CommSec
10 August 2026
Author: James Gruber is Equity Market Strategist at CommSec
The term ‘bubble’ is used a lot these days. But what is a bubble? And are we experiencing one now in artificial intelligence-related stocks?
The US semiconductor index – consisting of chip companies which provide the compute power for AI – has more than doubled over the past year – the only other time that has happened was in 1998-2000 at the height of the dotcom bubble and that did not end well.
A poster child for the index, Micron, is now trading at 19x sales, which is about double its peak valuation during the so-called ‘dot-com bubble’ of 1999-2000. Back then, DRAM (Dynamic Random-Access Memory) prices – tracking prices for memory used for temporary data storage in computers - collapsed and earnings evaporated, leading Micron to lose over 98% of its market value.
The question today is whether this is a bubble, or is it a structural trend that will prove more sustainable?
To answer the question, this article will examine the frameworks used in two classic books on market bubbles, Edward Chancellor’s Capital Account and William Quinn and John Turner’s Boom and Bust.
The capital cycle
In the late 1990s/early 2000s, UK hedge fund Marathon Asset Management gained fame for calling out the telecom/internet bubble and detailing the numbers for why it would falter.
In Capital Account, Chancellor collates the shareholder letters of Marathon from that period.
In these letters, Marathon outlines its framework for assessing investment cycles and bubbles.
Its so-called capital cycle theory suggests that financial markets repeatedly misallocate capital because investors extrapolate recent success into the future. Booms encourage overinvestment, which eventually destroys profitability. Marathon thought investors who pay attention to the capital cycle could identify when industries become overcrowded and when neglected sectors were poised for recovery.
It outlined the capital cycle as:
1. High returns attract investment.
2. Companies expand capacity.
3. Competition increases.
4. Returns fall.
5. Capital exits.
6. Capacity shrinks.
7. Returns recover.
According to Marathon, investors focus too much on demand in an industry, when the biggest swing factor is supply, especially the supply of money or capital.
The fund’s capital cycle matrix is a powerful tool to assess when industry excess and speculation may be gathering steam.
To see why, Marathon’s thesis on the telecoms/internet bubble on the 1990s is one example.
The fund detailed how they thought that the internet revolution was real, but the investment boom was so large that it had created excess capacity, competition, and debt. At the time, Marathon laid out the numbers behind the excess supply and how demand was unlikely to catch up in the short-term. Conversely, most investors at the time were focused on demand dynamics in the industry.
Marathon theorised that industry supply would have to be significantly cut, returns would shrink, and profits would crash. That would be exacerbated by investment leaving the industry. Businesses would have to consolidate to survive, and only then would returns start to normalise and turn up again.
As history shows, Marathon was accurate in its portrayal of the rise and fall of the telecoms industry.
The bubble triangle
Where Chancellor’s book focus on industry excesses, Quinn and Turner’s Boom and Bust specifically explores financial bubbles – what they are, and their primary characteristics.
The authors use fire as a metaphor for a financial bubble. The formation of a fire can be described in simple terms using a fire triangle, which consists of oxygen, fuel, and heat. Given sufficient levels of these three things, a fire can be started by a spark.
The book proposes a bubble triangle, where financial bubbles happen when three elements combine:
1. Marketability – assets can be easily traded
2. Money/credit – investors have access to financing
3. Mania or speculation – investors buy primarily because they expect prices to rise.
Their framework is often summarised as the 3M’s:
Marketability + Money + Mania = Bubble
As for a spark for a financial bubble, the authors believe that it usually comes from either the emergence of a new technology and government policy/intervention.
And a bubble bursts when one side of the bubble triangle weakens.
As a practical example, the book uses the triangle to explain the 2008 Global Financial Crisis. The thesis is that the crisis was not primarily caused by irrational borrowers or a single policy mistake. Instead, it emerged because the three sides of the triangle lined up:
1. Marketability: mortgages became tradable securities
Traditionally, a bank originated a mortgage and held it on its balance sheet. During the housing boom, mortgages were increasingly securitised into mortgage-backed securities (MBS), collateralised debt obligations (CDOs), and related products.
This dramatically increased the marketability of housing debt. Mortgages could be packaged, sold, and resold globally, attracting investors who never interacted with the underlying borrowers.
2. Money/Credit: abundant liquidity and easy lending
After the dot-com crash and the 2001 recession, interest rates were low and global savings were plentiful. Banks and shadow banks could obtain funding cheaply.
