Video Title: Kevin O'Leary Reveals His Next Big Bet
Video Author: Stock Sharks
Editor's Note: As AI investment moves from model competition into the infrastructure expansion phase, the market's focus is shifting from "who can develop the strongest model" to "who can provide the foundation for the entire AI industry to keep running." Chips, compute, and model capabilities still matter, but as data center scale expands, power supply, energy costs, and long-term infrastructure contracts are beginning to become key variables affecting AI investment returns. As technological progress gradually becomes a market consensus, a more fundamental question begins to emerge: if it is impossible to determine in advance the ultimate winner of technological competition, how should investors share in the returns generated by the expansion of the entire industry?
In an interview on The Deep End, a podcast under Stock Sharks, well-known Canadian investor and Shark Tank regular Kevin O'Leary (nicknamed "Mr. Wonderful") shared his latest judgments on AI, energy, Crypto, quantum computing, and asset allocation. O'Leary accumulated his wealth in his early years by founding the software company SoftKey, and today his investment portfolio spans technology, energy, and crypto assets, among other fields. Rather than chasing a single hot asset, he pays more attention to the resource constraints behind technological expansion, business models, and whether growth can ultimately translate into sustainable capital returns.
In this conversation, O'Leary is in effect breaking down "what should be invested in next" into a set of more fundamental structural questions: what resources does technological expansion depend on? Who ultimately captures the industry's value? How do new technologies translate into cash flow? When the market cannot determine the ultimate winner, how should investment risk be controlled?
First, the logic of AI investment is shifting from technological leadership to resource constraints. Over the past few years, AI investment has mainly revolved around large technology companies, advanced chips, and model capabilities, with investors trying to identify companies with technological advantages. But as compute demand increases, electricity is becoming an important condition affecting data center expansion. O'Leary therefore places his investment focus for the next three years on energy, with particular attention to natural gas, power infrastructure, uranium mining, and regions such as Canada and the Nordic countries that have low-cost power resources. He adopts a classic "picks and shovels strategy": there is no need to predict which AI company will ultimately win, but rather to find companies that can provide key resources to the entire industry. This means that investment opportunities in the AI industry may further spread from technology companies to energy and infrastructure segments, but whether resource advantages can translate into excess returns still depends on project costs, contract pricing, and execution capability.
Second, Crypto's valuation logic is shifting from market trading to actual adoption. Previously, institutional capital mainly focused on Bitcoin and Ethereum, while other crypto assets faced issues such as concentrated liquidity, regulatory uncertainty, and a lack of clear commercial demand. But O'Leary has begun to re-examine public blockchains such as Avalanche, not because of short-term coin price performance, but because enterprises in fields such as finance, logistics, sports assets, and contract management may need different types of blockchain infrastructure. The question that arises from this is whether future blockchain applications will concentrate on a few general-purpose networks or form multiple specialized networks serving different industries. This change may create opportunities for other public blockchains, but enterprise adoption of blockchain technology does not necessarily mean that the related tokens can capture the corresponding economic value. What truly needs to be verified is network usage, commercial revenue, and the mechanism by which value is transmitted to token holders.
Third, quantum computing is introducing a new long-term security variable for crypto assets. In the past, Bitcoin's main market risks were concentrated in price volatility, regulation, and capital flows. With the development of quantum computing, whether the existing cryptographic system can withstand potential attacks in the future is becoming another issue worth discussing. O'Leary suggests that even if a quantum computer truly capable of breaking encryption has not yet appeared, the market's expectations of this risk may affect asset pricing in advance. However, there is still no definite timetable for the so-called Q-Day, and progress in quantum technology cannot be directly equated with the failure of blockchain security mechanisms. Rather than predicting when Bitcoin will face a threat, he is also watching the positioning of companies such as IBM and Google in quantum computing and security technology. What is involved here is not only Crypto risk, but also the cryptographic infrastructure upgrade that the financial system may face in the future.
