[Alpha Analysis] Peter Thiel Allocates 72% of Portfolio to Energy... The Bottleneck of AI Investment Shifts

The center of the artificial intelligence (AI) investment market is showing signs of shifting from semiconductor companies like Nvidia to power, generation, and power grid infrastructure.

If the first bottleneck in the AI industry was the graphics processing unit (GPU), the market is now beginning to focus on the power generation and transmission, data center connections, transformers, and cooling facilities needed to actually operate hundreds of thousands of GPUs.

Notably, the recent portfolio changes of Peter Thiel, a prominent investor in Silicon Valley, symbolically illustrate this trend.

According to the 13F filings of U.S. institutional investors for the second quarter, Thiel Macro, managed by Thiel, has built a portfolio of U.S. listed stocks worth approximately $418 million, with about 72% concentrated in the energy and utilities sector.

Investment targets include Vistra, American Electric Power (AEP), DTE Energy, Vista Energy, and next-generation nuclear power company X-energy.

Even more noteworthy is the fact that Thiel has liquidated his investment in Nvidia by the end of 2025.

This has led to interpretations that he has not ended his bets on AI growth but has changed his judgment on what resources will become most scarce in the AI industry.

Act 1 of AI Investment is GPUs... Act 2 is Power

The core logic of the AI investment market over the past few years has been simple.

To develop more powerful AI models, more computational power is needed, and to secure more computational power, GPUs are required.

Nvidia was essentially the most important supplier in this structure. As competition for AI models intensified, securing GPUs became a competitive advantage for companies, making Nvidia the biggest beneficiary of the AI investment boom.

However, as large tech companies poured astronomical capital into GPUs and data centers, new problems began to emerge.

Even if GPUs are secured, if there is insufficient power to operate them, AI infrastructure cannot expand.

Large AI data centers require significantly more power than traditional internet service data centers. Simply building data centers is not enough. Power must be generated at power plants, power grids must be expanded, transformers and transmission facilities must be installed, and power must be connected to data centers.

Cooling infrastructure is also essential.

Ultimately, the bottleneck in the AI industry is shifting from semiconductor supply to physical infrastructure supply.

Market interest is naturally spreading from GPU manufacturers to power companies, power grid operators, transformer and power equipment companies, and on-site power solution providers.

Not Just Peter Thiel's Bet

This trend is not just Peter Thiel's strategy.

David Tepper's Appaloosa is investing in power companies Vistra and NRG. Other tech investors are also increasing their exposure to power and energy infrastructure companies like Bloom Energy.

The common judgment among investors is relatively clear.

It is a strategy to invest in resources that will inevitably become scarce as the AI industry grows, rather than the growth of the AI industry itself.

This resembles the investment adage that companies selling picks and jeans during the gold rush era made stable profits compared to those mining for gold.

If the pickaxe of the AI era was the GPU, now power plants and power grids may take on that role.

Diverging Investor Judgments on Memory

On the other hand, in the memory market, another key area of the AI semiconductor supply chain, investor judgments are diverging significantly.

Tepper has reduced his stake in Micron by 41%, while Stanley Druckenmiller has liquidated his position.

In contrast, some investors like Coatue and Atreides still maintain a high exposure to memory companies.

Certain funds have concentrated over 60% of their portfolios in Micron and SanDisk, only to experience significant losses.

This indicates that the AI investment market is already moving beyond the simple 'buy AI growth stocks' stage.

In areas like GPUs and memory, which have already risen significantly, price pressures have increased, and investors are beginning to search for the next growth phase.

That candidate is power.

'Insufficient Power' and 'Buy Utilities' are Different Stories

However, just because AI power demand is increasing does not mean that all power companies become investment opportunities.

In particular, regulated utility companies may face limited investment returns due to regulatory structures, even if they invest massive capital in power plants and power grids.

Even if the construction of AI data centers increases explosively, there is no guarantee that the economic benefits will be fully captured by power companies.

Some utility companies may have already seen their stock prices significantly re-evaluated in anticipation of AI power demand.

Therefore, investors are focusing on infrastructure one step below power companies.

Transformers, transmission equipment, data center power connections, on-site generation, and cooling systems are the 'bottleneck of the bottleneck.'

Understanding what equipment and services are essential each time a new AI data center is built is key.

A Familiar Problem in the Cryptocurrency Industry... Ultimately, the Competition is 'Power'

These changes are not unfamiliar to the cryptocurrency industry.

The Bitcoin mining industry has long evolved into a sector where power prices and the ability to secure power determine competitiveness, rather than computing equipment.

Even if the latest mining machines are secured, it is difficult to maintain profitability without securing cheap and stable power.

AI data centers are also entering a similar phase.

Even if the GPU performance competition continues, in the long run, who can secure more power more cheaply and stably is likely to determine the competitiveness of AI infrastructure.

As AI data centers and Bitcoin mining facilities compete for the same power resources in more regions, the strategic value of power prices, generation infrastructure, and access to power grids may increase further.

This shows the possibility of unexpected connections between the AI and cryptocurrency industries.

The most important question in AI investment is 'Where is the next bottleneck?'

Peter Thiel's portfolio changes do not guarantee a rise in power stocks.

The 13F filings only show investment positions at a specific point in time, and it is unclear whether investors have changed their positions afterward. It does not show investors' hedge strategies or unlisted assets either.

Nevertheless, it is noteworthy that several large investors are simultaneously showing interest in generation and power infrastructure.

Capital expenditures in the AI industry are rapidly expanding from GPUs to data centers, and from data centers to generation and power grids.

As the industry grows, so do the bottlenecks.

In the last AI investment cycle, the most scarce resource was GPUs. After that, memory, including high-bandwidth memory (HBM), emerged as a new bottleneck.

And now the market is asking the next question.

Where will the power to run millions of AI chips be secured?

The next winner in the AI investment market may not be the company that creates the flashiest AI models, but rather the one that supplies the electricity and infrastructure to keep those models running 24/7.

What investors need is not to blindly follow the portfolios of famous investors.

Instead of following the money, one must follow the bottlenecks.

If the next bottleneck in the AI industry is power, then the protagonist of Act 2 of AI investment may not be GPU companies either.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.