The result was an enormous expansion of credit:
- lower lending standards
- subprime mortgages
- teaser rates
- high leverage throughout the financial system
Credit expansion supplied the fuel for rising house prices.
3. Speculation: belief that house prices could only rise
As prices climbed, expectations became self-reinforcing.
People bought homes because prices were rising.
Investors bought mortgage securities because housing seemed safe.
Lenders relaxed standards because rising house prices appeared to protect them from losses.
The key speculative narrative was essentially: House prices do not fall nationally.
Once that belief became widespread, participants stopped focusing on fundamentals and focused on expected future price appreciation.
Why did it crash? Because:
- house price growth slowed
- mortgage defaults rose
- confidence in mortgage securities evaporated
- funding markets froze
- leverage amplified losses
How the AI boom compares
Does the current AI boom constitute a bubble?
Running through Quinn and Turner’s bubble triangle:
1. Marketability: trading has become easier.
Investors have gained greater access to publicly listed AI companies through online trading platforms, micro investing firms and Exchange-Trade Funds (ETFs), including leveraged ETFs.
They have also been able to access the privately listed AI companies as private funds have become increasingly open to retail investors.
These factors have undoubtedly helped fuel the enthusiasm for AI-related firms.
2. Money/Credit: hundreds of billions in private and public money
The large AI firms have been funded by hundreds of billions in venture capital and private equity money.
OpenAI has raised more than US$160 billion over the past 15 months from the likes of Amazon, Microsoft, and Nvidia.
Meanwhile, Anthropic raised US$30 billion in February this year, bringing its cumulative funding to roughly US$60 billion since its founding in 2021.
SpaceX also raised about US$12 billion cumulatively in private markets before going public in June, raising a further US$86 billion.
The Magnificent Seven stocks have also tapped debt markets to fund their massive spending on data centres and other AI infrastructure, expected to total almost US$700 billion this year.
Recently, Alphabet, the parent company of Google, raised US$85 billion for investment in AI.
3. Speculation: hype around AI stocks has intensified
It seems many investors have come to see US AI stocks as one-way bets. Chip makers – which provide the chips for AI computation – have had a great run of late. Micron has risen more than more than 8x while Sandisk has soared about 28x over the past 12 months.
Even giants such as Alphabet, seen as being at the forefront of AI technology, have seen massive gains – with the stock jumping ~85%% over the past years.
This has been accompanied by growing speculation in smaller stocks. Some companies that have switched their core businesses to AI, such as Allbirds and Myseum, have seen sharp one-day gains on the news.
That is not to mention what has happened in private markets. Within weeks of forming in early 2025, Thinking Machine Labs, headed by former OpenAI CTO, Mira Murati, raised US$2 billion at a US$10 billion valuation, and that was without the company revealing exactly what it was building.
The spark for the AI boom has obviously been the rise of a new technology, with AI coming to public attention when ChatGPT was unveiled in November 2022, and it gathered further steam when Nvidia released earnings results in May 2023 and forecasted explosive growth in coming quarters.
The question marks for the AI bubble theory
There are three queries about whether AI is conclusively a bubble. The first is about money or credit. It is something that Edward Chancellor’s book emphasises more: that any boom or bubble is accompanied by low interest rates, abundant liquidity, and easy financing.
The issue today is that interest rates are not that low, and financing is not that easy. Yes, there is plenty of money in private markets but a lot of that is equity rather than debt. Only recently have AI-related firms started to access debt markets to fund their plans.
Second is something that Chancellor also focused on more: that any boom or bubble involves significant over-investment or malinvestment. There is certainly an extraordinary degree of investment in AI at present. But whether it is excessive compared with demand is difficult to assess. No one quite knows what the demand for AI will be in the short or long term, so definite conclusions about the current investments are difficult.
The third question mark is around earnings. Today’s AI leaders such as Alphabet, Amazon, Microsoft, and Nvidia, are highly profitable. Some pundits believe that is a big contrast from the internet era, which had some loss-making firms that quickly went bust.
It is a less convincing argument as the late 1990s had several large, highly profitable, and high-growth companies including Intel and Cisco.
Edward Chancellor also addresses the earnings question in his book, saying the high returns from companies are often a prerequisite for excessive investments, and greater competition filters through an industry to bring down both margins and profits.
Ultimately, whether AI proves to be a bubble will only become clear with time.
Back to 