Fourth, the investment approach is shifting from betting on winners to managing uncertainty. O'Leary does not deny the appeal of high-growth assets, but his early experience shorting Yahoo made him place greater emphasis on position limits. He insists that a single stock should not exceed 5% of the portfolio and a single industry should not exceed 20%, and he controls concentration risk by reducing positions in batches. This approach is consistent with his investment logic in energy, Crypto, and quantum computing: when the technological path is still unclear, participate in potential growth through diversified allocation rather than letting a single prediction determine the performance of the entire portfolio. At the same time, his analysis of private companies always revolves around customer acquisition cost, retention rate, and cash flow, emphasizing that technology must ultimately be reflected in real operational improvement.
If this conversation is compressed into one judgment, it is this: the next stage of investment competition may no longer be simply about finding the companies with the strongest technological capabilities, but about identifying which resources, infrastructure, and business models can reliably capture value in the process of continued technological expansion.
In this sense, what O'Leary is discussing is no longer just specific assets such as AI, Crypto, or energy, but how the market reallocates capital, measures risk, and determines asset value after the technology industry moves from narrative-driven to commercial realization. What truly determines long-term returns may not be who was earliest in betting on a technological trend, but who can continuously obtain cash flow during the industry's development and avoid paying too high a price for growth that has not yet been realized.
The following is the original content (the original content has been edited for easier reading and comprehension):
If only one investment theme could be chosen for the next three years, Kevin O'Leary's answer is energy.
This may differ from the market's expectations for a technology investor. Over the past few years, the core of AI investment has been NVIDIA, model companies, and large technology enterprises. Investors tried to judge who had the most advanced chips, the most powerful models, and the technology platform most likely to dominate the future.
But O'Leary believes that as the AI industry continues to expand, investors need to rethink where value actually comes from.
His approach is not to exit technology investment, but to reduce dependence on a single technology winner. No matter which company ultimately wins the model competition, it will need electricity, land, data centers, and other infrastructure.
This is also why he has shifted his investment focus to energy.
And in the Crypto market, O'Leary is conducting a similar reassessment: can blockchain assets that previously obtained valuations mainly through market trading and capital inflows prove in the future, through actual commercial applications, that they have sustainable economic value?
The next round of AI opportunities is not in models, but in powering models
O'Leary is not pessimistic about AI's long-term prospects.
He acknowledges that large technology companies created considerable returns in the previous round of AI investment, and investors holding related stocks or indices have already benefited from it. But for the next stage of investment, he is more focused on one question: if it is uncertain which model company will ultimately win, is it still possible to participate in the growth of the AI industry?
His answer is to invest in energy.
O'Leary calls this Picks and Shovels. In a gold rush, rather than predicting who can dig up gold, it is better to invest in companies that provide tools to the gold diggers.
The AI industry has a similar commercial structure. Model companies need chips, chips need data centers to run, and data centers need large amounts of stable electricity. Therefore, whether OpenAI, Anthropic, or other technology companies gain more market share, electricity demand may increase with the expansion of computing infrastructure.
O'Leary said that over the next 36 months, he will focus on energy companies, as well as fields such as natural gas power generation, turbines, power infrastructure, and uranium mining. What he values is not that these assets have an AI concept, but that they may become necessary inputs for the development of the entire industry. This also explains why he is bullish on Canada and the Nordic countries.
In the interview, O'Leary repeatedly mentioned the energy advantages of Finland, Norway, and Alberta, Canada. He believes that these regions have relatively low-cost power resources, and some long-term power supply contracts can be priced below 6 cents per kilowatt-hour.For large data centers that need to operate over the long term, electricity costs can directly affect project profit margins. Even if the competitive landscape among model companies changes, enterprises that own land, grid access conditions, and long-term power contracts may still be able to provide services to different customers.
The Bitzero he invested in is a concrete embodiment of this idea. According to O'Leary, Bitzero once used Bitcoin mining to generate cash flow, but what he values more is the long-term power contracts and infrastructure resources the company obtained in Norway and Finland.
For him, mining is just one of the existing businesses; what may truly have long-term value is the company's ability to secure energy. If cloud computing or AI companies need to build data centers locally in the future, Bitzero has the opportunity to use its existing resources to participate in related projects. The appeal of this business model is that infrastructure providers do not necessarily need to judge which AI model customers will ultimately use.
