how disruptive this will be once every piece of information on their computer and phone can be retrieved by AI in a second .
Jordi: Absolutely agree. The most typical example is dealing with estate documents as an executor when a loved one passes away – it's a huge nightmare. I've been going through that for the past 18 months, and I've compiled all the documents into one folder and connected it to Claude Opus 4.5. When anyone calls to ask about the status of assets or documents, Claude can immediately find the accurate answer and send it out, demonstrating the power of a future personal assistant and database .
Host (Raoul): I'm currently using a tool called Granola, which not only performs real-time meeting transcriptions but also integrates with large models to become my long-term knowledge base. All conversations are fed into my knowledge brain, and it never forgets what we discussed last time. The biggest bottleneck for AI companies right now is "insufficient memory persistence," and this long-term database layer breaks through this limitation.
Jordi: I currently run OpenClaw on my Mac Mini using the Chinese model Kim K2.5 system, and also GPT-5.5 on my top-of-the-line M5 chip Apple laptop as my assistant and to run my portfolio technology algorithms. It helps me recall those inspirational notes I took on a plane two weeks ago.
Host (Raoul): To avoid the friction caused by exporting data back and forth between different devices, I also just got the top-of-the-line M5 Apple laptop, and now I take it everywhere I go .
Jordi: By the way, if you don't hire people, the cost savings and incredibly high profit margins of AI will tempt you to invest heavily in hardware. I'm planning to buy an Nvidia DGX to run the Hermes agent. After listening to Dennis Casabus's interview on Y Combinator, I'm convinced that open-source models will become smaller and better, and eventually we'll move towards "edge AI." In the future, running local models on your own devices and computers will be an extremely important part .
But learning to use AI (especially on edge devices) is like learning to ski or golf; you have to abandon old habits and put in a lot of practice. When I get an inspiration while listening to a podcast on a walk, I immediately pause, jot it down with Whisper, and then generate a draft of an article using ChatGPT. So people have to buy the best phones and computers, treat them as an educational investment, and use them constantly while walking, driving, and flying, figuring out their own workflow .
Host (Raoul): This kind of 24/7 thinking certainly generates a lot of inspiration. I've been focusing on government debt and healthcare longevity. American households have a net worth of $180 trillion and debt of only around $40 trillion, so it's not a huge problem. But what I'm more concerned about is how AI will profoundly impact this welfare system as life expectancy increases and healthcare spending rises as a percentage of welfare spending.
Jordi: The debt-to-GDP ratio will collapse. Regarding welfare and an aging population, while AI can certainly help reduce welfare costs, the more difficult challenge is helping the elderly rediscover their sense of value in their communities or the economy. The boundaries of "work" are already blurred; many YouTubers' podcasts are designed to attract human attention—a typical example of work in the post-AI era. Regarding anxieties about the future of human work, I recommend reading *The Daily Stoic*. For millennia, humanity has been anxious about "what if we don't have to work anymore," but we always find ways to adapt and evolve.
Host (Raoul): Absolutely agree. Regarding debt resolution, another key point comes to mind: tokenization. Two-thirds of the world's vast assets (such as real estate, private equity, venture capital, and art) are not tradable. Once these assets are tokenized, they will bring transparency and liquidity to dormant assets, and with increased liquidity, GDP will inevitably rise. Therefore, tokenization is a key focus for me in addressing issues of longevity, ownership, and welfare.
Jordi: I've written a series of articles about the "invisible economy" (i.e., the AI agent economy). Cryptocurrencies are essentially machine-readable packets of information. Google processed trillions of tokens last year, and now it's in the trillions. To train AGI to ASI (super artificial intelligence), they need to absorb all digitizable information, including university and scientific data. The largest market on Earth will no longer be the human asset market, but the data market that AI craves. In the future, countless AI agents will engage in millisecond-level API call transactions—an astonishingly large, yet invisible, invisible trading market.
Host (Raoul): That's an excellent perspective. Because of this major trend, we need to re-examine the concept of a "bubble." A bubble is a combination of price and time. If the Big Seven tech companies (Mag 7) grew to $20 trillion in a year, that's a bubble; but if they grew twenty or thirty times in 15 years, that's a structural dividend. People always think that if prices rise too fast, it's a bubble, but in reality, the profit growth of large companies is now synchronized with the soaring stock prices, and the price-to-earnings ratio has even decreased. Indicators across all industries are rising in a hockey stick-like straight line .
