How Do Bubbles Form? -

Talk of bubbles has returned to financial markets, particularly around artificial intelligence. Private companies are raising capital at extraordinary valuations, public markets remain highly concentrated in a small number of large technology firms, and investment into AI infrastructure continues at an unusual pace.

Recent funding rounds have only intensified the discussion. For example, Anthropic recently raised around $65 billion at a valuation approaching $1 trillion, reflecting both rapid revenue growth and extraordinary investor appetite for exposure to AI development . At the same time, hyperscalers are committing hundreds of billions of dollars toward data centres, chips, and energy infrastructure to support AI workloads .

Yet despite the scale of these developments, the more useful question may not be whether AI represents a bubble. That question tends to be asked too late or answered too simplistically.

A better starting point is to understand how bubbles actually form, why they recur across very different eras, and what conditions tend to produce them. Because once those conditions are visible, the distinction between innovation and excess becomes much harder to draw in real time.

What a Bubble Actually Is (and Isn’t)

A financial bubble is often misunderstood as a situation where prices rise too far or too quickly. But that definition is too shallow to be useful.

Markets can rise for long periods for rational reasons: productivity growth, falling interest rates, or structural changes in how economies operate. Rising prices alone do not define a bubble.

A more accurate way to think about a bubble is as a process rather than a level. It is a self-reinforcing cycle in which expectations about the future begin to grow faster than the economy’s ability to realistically deliver those expectations in the present.

That cycle tends to follow a familiar pattern. Rising prices attract attention. Attention brings in more capital. That capital pushes prices higher again, reinforcing the belief that the trend is justified. Over time, narratives strengthen and become more confident. What begins as cautious optimism can gradually turn into a widely shared assumption that “this time is different”.

The key issue is the feedback loop between prices, expectations, and capital flows. Once that loop becomes dominant, traditional measures like earnings or cash flow can matter less in the short term than momentum and belief.

Common Ingredients in Past Bubbles

While every bubble has its own context, they often share structural similarities. Importantly, these similarities appear across very different sectors and time periods, suggesting that bubbles are not anomalies but recurring financial dynamics.

One of the clearest historical examples is the Railway Mania of the 1840s. Railways were a genuine technological breakthrough that transformed trade, mobility, and industrial production. However, massive amounts of capital were deployed on the assumption that demand and profitability would match the speed of expansion. In practice, many routes were overbuilt, returns were uneven, and expectations about short-term profitability were overly optimistic. The key point is that railways were not a “bad idea”. They were a transformative infrastructure system whose financial expectations moved ahead of reality.

A more modern example is the dot-com bubble of the late 1990s. The internet was a genuine technological shift with long-term economic significance. However, capital flooded into companies simply because they were associated with the internet, regardless of whether they had viable business models or earnings potential. Valuations expanded rapidly, often based on traffic, growth projections, or narrative appeal rather than profitability. Many firms failed, but the underlying technology did not. The issue was timing and pricing, not direction.

Crypto markets offer a more recent example of similar behaviour, though with a different mechanism. Rather than infrastructure or corporate earnings, crypto cycles have been heavily driven by liquidity conditions, retail participation, and narrative momentum. Prices have often moved sharply in response to sentiment shifts, with booms and corrections occurring in relatively short cycles. While blockchain technology has clear use cases, pricing has frequently reflected expectations and liquidity conditions more than adoption or cash-flow generation.

Across all three cases, a consistent structure appears. Each involved a genuine underlying shift, whether technological or financial. Each was characterised by uncertainty about future winners and timelines. Each attracted significant inflows of capital. And each saw narratives grow stronger as prices rose.

The common thread is not irrationality, but acceleration. When capital, narrative, and expectation reinforce each other, pricing can detach from measurable outcomes for extended periods.

Where AI Fits In?

Artificial intelligence sits in a more complex position than most historical examples because it contains elements of all three patterns.

In some ways, AI resembles the railway era. It is not purely a software phenomenon but a capital-intensive infrastructure buildout. The expansion of data centres, semiconductor supply chains, and energy demand reflects a physical investment cycle similar to earlier industrial transformations. Recent estimates suggest that global investment in AI infrastructure is already running at hundreds of billions of dollars annually, with further growth expected as demand increases . This creates uncertainty around long-term utilisation, efficiency, and returns on investment, similar to earlier infrastructure booms.

In other ways, AI resembles the dot-com period. It is widely seen as a general-purpose technology that will reshape industries, productivity, and business models. This has led to rapid valuation expansion in both public and private markets, often driven by expectations of future dominance rather than current profitability. As with the internet, there is little disagreement that the underlying technology is real. The uncertainty lies in how quickly value will be captured, and which firms will ultimately benefit.

Finally, AI also shares characteristics with crypto cycles. The speed of capital inflows, the intensity of media attention, and the momentum-driven nature of certain market segments all contribute to rapid repricing. Investor behaviour can become self-reinforcing, particularly in sectors where future outcomes are highly uncertain and widely debated.

At the same time, there are important differences that make AI harder to classify. Unlike many historical bubbles, AI is already being integrated into real business processes across industries. Productivity gains are being reported in specific use cases, and major technology firms are generating substantial revenues while simultaneously investing heavily in further expansion. This means the boundary between speculative expectation and actual economic utility is less clear than in previous cycles.

The result is a more ambiguous picture. AI shares characteristics with historical bubbles, but it also sits on top of genuine and already-deploying technological infrastructure. That combination makes it difficult to determine whether current pricing reflects excessive optimism or a rational, if highly uncertain, discounting of future productivity gains.

Are We in a Bubble?

The challenge with bubbles is rarely identifying them in hindsight. It is recognising the conditions that tend to produce them while they are still forming.

Financial history suggests that bubbles are not rare exceptions. They are recurring phases that emerge when genuine technological or economic shifts interact with uncertainty, capital availability, and rapidly changing expectations.

Artificial intelligence may ultimately prove to be one of the most important technological developments in decades. It may also pass through periods where expectations outpace measurable outcomes. Both things can be true at the same time.

The more useful question is not whether a bubble exists in isolation, but how far expectations are moving relative to what can realistically be delivered in the near term.

💼 Unpacked

Hyperscaler

A hyperscaler is a company that operates massive cloud computing and data centre infrastructure at global scale. Firms such as Amazon, Microsoft and Google are considered hyperscalers because they can rapidly expand computing capacity to support millions of users and businesses.

Capital Expenditure (CapEx)

Capital expenditure, or CapEx, is money spent by a business on long-term assets that are expected to generate value over many years. Examples include building factories, purchasing equipment, constructing data centres, or upgrading infrastructure.

Market Cycle

A market cycle describes the recurring pattern of expansion and contraction in asset prices over time. Cycles often move through phases of optimism, growth, peak valuations, decline, and recovery.

📣 Support The Fiscal Compass

If you found this insightful, consider sharing with friends or colleagues. For weekly economics-led takes on markets, policy, and macro trends, subscribe to The Fiscal Compass.

Follow along on social media for concise updates throughout the week:

Instagram: @thefiscalcompassofficial

X: @FiscalCompass.

LinkedIn: Vinay Meisuria