Of course, controlling power resources does not mean already having stable profits. Projects still need to solve issues such as grid access, equipment procurement, customer contracts, and capital expenditure, and the price advantage of long-term power supply contracts must also be able to cover comprehensive construction and operating costs. But O'Leary believes that, compared with looking for a single winner in model competition, energy infrastructure provides another way to participate in AI growth.
The growth in AI demand may be reflected not only in the revenue of technology companies, but also in the economic value of upstream energy, equipment, and infrastructure.
From electricity further extending to uranium mining
O'Leary's investment in energy has not stopped at existing power supply. He is also studying the energy structure of the next seven or eight years, and is especially optimistic about the development potential of Small Modular Reactors (SMRs).
SMR is a nuclear energy technology route that uses smaller reactor modules, and some designs hope to improve the flexibility of nuclear power projects through modular manufacturing and deployment. In O'Leary's view, if the electricity demand of AI data centers grows over the long term, nuclear energy has the opportunity to become an important choice for stable power supply, and uranium mining is the upstream resource in the related industrial chain.
Therefore, he began to pay more attention to uranium mining investments. This judgment still depends on the commercialization progress of nuclear power technology. Whether SMRs can achieve competitive power generation costs within the expected time still depends on engineering construction, regulatory approval, financing, and fuel supply.
But O'Leary's investment logic is consistent: he is not in a hurry to bet on whether a certain technology can ultimately become the market winner, but instead studies in advance the basic resources on which technological expansion depends.
Canada is an important part of this idea. He believes that Canada has natural gas, uranium, potash, aluminum, and other strategic resources, while also being adjacent to the large U.S. energy consumption market. If North America continues to increase investment in computing power and power infrastructure in the future, Canada may benefit from it.
O'Leary disclosed that his team established about a 10% exposure to the Canadian market through the large-cap Canadian ETF XIU. This trade initially did not receive unanimous support from the team, because at the time U.S.-Canada trade friction increased market concerns about the Canadian economy.
But O'Leary regarded this pessimism as a contrarian investment opportunity. He believes that, as Canada's policy environment changes and resource development projects advance, the market may reassess the value of local assets.
He said in the interview that this investment at one point outperformed the S&P 500 by 158 basis points, or 1.58 percentage points. However, the interview did not disclose the complete comparison period or pricing basis, so this figure can only be regarded as a description of his personal investment performance.
Whether Canada can outperform over the long term still depends on the implementation of energy projects, commodity prices, trade relations, and policy execution efficiency. For O'Leary, betting on Canada is not simply a judgment on economic recovery, but a bet that global technological expansion may once again raise the strategic value of energy and natural resources.
Crypto can no longer rely only on capital inflows; the real question is whether public blockchains can generate revenue
If AI investment is shifting from models to energy, then O'Leary's reassessment of Crypto is shifting from financial market trading to actual business demand.
In the past Crypto investment framework, an important assumption was that institutional capital entering the market would drive broader asset allocation demand for the entire industry. But O'Leary found that institutional capital may not necessarily flow evenly into all crypto assets.
He recalled that at an industry conference previously attended by a large number of institutional investors, some analysts believed that through Bitcoin and Ethereum alone, one could obtain about 97% of the price volatility exposure of the crypto market they studied.
This figure is a research judgment he relayed, not the market capitalization share of BTC and ETH, but the allocation logic behind it is worth noting. For large institutions, if a small number of leading assets can already satisfy their main Crypto allocation needs, then the necessity of buying a large number of small and medium-sized tokens may decline. This would lead to more obvious capital divergence in the Crypto market.
O'Leary said that his portfolio had previously held 27 Crypto-related positions, and later sold some assets, while some of the remaining investments also suffered relatively large losses, with the main assets ultimately retained including Bitcoin and Ethereum.
But he did not therefore conclude that other public blockchains had completely lost investment value. On the contrary, he began to look for a new standard of judgment: if it no longer relies on the entire Crypto market rising, how exactly can other public blockchains prove their economic value?
He believes the answer may come from actual enterprise adoption of blockchain technology.