Jordi: Experienced traders who know too much are more likely to miss opportunities. They carry too many memories of the "dot-com bubble" or the Great Depression of 1929, causing them to make incorrect assumptions and comparisons to the current market. It's like writing code using a Claude model; sometimes admitting you don't know anything and abandoning unnecessary context can actually make the system work better .
Host (Raoul): I'm curious about your perspective. Currently, earnings prospects for AI hardware and infrastructure-based stocks are very good, which is detrimental to narrative-driven assets like Bitcoin and cryptocurrencies because funds have been drawn away. Many people haven't grasped the explosive potential of the AI agent economy, and now large funds managing trillions of dollars are all on board, enjoying very stable profits. This will not only lead to the normal rotation of funds towards longevity-themed drugs and other sectors, but also pose a challenge to the crypto market. When traditional stocks offer such huge returns, what forces will cause funds to rotate back into the crypto market?
Jordi: First, even giant stocks can stop surging due to insufficient liquidity and attention; we saw this rotation between 1995 and 2000. As you mentioned before, "infrastructure bottlenecks" will cause these AI giants to experience a period of stagnation and consolidation as production capacity cannot keep up with demand. When funds overflow, decentralized identity (ID) and blockchain technology, which meets the attributes of AI proxy transactions, will once again demonstrate their absolute advantage. I am particularly interested in Layer 1 protocols. Although some believe that AI will disrupt the existing SaaS software ecosystem, I think that people are still embedding AI into existing billing and accounting software every day through APIs, and software still has its value .
Host (Raoul): The AI hardware world does indeed face limitations in commodity production. For example, semiconductors require specific lithography gases and petrochemical raw materials. Once production encounters bottlenecks, the expansion of hardware infrastructure may not reach the expected scale, which will squeeze out some of the valuation bubble in Capex. I believe there are two prerequisites for cryptocurrencies to usher in a "third wave": First, bottlenecks in AI physical infrastructure lead to capital flight, seeking safe targets based on the digital world that only require capacity and not physical expansion (such as AI agent application layers); second, tokenization gives traditional, large, dormant assets liquidity, causing institutional funds to re-question their holdings and seek entry into the crypto world. We are currently in a bottleneck digestion period that requires great patience. What are your thoughts on the recent large-scale IPOs of some tech companies?
Jordi: Companies like Google are raising funds at this time because it's a battle for limited capital. Companies like OpenAI, Anthropic, or SpaceX can't borrow through the credit market; they can only draw capital from the stock market. I think these three mega-IPOs may mark the peak of infrastructure capital expenditure (Capex) transactions in the short term, but not the peak of the overall market. Funds will flow to other targets such as software.
Host (Raoul): Yes. As I mentioned in my report, we just went through an "AI capital expenditure buffet," and everyone ate their fill. Now, a consolidation and digestion period of 3 to 6 months is needed to reset capital. As these underlying major trends continue to develop, if stocks like Nvidia stop rising, the probability and speed of funds flowing back into assets like cryptocurrencies will increase significantly.
Jordi: Cryptocurrencies do indeed need this digestion period. It's similar to the launch of gold ETFs. The approval of Bitcoin ETFs and the support of the US president prematurely drew up massive demand, and with global liquidity expanding by only 10% annually, the capital pre-borrowed for a year inevitably needs a year to be digested. After the devastating bear market of 2022 and the emotional release in 2024, this painful consolidation period is unavoidable.
Host (Raoul): But when we go to large-scale crypto conferences like Consensus in Miami, with tens of thousands of attendees, you'll find almost no retail investors (although not many can afford the expensive tickets). The venue is filled with major banks, financial institutions, and traditional players focused on stablecoins and asset tokenization. This proves that the underlying technology trend is extremely strong. Once the infrastructure benefits of blockchain space begin to materialize and are monopolized by a few core players, the price of tokens will inevitably rise.
Jordi: Many people unfortunately get lost in the "trees" of short-term trading, ignoring the whole "forest." If you've read Elliott Wave Theory, you'll know that we're waiting for the "third wave" that will make you big money. Just like how I caught Micron Technology's surge, I'm patiently waiting for the arrival of that "third wave" (the so-called banana zone) in the crypto market.
Related reading:Interview with a Wall Street strategist: AI-driven deflation accelerates capital flows to scarce assets
