Brace Yourself.

Summary

The video explores the concept of asset bubbles, focusing on semiconductor stocks, which meet all the criteria for a potential bubble. While the tech sector has shown moderate growth, semiconductor stocks have surged significantly, raising concerns about an impending bubble. The video also highlights AI infrastructure spending, projecting massive future investments. However, despite the hype, productivity gains from AI remain limited, and companies are not yet seeing significant benefits. The video suggests that while AI-related growth may continue in the short term, unrealistic expectations could lead to a downturn by 2026, especially if return on investment fails to meet predictions.

Takeaways

• 😀 The National Bureau of Economic Research defines an asset bubble with three key conditions: a 100% rise over 2 years, outperformance of the broader market by 100%, and a 50% return over 5 years.
• 😀 Historical examples of asset bubbles include railways in the 1840s, the Dow Jones in the 1920s, and Japan’s stock market in the 1980s, all of which eventually burst, leading to major economic downturns.
• 😀 The current tech market, particularly the NASDAQ 100, does not meet the criteria for an asset bubble, having only risen by 45% in the last 2 years, outperforming the S&P 500 by 3%, and gaining 90% in the last 5 years.
• 😀 Semiconductor stocks have been exhibiting bubble-like behavior, with a 110% rise in the last 2 years, a 275% gain over the last 5 years, and outperformance of the S&P 500 by almost 100%.
• 😀 Despite semiconductor stocks' bubble-like characteristics, they remain a small part of the overall economy compared to the tech boom of the late 1990s.
• 😀 Semiconductor stocks are crucial to AI infrastructure, including data centers, cloud computing, and large language models, making the entire sector highly interconnected with AI's future.
• 😀 AI infrastructure spending is having a significant impact on U.S. economic growth, with $375 billion invested in 2025, contributing 1.2% of GDP, expected to rise to $875 billion in 2026, equaling 3% of GDP.
• 😀 The multiplier effect of AI spending is substantial; for example, $177 billion in tech AI capex in 2025 led to $700 billion in U.S. economic growth, creating a multiplier of 3.5x.
• 😀 Taiwan’s chip exports, which constitute 60% of the world’s semiconductor supply, play a crucial role in U.S. economic growth, as strong exports correlate with periods of economic expansion in the U.S.
• 😀 By 2030, AI capex is projected to reach $7 trillion, equivalent to 20% of U.S. GDP, but whether this spending will result in real economic value and a positive return on investment is uncertain.
• 😀 Despite the large AI spending, there is little concrete evidence of significant productivity improvements, and many companies are reporting minimal benefits from AI adoption. This raises concerns about whether the high expectations for AI spending will be met.

Q & A

What are the three conditions that qualify an asset as a bubble according to the National Bureau of Economic Research?

The three conditions are: 1) The asset must rise by more than 100% over a 2-year period. 2) It must outperform the broader market by at least 100% over the same period. 3) It needs to deliver a 5-year return of over 50%.

Which historical examples met the conditions for an asset bubble?

Examples include the railways in the 1840s, the Dow Jones in the 1920s, and Japan's stock market in the 1980s, all of which met the three conditions for a bubble before their subsequent burst.

How does the current NASDAQ 100 performance compare to the definition of an asset bubble?

The NASDAQ 100 has risen by 45% in the last 2 years, outperformed the S&P 500 by only 3%, and gained 90% over the last 5 years. This does not meet the three conditions of an asset bubble.

How do semiconductor stocks compare to the NASDAQ 100 in terms of meeting the criteria for a bubble?

Semiconductor stocks have risen by 110% over the last 2 years, gained 275% over the last 5 years, and outperformed the S&P 500 by almost 100% over the last 2 years. This meets all three conditions for an asset bubble.

What role do semiconductors play in the AI infrastructure?

Semiconductor chips are central to AI infrastructure, including data centers, cloud computing, and large language models. These technologies rely heavily on semiconductor chips to function.

How much is expected to be spent on AI infrastructure by 2025, and what is the projected impact on US GDP?

In 2025, about $375 billion is projected to be spent on AI infrastructure, which is roughly 1.2% of US GDP. This figure is expected to more than double to $875 billion in 2026, making up 3% of US GDP.

What is the multiplier effect in relation to AI capital expenditure (capex)?

The multiplier effect refers to the additional economic growth generated when AI capex spending circulates through the economy. For example, in the first half of 2025, $177 billion of tech spending led to $700 billion in US economic growth, resulting in a multiplier effect of 3.5x.

How does Taiwan’s semiconductor export data relate to US economic growth?

Taiwan produces over 60% of the world’s semiconductors. Historical data shows that when Taiwan’s chip exports increase, US economic growth tends to strengthen, while a slowdown in exports often coincides with a weaker US economy.

What is the projected growth of AI capex spending by 2030, and why is it seen as a potential risk to the economy?

AI capex spending is projected to reach $7 trillion by 2030, which is approximately 20% of current US GDP. If these expectations do not materialize into real economic value or positive returns, the resulting collapse in spending could significantly impact GDP growth, creating a risk to the economy.

Is AI currently delivering on its productivity promises, and how does this affect economic expectations?

AI has not yet delivered substantial productivity gains. Since the mainstream adoption of ChatGPT in 2022, US labor productivity has increased by just 2%, which is only slightly higher than before. Many companies report no significant benefits from AI adoption, indicating that AI’s productivity promise has not yet been realized.