In past market narratives, Ethereum was often seen as one of the networks most likely to become general-purpose blockchain infrastructure. But O'Leary began to doubt whether all industries in the future would really adopt the same public blockchain. Industries such as finance, logistics, real estate, contract management, and sports assets do not have exactly the same needs for blockchain.
Financial institutions may focus on stablecoin payments and asset transfers; logistics companies place more emphasis on supply chain management and contract execution; sports clubs may want to use blockchain to manage collectibles and digital assets.
Different business needs may give rise to different technical architectures. This is also why O'Leary reexamined Avalanche. He specifically mentioned Avalanche's ability to support enterprises in deploying customized blockchains, believing that this model may be suitable for sports assets, collectibles, and other scenarios that require independent business networks.
In his view, enterprises may no longer need to put all businesses on the same public network, but instead choose different blockchain infrastructures according to industry needs. This means that future blockchain competition may form two paths: one in which a small number of general-purpose networks dominate, and another in which multiple networks separately serve different industries and achieve interconnection through technical interfaces.
It is currently impossible to determine which model will ultimately win. More importantly, there is no inevitable link between business adoption rates and token investment returns. Even if an enterprise adopts Avalanche technology, it does not mean that demand for the AVAX token will necessarily grow in sync. Investors also need to study how enterprises pay network fees, how business revenue is generated, and whether this economic value can be transmitted to the token.
Therefore, O'Leary did not say he intended to resume large-scale buying of altcoins, but rather that he is considering establishing smaller public blockchain positions after studying specific applications. Unlike past Crypto investments driven by market liquidity, he has begun to focus on whether enterprises truly need these blockchains and whether demand can form sustained business revenue.
From this perspective, the next stage of Crypto competition may no longer be only about competing for investors' trading funds, but about competing for the business needs of real enterprises.
Quantum computing is changing the long-term risk assessment of Crypto
When O'Leary began to reexamine the commercial value of public blockchains, he also raised another risk that could affect the entire Crypto industry: quantum computing.
In the past, when investors assessed Bitcoin's risks, they mainly focused on market prices, liquidity, policy regulation, and the macroeconomic environment. But with the development of quantum computing technology, whether existing cryptographic mechanisms will need to be upgraded in the future has also begun to become an issue worth studying.
O'Leary calls the point in the future when quantum computing may break through some existing cryptographic algorithms Q-Day.His judgment is not that Bitcoin has already suffered a quantum attack, but that market expectations of a potential threat may affect prices before a technological breakthrough occurs. If institutional investors begin to worry that future quantum computers could attack existing digital signature systems, then even if a real attack has not yet occurred, it could still reduce their willingness to allocate assets.
He personally speculates that some institutional investors may refrain from allocating to Bitcoin for a long time because of quantum risk. However, this proportion has not been verified by public statistics and remains his personal judgment.
At the technical level, a distinction also needs to be made: quantum computing poses a potential threat to some public-key cryptographic algorithms, but that does not mean all of Bitcoin's security mechanisms will fail simultaneously at some point. The relevant network may also adopt post-quantum cryptography through protocol upgrades.
For O'Leary, this type of risk gives rise to another investment opportunity. If financial institutions need to upgrade their security systems in the future, then companies capable of providing quantum computing technology and quantum-resistant solutions may gain new business demand. He therefore began paying attention to the quantum computing businesses of companies such as IBM and Google.
But there are still differences between quantum computing R&D capabilities, commercialization revenue, and post-quantum security products. Investors cannot simply equate technological progress with corporate profit growth.
From energy to quantum computing, O'Leary is in fact studying the same type of problem. The large-scale adoption of a technology often not only creates new products, but also places higher demands on existing infrastructure. AI needs more energy, enterprise blockchain needs networks that can meet actual business needs, and the development of quantum computing may drive upgrades in cryptography and information security systems.
The investment opportunities brought by industrial change do not necessarily exist only in the new technology itself; they may also exist in the infrastructure that must be upgraded to adapt to the new technology.
