During the Sustainable Global Leaders Conference, renowned Nordic investor and author of "The Psychology of Finance," Lars Tvede, spoke with Sina Finance, offering an in-depth analysis of the growth logic and investment opportunities in the age of artificial intelligence.
In Tvede's view, AI is not merely a technological upgrade, but could potentially drive the global economy into a new growth model—what he calls the "Precision Economy." Technological progress is expected to allow humanity to create more output with less energy, land, and material resources. "What truly matters is not simply reducing the use of technology, but leveraging technology itself to make the entire economic system more efficient and resource-saving."
Regarding the widely watched AI investment bubble, he believes that the history of general-purpose technology development shows that bubbles are almost impossible to completely avoid, yet it is still too early to declare that AI has entered a systemic bubble. Rather than focusing solely on the valuations of tech giants, it is more worthwhile to pay attention to supply and demand changes in infrastructure such as chips and electricity, and how AI can truly enter traditional industries and create economic value.
On the topic of global industrial competition, Tvede particularly emphasized the importance of talent and education. In his view, while owning models, chips, and data centers is important, what truly determines a country's long-term competitiveness is how many people can understand the operational logic of specific industries and transform AI into real productivity. "In the AI era, chips, computing power, and electricity determine the foundational conditions for technological development, while talent determines how much economic value these resources can ultimately create," Tvede told Sina Finance.
AI May Help Humanity Find More Sustainable Energy Solutions
The rapid development of AI is creating a real contradiction: on one hand, artificial intelligence promises to significantly boost production efficiency; on the other hand, the expansion of data centers brings enormous electricity demand, making energy consumption a key constraint on industrial development. As for what long-term impact AI will have on sustainable development, Tvede said it is currently impossible to give a definitive answer. But he believes what deserves more attention is whether technological progress can change economic growth's dependence on resource inputs.
"I believe we have already entered a new economic development model, which I call the 'Precision Economy.'" The so-called precision economy refers to achieving more output with fewer resources through technological progress, so that economic growth is no longer necessarily accompanied by proportional increases in energy consumption, land use, and material resource inputs. Tvede used global agriculture as an example to explain this logic. In the past, as the global population continued to grow, people widely feared that food supply would struggle to meet demand. American biologist Paul Ehrlich warned in his book "The Population Bomb" about the food crisis that population growth could trigger. However, advances in agricultural technology, particularly the Green Revolution, improved food production efficiency. Since then, the global population has continued to grow, per capita food supply has also improved, while arable land area has not continued to expand at a rate proportional to population growth. In Tvede's view, this is a typical manifestation of the precision economy: humanity does not necessarily have to obtain more output by investing more land and resources—technological innovation can also change the relationship between resources and output. He believes AI could play a similar role.
However, Tvede cautioned that the current greater concern is not AI itself, but whether energy supply and its solutions are sufficiently efficient. He hopes AI can help humanity find more compact, efficient, and sustainable energy solutions. At the same time, the AI industry itself must continuously improve efficiency. In specific scenarios, software advances could even bring hundreds of times or higher efficiency improvements, with the exact magnitude depending on the task and benchmark. Therefore, Tvede does not support attempting to slow down AI development merely because AI consumes energy. "What truly matters is not simply reducing the use of technology, but leveraging technology itself to make the entire economic system more efficient and resource-saving."
AI Computing Power Is Moving Toward Distributed Models; Europe Lags Behind But Still Has Opportunities
Against the backdrop of continuously growing computing power demand, Tvede believes that large data centers will remain an important part of AI infrastructure. But at the same time, AI computing is also extending to more terminal devices. Tvede told Sina Finance that the future of AI is not only about building more large data centers, but also about distributing intelligent capabilities to enterprises, devices, and daily economic activities, making AI gradually become a ubiquitous foundational capability.
When discussing Europe's competitiveness in the AI field, Tvede's assessment was more cautious. "Frankly speaking, in the AI field, Europe has clearly fallen behind." He believes that in developing the most advanced large foundation models, Europe faces obvious disadvantages. Developing globally competitive foundation models requires massive investment and sustained costs in computing power, research and development, and talent. Whether relevant European companies can obtain sufficient capital to engage in long-term competition with the strongest American competitors remains uncertain. French AI company Mistral is a noteworthy representative in Europe, but overall, Europe still lacks enough foundation model developers with international influence. However, Tvede does not believe Europe has lost its chance to participate in AI competition. He is particularly optimistic about decentralized applications, small language models, and AI deployment in industrial sectors. In his view, the participants in future AI competition will be more diverse. The leaders in foundation models may not necessarily monopolize all economic returns, and companies skilled at embedding AI into actual production processes could also gain important competitive advantages.
Watch Chips and Electricity Over the Next Three to Five Years; AI Investment Value Will Shift from Infrastructure to the Application Layer
As for the AI investment opportunities that investors care most about, Tvede places them within the historical evolution of general-purpose technologies. He believes that whenever a general-purpose technology emerges, its economic value typically goes through three stages. In the first stage, returns are mainly concentrated in the technology itself and its infrastructure providers. In the early days of internet development, demand for networking equipment, routers, and communication infrastructure grew rapidly, making companies like Cisco the focus of market attention. In the second stage, value gradually shifts to enterprises that apply these technologies. As internet infrastructure matured, companies like Amazon and Alibaba used the internet to reshape business models, creating new products, services, and consumption scenarios. Those who truly captured enormous economic returns were no longer just infrastructure producers, but also enterprises skilled at using infrastructure to conduct business. In the third stage, the benefits of technological progress are further passed on to consumers. Tvede cited American economist William Nordhaus's research on the distribution of economic returns from innovation, noting that while profits initially created by innovation may be concentrated among inventors and technology producers, in the long run, a considerable portion of the benefits society receives ultimately manifests as consumers being able to obtain products and services at lower prices. He believes AI may follow a similar path.
