What are the main concerns about the AI bubble?

Concerns include circular investments (e.g., NVIDIA investing in OpenAI, which then buys NVIDIA chips) and data center overspending. This creates a virtual demand cycle driven by wishful thinking, potentially leading to evaporation of value if AI doesn't deliver expected productivity gains.

How is Google competing in the AI hardware market?

Google is developing its own specialized AI chips (Tensor Core) that are more energy-efficient and cheaper than NVIDIA's general GPUs. They are also building a complete alternative AI supply chain, evidenced by their significant deal with Anthropic.

When might we see a real-world economic impact from AI?

Market analysis companies do not expect a significant real-world impact of AI on economies until 2027 or later. This delay contributes to investor comfort, as the expected productivity gains are not yet on the horizon.

Are AI investments irrational, or is there a rational basis?

While AI capital spending is high and productivity gains are not yet visible, the investment isn't necessarily irrational. Much of it goes into general-purpose infrastructure computation, similar to past technological waves like fiber networks, which ultimately spurred economic growth.

What was the mistake in previous infrastructure bubbles?

The mistake in historical bubbles like railways or electrification wasn't building the infrastructure itself. Instead, it was the belief that the initial companies involved would capture most of the long-term profits, when in reality, the infrastructure benefits were broader.

How can I stay updated with AI advancements?

It's crucial to keep your skills up-to-date in the rapidly changing AI landscape. Resources like Outskill offer training events and materials that teach how to build AI agents, automate workflows, and prepare for future AI shifts.

Why The AI Bubble May Be Good

Sabine Hossenfelder

Science & Technology4 min read8 min video

Mar 10, 2026|33,076 views|2,774|433

hossenfelder science with sabine science humour science news sabine science news science humor ai bubble tech artificial intelligence chatgpt gemini

Save to Pod

Key Moments

TL;DR

AI bubble may be rational; investments in computation infrastructure are key for complexity growth.

Key Insights

The AI boom resembles past infrastructure bubbles like railways and electrification, where initial misallocation of capital ultimately led to significant long-term economic growth.

Circularity in AI investments, such as NVIDIA investing in OpenAI which then buys NVIDIA chips, creates artificial demand and inflates valuations.

Google's strategic shift to develop and push its specialized, more efficient AI chips poses significant competition to NVIDIA's more general-purpose GPUs.

Overspending on data centers could pose a systemic risk to the stock market if AI demand doesn't materialize as expected.

The timing of AI's real-world economic impact is uncertain, with market analysts not expecting significant effects until 2027 or later, which contributes to investor comfort.

The US administration's support and the bubble-like characteristics of high capital spending before visible productivity gains are noted, yet underlying infrastructure investment is seen as potentially rational.

THE RATIONALE BEHIND AI INVESTMENT

The current boom in artificial intelligence, characterized by massive investments from tech giants and high valuations for AI startups, is being examined for its underlying rationality. The speaker posits that civilizations are fundamentally driven by increasing complexity, which in turn requires significant energy and information processing. Consequently, investing in information processing capacity is seen as a sound, physically grounded strategy, with the United States currently leading in this domain. This perspective suggests that the AI investments, despite their bubble-like appearance, might be essential for future progress.

CIRCULAR INVESTMENTS AND ARTIFICIAL DEMAND

A significant concern regarding the AI bubble is the presence of circular investment strategies. Examples include chip manufacturers like NVIDIA investing in AI companies such as OpenAI, which then use NVIDIA's chips, effectively circulating NVIDIA's own money. Similarly, AI companies invest in data centers that then rent space back to these AI firms. This creates a self-sustaining cycle that bolsters perceived demand and valuation, driven by wishful thinking about future actual demand rather than current market realities. This dynamic can be likened to a company lending money to customers to buy its own products.

THE ROLE OF HARDWARE AND DATA CENTERS

The intense investment in AI hardware, particularly semiconductors, has driven up the stock prices of companies like NVIDIA. However, this value is contingent on AI beginning to deliver tangible results; otherwise, the inflated valuations risk evaporation. Furthermore, substantial investments are being directed into data centers. Estimates suggest trillions of dollars will be poured into data centers in the coming years, primarily in the US. A potential mismatch between this build-out and actual AI-driven demand could lead to significant overcapacity and drag down the broader stock market.

GOOGLE'S STRATEGIC CHALLENGE TO NVIDIA

Google has initiated a significant strategic pivot by developing and promoting its own specialized AI chips, creating an alternative AI supply chain. Unlike NVIDIA's general-purpose Graphics Processing Units (GPUs), Google's Tensor Core chips are highly optimized for AI training, offering greater cost and energy efficiency. Reports indicate Google's latest chips provide a substantial performance-per-watt advantage over NVIDIA's best offerings. This competitive push, exemplified by a major deal with Anthropic for its chips, signals a long-term challenge that could gradually erode NVIDIA's dominant market position.

THE HISTORICAL PARALLEL OF INFRASTRUCTURE BUBBLES

The current AI investment landscape shares similarities with historical infrastructure bubbles, such as the railway boom in the 19th century, the electrification wave in the early 20th century, and the fiber optic network expansion in the late 1990s. In these past episodes, significant capital was initially misallocated, leading to company failures and investor losses. However, the underlying infrastructure built ultimately laid the groundwork for subsequent phases of economic growth. The crucial distinction from less successful bubbles, like the tulip mania, lies in the enduring utility of the infrastructure, rather than the speculative frenzy itself.

PROSPECTS FOR ECONOMIC IMPACT AND SURVIVAL

The projected timeline for AI's substantial real-world economic impact remains distant, with market analysis firms not anticipating major effects before 2027 or later. This extended horizon allows investors to remain comfortable, as immediate productivity gains are not expected. Much of the current enthusiasm is US-centric and supported by political factors, which may continue as long as the current administration remains in power. While the AI boom exhibits bubble characteristics with high capital spending preceding visible productivity, the core investments are in general computation infrastructure. The key challenge for investors will be identifying which few companies will ultimately endure and capture long-term value.

Mentioned in This Episode

●Software & Apps

●Companies

●Books

●Concepts

●People Referenced

Common Questions

The AI bubble might be beneficial because it drives investment into information processing capacity, which is crucial for growing complexity and civilization. Past infrastructure booms, like railways and electrification, also saw overinvestment and failed companies, but ultimately led to significant economic growth.