Compiled & Translated: Deep Tide TechFlow

Podcast Source: Bloomberg Originals

Original Title: How Circular Deals Are Driving the AI Boom

Broadcast Date: January 23, 2026

Key Points Summary

The AI boom is everywhere, but a large part of it is just a surface phenomenon. Funds are circulating among a few unprofitable companies. If this is a bubble and it eventually bursts, its impact could affect everyone, and the consequences cannot be ignored.

Many warn that if it ultimately proves to be a bubble and bursts, it could have far-reaching effects on the entire economy. Bloomberg Originals explores the circular investment deals among AI companies and how these deals constitute the so-called "ultimate gamble."

Highlights Summary

Specific Chain of Circular Investments

• Nvidia plans to invest up to $100 billion in OpenAI, while OpenAI is also a major customer of Nvidia's chips
• OpenAI leases computing services from Oracle, which is itself a customer of Nvidia—funds form a closed loop among several companies

Profitability Dilemma

• Major AI projects like OpenAI and Anthropic are currently unprofitable; every time a user uses ChatGPT, OpenAI may be losing money
• Sam Altman stated that the company expects to break even around 2029 to 2030

Infrastructure Race

• Morgan Stanley estimates total corporate investment in AI data centers is projected to reach $3 trillion
• A 1 million square foot textile factory was converted into a data center; retrofitting existing facilities takes 6 months to start up, while building from scratch takes 2 years

Precedent of the Dot-com Bubble

• The 2000 dot-com bubble burst wiped out about $5 trillion globally
• Amazon's stock took 8 years to recover to pre-bubble levels, Cisco took 25 years

Concerns of "Too Big to Fail"

• The AI investment boom has become a significant driver of GDP growth
• Ordinary Americans' retirement accounts indirectly hold shares in these tech companies, creating broader exposure than imagined
• Some worry this could be like the 2008 financial crisis—where large institutions needed massive financial support to prevent a total economic collapse
• It could be argued that AI is Wall Street's biggest gamble ever, and Wall Street itself is known for taking risks, making this investment the "ultimate gamble."

The AI Boom and Circular Investments

Artificial Intelligence (AI) is expanding from Wall Street to rural America, becoming a core driver of economic development. The market is confident in AI's potential, viewing it as an infallible miracle. Investors have very high growth expectations for AI. Tech giants like Microsoft, Meta, and Alphabet have already invested tens of billions of dollars in capital expenditures for related areas and plan to increase investments further in the future.

The AI boom is not limited to software development; it also drives infrastructure construction. For example, supporting AI development requires building more data centers and ensuring energy and water supply. However, this rapidly growing industry also presents some risks, particularly in how funds flow. A new investment strategy is emerging—circular investments amounting to tens of billions of dollars. For instance, Nvidia plans to invest up to $100 billion in OpenAI, and these huge sums circulate among tech giants, forming a fund chain akin to a "merry-go-round."

Nevertheless, AI's potential remains enormous. Currently, about 80% of U.S. companies have started using AI, signaling a structural revolution similar to electricity or the internet.

On the Bubble Issue and Complex Fund Flows

Although Artificial Intelligence (AI) has huge potential, its profitability has not yet been fully proven. Today, the biggest question in San Francisco's tech world is: Are we in an AI investment bubble? If so, how big is this bubble? What would be the consequences if it bursts? This is an important question. We might be in a new era of AI-driven growth, or we might be facing an unprecedented investment bubble.

So-called "circular investments" refer to the flow of funds, products, and services between companies. For example, Nvidia plans to invest up to $100 billion in OpenAI, while OpenAI is also a major customer of Nvidia's chips. This fund flow also involves other intermediaries, such as Oracle. OpenAI sometimes leases computing services from Oracle, and Oracle itself is also a customer of Nvidia. This complex web of fund flows makes the entire industry an interconnected network involving many well-known companies.

Concerns About Industry Interdependence and Infrastructure Construction Race

Funds frequently circulate among these companies. While there's nothing inherently wrong with this model, when transaction amounts become too large, it can lead to overexpansion. The current concern is, could this symbiotic relationship make the entire system fragile? If one of these companies underperforms or encounters problems, could it affect the stability of the entire industry?

Meanwhile, massive investments are pouring into data center construction, driving infrastructure expansion nationwide. We are experiencing an infrastructure "arms race." For example, construction spending in most industries was declining in 2025, but spending on data centers and power plants was increasing. Many companies are acting as "foundation builders" for the AI industry, actively investing in these projects. According to Morgan Stanley's latest estimates, total corporate investment in AI data centers is projected to reach $3 trillion.