From betting on technology winners to managing investment risk, it ultimately still comes back to cash flow
Although O'Leary has made many positive judgments about energy, blockchain, and quantum computing, he does not believe investors should concentrate large amounts of capital betting on a single technology trend. Instead, he regards diversification as one of the most important investment disciplines.
His rule is: a single stock should not exceed 5% of the portfolio, and a single industry should not exceed 20%.
This rule is related to his early investment experience. During the internet bubble, O'Leary once shorted Yahoo. As the stock price continued to rise, he repeatedly faced margin call pressure. According to his recollection, this trade once caused a huge drawdown of about 40% to 50% in his net worth. Although Yahoo later experienced a stock price decline, the experience of enduring unrealized losses and margin calls for a long time made him decide not to engage in stock short selling again. This also changed the way he managed his portfolio.
Today, he tends to build a base position through index funds, then appropriately increase allocation to companies he is optimistic about, but always controls the weight of any single stock.
Tesla is one example. O'Leary recalls that he initially did not agree with Tesla's high valuation, but an argument made by his son changed his judgment: Tesla is not just an automaker, but also a technology company that accumulates a large amount of data through its vehicles. He eventually built a position and trimmed it multiple times as the stock price rose, avoiding Tesla becoming too large a share of his portfolio.
This approach may cause him to miss part of the subsequent upside, but it also reduces the impact of a single stock on overall asset performance. In the interview, he mentioned that for some individual stock trades, he usually uses a return of about 17% as the minimum target threshold, and after reaching the target considers selling more than half of the position.
These are his personal trading rules, and they do not mean that a fixed take-profit level can guarantee higher long-term returns. What really matters is that he does not want the performance of his portfolio to depend entirely on a judgment about any one company.
In private company investments, O'Leary similarly emphasizes using operating data to validate the investment thesis. He said that when evaluating a company, he places the greatest importance on Customer Acquisition Cost (CAC) and Churn Rate. The former reflects how much money a company needs to spend to acquire customers, while the latter is used to judge whether existing customers can be retained over time.
If a company must continuously increase marketing spending to maintain revenue growth, while customers are also continuously churning, then the apparent rapid growth may not have good economics.
O'Leary also studies a company's cash flow, debt levels, and operating data from multiple past quarters. This is also one of the reasons he is optimistic about enterprise applications of AI. Among the private companies he has invested in, AI tools have already been used for financial analysis, customer acquisition, and daily operations. He believes that one of AI's real commercial values is helping companies reduce operating costs and improve profit margins.
He also mentioned that his company team chose Anthropic's products as internal tools. This made him realize that different models do not perform exactly the same in actual enterprise use, and enterprise procurement does not depend only on public model rankings.
But regardless of which vendor is chosen, the ultimate standard for measuring the value of AI tools is still operational improvement. Whether the technology is advanced enough and whether a company can profit from it are two different questions. This also explains why O'Leary, while optimistic about AI's long-term development, devotes a great deal of attention to energy, infrastructure, and corporate cash flow.
What he focuses on is no longer just how much new market demand technology can create, but through what business mechanisms these demands ultimately translate into profits.
For energy companies, it is necessary to observe whether low-cost electricity resources can form profitable long-term contracts; for public blockchains, it is necessary to verify whether enterprise adoption can translate into real network revenue; for quantum computing, it is necessary to distinguish research progress from actual commercial products; and for AI applications, it is necessary to focus on whether companies can continuously reduce costs and improve profit margins.
If these conditions are not fulfilled, then even if the industry grows over the long term, the related assets may not necessarily deliver returns in line with market expectations.
Therefore, O'Leary's entire investment approach can be summed up in one judgment: rather than trying to predict the ultimate winner of each round of technological competition, it is better to study which companies can continuously provide necessary resources, obtain commercial revenue, and convert growth into cash flow throughout the process of industrial expansion.
From AI power infrastructure to enterprise blockchain and quantum security, these investment themes appear on the surface to belong to different industries, but behind them all is the question of how economic value is distributed after technology becomes widespread.
For investors, what truly needs to be verified next is not only whether technology will continue to advance, but whether the market's pricing of future growth can be supported by actual operating results.