For investment directions over the next three to five years, Tvede is first bullish on chip manufacturers. In his view, high-performance computing power is still constrained by supply capacity and delivery speed. Even if chip supply continues to increase, the expansion of AI applications could bring even greater computing power demand. Therefore, chips remain a field worthy of close attention. The second direction is energy, especially companies that can quickly provide the electricity and energy solutions AI needs in the U.S. market. As data center construction accelerates, power supply has become an important factor affecting project implementation. Companies that can quickly provide available energy and help data centers obtain electricity in a timely manner may gain a competitive advantage as a result. However, Tvede emphasized that AI investment opportunities will not remain at the infrastructure level forever. As technology gradually matures, industries such as biotechnology, media, finance, and insurance are expected to use AI to transform production processes, reduce operating costs, and even create entirely new products and services. Conversely, if a company consistently fails to effectively apply AI, it may find itself at a disadvantage in competition: rivals can use AI to offer the same products at lower cost, or better products at the same cost. Therefore, Tvede's investment logic is not simply betting on all AI-related companies, but rather following the path of technological evolution to find value creators at different stages. He believes that in the short term, look at chips and electricity; in the medium to long term, focus on companies that can truly use AI to improve operational efficiency.
Is There a Systemic AI Bubble? It's Too Early to Judge
As the AI investment boom continues to heat up, market concerns about a valuation bubble are also increasing. Tvede told Sina Finance, "If AI ultimately experiences an obvious investment bubble, I wouldn't be surprised. But I don't think we can now assert that AI is already in a bubble." He recalled that when the internet bubble peaked in March 2000, Cisco was one of the most representative companies in the market. At that time, Cisco's stock price had a price-to-earnings ratio of about 135 times its trailing 12-month earnings, and its valuation was also extremely high based on forward earnings. By comparison, the forward price-to-earnings ratio of NVIDIA (NVDA) cited by Tvede in the interview is approximately 18 to 20 times. The stock prices, earnings levels, and valuation bases of the two at different points in time are not entirely consistent, so this comparison cannot be directly used as a basis for judging AI's overall valuation. But he used it to emphasize that one cannot conclude that the entire industry has formed a bubble simply because AI-related company stock prices have risen. He also mentioned that for memory chip companies such as SK Hynix, Samsung Electronics, and Micron Technology, based on market earnings forecasts for 2026 to 2028, some companies' forward price-to-earnings ratios may be at relatively low levels, with some forecasts even reaching single digits. Of course, a low price-to-earnings ratio does not mean there is no risk. Earnings forecasts for cyclical industries may be overly optimistic, and once demand growth slows, prices fall, or supply increases, actual earnings could fall short of expectations.
In Tvede's view, to judge whether AI has a bubble, one must also pay attention to a more important issue: whether the industry is experiencing competitive overinvestment. During the internet bubble, many companies built large-scale fiber optic networks based on extremely optimistic growth expectations. Investors believed these companies would dominate the market in the future, but actual data traffic was insufficient to support the massive infrastructure scale, ultimately resulting in a large number of idle assets. Tvede believes the current AI industry situation is different—the main problem in some key segments is still insufficient supply, not insufficient demand. "But this does not mean there is no risk of overexpansion in AI investment. As companies continue to invest, if computing power supply grows too quickly in the future and application demand fails to meet expectations, overcapacity and declining returns on assets could still emerge in certain areas," Tvede told Sina Finance. Therefore, Tvede believes AI could experience a bubble, but judging a bubble cannot rely solely on valuations—it also requires observing real demand, supply constraints, and whether capital expenditures can ultimately translate into economic returns.
The Key to AI's Long-Term Competitiveness Lies in Cultivating Talent That Can Create AI
Regarding China's position in global AI competition, Tvede believes that energy, digitalization, and technical talent form an important foundation, but whether these advantages can translate into lasting industrial strength ultimately depends on talent cultivation and application capabilities. In his view, owning advanced models does not mean companies will automatically achieve productivity gains. When AI truly enters specific work scenarios, companies must have people who understand the business itself, know how to adjust workflows, reorganize production, and transform technological capabilities into economic value. "If China can enable more students to learn AI, understand the basic principles of AI, and master how to develop AI applications, agentic AI, and small language models, then this will become a huge competitive advantage." In Tvede's view, future AI education should not be limited to teaching students to use ready-made tools, but should also help them understand technical principles, master development capabilities, and learn how to apply AI to different professional fields. This capability needs to be cultivated starting from school education and carried through secondary school and other educational stages. He believes China has a large pool of technical talent, and if it can further strengthen AI education so that more people have the ability to create AI applications, develop agents, and transform industry processes, it will have the opportunity to turn existing advantages into long-term competitiveness. This logic also applies to the United States and Europe. "What truly determines long-term competitiveness is not just how many chips, how many data centers, or how much electricity you have, but how many people can transform AI into real productivity."
From the "precision economy" to investment cycles, from infrastructure competition to talent cultivation, Tvede's analysis points to the same question: technology itself does not automatically create sustained economic prosperity. What truly matters is how humanity uses technology, organizes production, and transforms innovation into long-term value. In this AI transformation that is still in its early stages, whoever can complete this transformation more effectively will be more likely to seize the initiative in future global competition.
Comments