AI is one of the most transformative technologies of modern times, with a potential on par with the railway, electricity and the internet. AI not only increases productivity, but also reshapes decision-making processes, organisations and ultimately power structures. For this reason, AI also follows a historically recognisable pattern: high expectations, massive capital allocation, and uncertain timing of returns. That dynamic is well known, but what is new is the scale, the speed and the concentration. So what could trigger a break in the AI bubble?
This is the second of two blog posts on the topic.
There are also risks related to value creation …
AI also risks becoming a commodity. For most users, the differences between systems such as ChatGPT, Claude and Gemini are relatively marginal. This increases the risk that monetisation potential declines further, particularly for OpenAI and Anthropic. Gemini, by contrast, can leverage Google’s targeted advertising model to support self-financing.
There are also signs that top AI talent is increasingly moving between companies. This mobility contributes to the spread of know-how across actors, reducing durable competitive advantages and increasing the likelihood that the technology shifts from differentiated innovation to broadly accessible capacity. As a result, both open-source and smaller players can close the gap more quickly.
AI may therefore follow the pattern of previous technologies: shifting from differentiated innovation to widely available capacity, a form of commodity. Value creation then moves rapidly from model development to distribution, data and customer access. This favours companies with existing cash flow and loyal user bases, rather than players such as OpenAI and Anthropic.
… and the financing model is showing signs of strain
AI investments depend on both new capital and the continuous refinancing of existing investments. If this chain is disrupted, even fundamentally viable projects can become illiquid. Projects with long-term potential may be forced to scale down or halt, as their time horizon can no longer be financed.
Until now, there has been an abundance of risk-tolerant capital. Big Tech has invested heavily, but private credit has been particularly aggressive. This includes large institutional investors such as SoftBank, as well as more traditional private equity funds.
There are now increasing signs of distress among private credit investors. In 2026, firms such as Ares Management and Apollo Global have begun limiting withdrawals by their clients. Such restrictions are always a concern, and similar measures have been introduced by Blue Owl and Cliffwater.
This makes the system sensitive to:
- Changes in interest rates (and inflation)
- Geopolitical fragmentation
- Confidence in US assets (e.g. Treasuries), which is currently declining due to a more transactional and less alliance-based US policy
Overall, this increases the likelihood that a correction evolves into more than a sector rotation, becoming a broader financial event.
The AI boom increasingly resembles dot-com
The similarities to the dot-com era are clear:
- Overinvestment in infrastructure
- Unclear business models
- Narrative-driven capital allocation
The key difference, however, is that AI is far more capital-intensive and geopolitically integrated. This means that a correction could be deeper and have broader effects than “just” equity markets. Credit markets, energy consumption and supply chains would also be affected. The time horizon may therefore be longerreducing the ability of central banks to ensure financial stability without risking overstimulation of the economy.
Conclusion: Two simultaneous truths
A break in the AI bubble is unlikely to have a single cause, but rather to emerge from a combination of multiple simultaneous disruptions, as outlined above. At present, two narratives are both true:
- AI will fundamentally transform economies and societies
- The current investment structure is fragile
The key question is therefore not whether AI will be transformative, but when value will be realised, and who will survive until then.
The AI bubble is not primarily driven by hype, but by a logical, yet potentially flawed, assumption: the scaling hypothesis. If this paradigm has reached its limits, and financing simultaneously becomes more fragile, the probability of a correction in AI equities increases. As these are broadly held, such a correction is likely to be prolonged and lead to consolidation within the sector.
At the same time, value within the AI value chain will shift, from model development to actors with distribution, data and existing cash flows. In a world shaped by geopolitical fragmentation, the key question is therefore not only whether a technology bubble may burst, but how such a break will accelerate an already ongoing fragmentation of the global economic system.