The Data Center Construction Boom: The "Picks and Shovels" of Infrastructure

Currently, data center construction is in a phase of rapid development. If your business provides infrastructure and services for data centers, you are in a very favorable position. Market demand far exceeds supply capacity, funding support is ample, and the industry outlook is positive. For example, the facility we are currently in was once a roughly 1 million square foot textile factory but was later converted into a data center.

The demand for data centers is almost insatiable, encompassing power supply, infrastructure construction, and specialized technical support, among other aspects. This demand will not slow down in the short term. For the AI industry, time is critical. If operations can start within 6 months by retrofitting existing facilities, rather than taking 2 years to build new ones from scratch, it is undoubtedly a better choice. Meanwhile, the power demand from data centers is also driving rapid growth in utility costs, even exceeding inflation rates. Utility companies and construction-related businesses that specifically serve data centers are performing particularly well.

The Profitability Puzzle: Challenges and Risks of AI Projects

However, building data centers quickly does not mean easy profitability. Data centers require continuous investment to maintain normal technological operations; otherwise, they quickly lose appeal to customers. So far, major AI projects are still operating at a loss. Taking OpenAI as an example, every time a user uses ChatGPT, OpenAI may face losses; companies like OpenAI and Anthropic are currently not profitable.

OpenAI's CEO Sam Altman stated that the company expects to break even around 2029 to 2030, but considering the current massive cash burn and the need for even more funds to build data centers and purchase computing resources in the future, this goal seems challenging. People worry whether these AI startups can bear such high costs, especially when they commit huge investments to build data centers. These data center companies can be seen as "early warning signals" for changes in industry demand. If demand for AI products suddenly weakens, the entire industry could be affected. Although all companies currently claim that demand for AI products is very strong, once demand drops, problems will emerge.

Historical Parallels: Dot-com Bubble vs. AI Boom

To understand the potential risks of today's AI boom, one only needs to look back at the 2000 dot-com bubble. At that time, internet companies promised a hopeful new era but ultimately led to huge losses. Savings were wiped out, office parks stood empty, and about $5 trillion in value evaporated globally. Tech stocks were hit the hardest, including many internet companies. Even the strongest companies took years to recover. For example, Amazon, a famous survivor, saw its stock take 8 years to return to pre-bubble levels. Cisco, a company providing infrastructure, took a full 25 years to recover its stock price.

There are indeed some similarities between the two booms, such as the phenomenon of circular investment deals. The question is, will the AI boom transcend the normal fluctuations of the tech industry and have a profound impact on the entire economy?

Economic Impact and "Too Big to Fail" Concerns

The dot-com bubble severely hurt the economy, but if the AI boom collapses, its impact could be even more far-reaching. The AI investment boom has become a significant driver of GDP growth, boosting the U.S. economy amid tariffs and inflation pressures. However, this also exposes ordinary Americans indirectly to risk, as many people's retirement accounts and other investment accounts hold shares in the big tech companies participating in AI investments.

Does this mean the AI boom has become "too big to fail"? The current concern is whether these companies have become "too big to fail." If they fail, it would not only cause economic problems but could also have broader impacts. Some even worry that this situation could be like the 2008 global financial crisis, where large financial institutions needed massive financial support to prevent a total economic collapse. If the AI boom truly collapses, the U.S. economy could face even greater challenges.

Long-Term Outlook: AI's Future Remains Optimistic

Despite the risks brought by the AI boom, many remain confident in its future because technology is constantly advancing. During the dot-com bubble, many companies invested in laying fiber optic cables. At the time, these investments seemed excessive and wasteful, but ultimately, this fiber became the foundation of internet broadband. The unused fiber built in the 90s later became crucial to internet development. Similarly, the data centers being built today, even if they result in temporary overcapacity, could potentially be fully utilized in the future.

Of course, AI development might take longer than expected. In this process, although some well-capitalized companies may survive, their valuations could experience significant fluctuations. However, AI technology itself will not burst like a bubble. Even though some companies might not withstand market tests, the AI industry is not an illusory bubble. It has developed actual products and demonstrated enormous potential. It could be argued that AI is Wall Street's biggest gamble ever, and Wall Street itself is known for taking risks, making this investment the "ultimate gamble."