An AI Bubble Burst Scenario You Should Watch!

At Straterix, we have built some of the most powerful tools for generating forward looking scenarios. These tools not only enable more efficient stress tests and risk-aware strategic planning, but also make it possible to ask meaningful “what if” questions and obtain highly insightful results within minutes. Given the widespread concern that current AI related valuations may be displaying bubble-like characteristics, we ran such a "what-if" query - and are sharing some of our findings here.

Here is how we do it – and what we found.

To construct an initial adverse shock, we defined a “what‑if” scenario calibrated to the deterioration observed in the technology sector during the early‑2000s dot‑com bust. There are, of course, many possible objections about the historical similarities, but we project only a few historical observations of a particular crisis and expand all other variables incorporating modern world dynamics. We observe, learn from it, and then, with the right tools, test alternative configurations within minutes.

From a comprehensive set of forward‑looking, internally consistent scenarios that incorporate various combinations of shocks and their consequences, we extract the cluster in which the technology sector follows a prescribed trajectory. All variables in the framework are generated jointly and coherently and incorporate dynamic correlations. This enables AI algorithms to trace the implied paths not only for traditional macro‑financial indicators, but also for asset classes and variables that did not exist in the original episode, such as cryptocurrencies.

In parallel, the framework enables a direct comparison between historical outcomes and current projections for well‑established variables—such as equity market performance, credit spreads, GDP growth, and unemployment—highlighting both structural similarities and important regime differences between the dot‑com era and today’s environment. This dual perspective provides a disciplined way to assess downside risks from a potential AI‑led correction, while avoiding simplistic analogies to past bubbles.

In the early‑2000s episode, the Information Technology sector fell by more than 70% from its peak in 2000 to its trough in 2003 over a 12‑quarter horizon. Applying a decline of this magnitude to today’s technology sector, the projected path over the next 12 quarters looks as follows (Figure 1):

Figure 1: Projection of a Tech Sector decline in an AI bubble burst scenario

From the cluster of scenarios extracted from the full-range scenario analysis, we can obtain the values of any macroeconomic or market variable. Below is a sample extract (Table 1):

This is a recessionary scenario with negative GDP growth, unemployment exceeding 7%, 66% drop in Private Equity and 30-45% in most market indices. Credit spreads are more than doubling and delinquency rates in CRE sectors increasing by 90 bps. Bitcoin drops by nearly 30%.

Comparing the variables that existed 25 years ago against this scenario (Table 2), we can see that the forecasted values somewhat exceed the historical ones for unemployment, CRE delinquencies and investment grade spreads while trailing a bit lower for high-yield spreads. The initial values for the current credit spreads are substantially lower than the spreads before the dotcom crisis. 

Table 2: Comparison between projected AI bubble burst scenario (12 quarters) and the dotcom crash (2000-2003)Table 2: Comparison between projected AI bubble burst scenario (12 quarters) and the dotcom crash (2000-2003

This is just one example showcasing a handful of core variables out of hundreds that are automatically generated. Using the Straterix platform, banks and insurance companies analyze “what-if” scenarios in minutes allowing them to effectively explore both the anticipated futures as well as highly hypothetical ones, providing both confidence and resilience no matter what the future holds.

Table 1: Selection of projected variables for an AI bubble burst scenario over 12 quarters

This is a recessionary scenario with negative GDP growth, unemployment exceeding 7%, 66% drop in Private Equity and 30-45% in most market indices. Credit spreads are more than doubling and delinquency rates in CRE sectors increasing by 90 bps. Bitcoin drops by nearly 30%.

Comparing the variables that existed 25 years ago against this scenario (Table 2), we can see that the forecasted values somewhat exceed the historical ones for unemployment, CRE delinquencies and investment grade spreads while trailing a bit lower for high-yield spreads. The initial values for the current credit spreads are substantially lower than the spreads before the dotcom crisis.Table 2: Comparison between projected AI bubble burst scenario (12 quarters) and the dotcom crash (2000-2003)Table 2: Comparison between projected AI bubble burst scenario (12 quarters) and the dotcom crash (2000-2003).

Table 2: Comparison between projected AI bubble burst scenario (12 quarters) and the Dotcom crash (2000-2003)

Alla Gil

is the Founder and CEO of Straterix Inc., a firm focused on financial modeling and risk analytics. A financial risk expert, she has also held senior leadership roles at Goldman Sachs, Nomura, and Novantas. Her work focuses on risk management, stress testing, and helping organizations navigate complex financial risks.