The Two Cycles of AI
Mapping the Financial Bubble and the Technological Boom
Aims
This project aims to decouple the current AI market boom into two distinct phenomena: the Financial Cycle (speculation) and the Technological Cycle (utility). By visualising valuation metrics alongside real-world economic and labour data, this analysis seeks to demonstrate that a potential financial correction will not derail the underlying technological integration, mirroring the post-2000 trajectory of the internet era.
Section I: The Financial Cycle (The Bubble & Risk)
This section establishes the market's current state of extreme risk and valuation concentration using four core financial metrics.
When indexed against the broader S&P 500, the trajectory of the "Magnificent 7" reveals a sharp, non-linear divergence that mirrors the dynamics of the 1999 Dot-com Bubble. This gap represents the core of the financial risk: market enthusiasm is narrowly concentrated in a few entities, driven by future faith rather than current broad-based growth.
This matrix explains why the divergence exists. By mapping future growth against current P/E ratios, I isolate the Magnificent 7 as the epicenter of speculation. The valuation premium establishes immediate structural risk: these firms are priced for perfection. Any failure to deliver exponential growth will compress these multiples, regardless of the underlying technology's utility.
To gauge the systemic magnitude of this boom, I use the Cyclically Adjusted P/E (CAPE). The current ratio places the market in historically extreme territory—second only to the peaks of 1929 and 2000. This provides the objective context: the financial cycle is operating above the "Red Zone" (30x), a threshold that has traditionally preceded severe market corrections.
Finally, I visualise the "spending vs. earning" imbalance for the four Hyperscalers (Amazon, Alphabet, Microsoft, Meta). I exclude Apple, Nvidia, and Tesla because their Capex is driven primarily by consumer products and chip fabrication, not the build-out of AI cloud infrastructure. The colossal gap for the Hyperscalers proves the current financial model is speculative; tens of billions are flowing out in Capital Expenditure with minimal immediate return.
Methodology Note: The AI Spend (Red) is estimated as 90% of TTM Capex for Hyperscalers (sourced via Yahoo Finance API). This 90% ratio is based on the industry consensus and analysis by firms such as Morgan Stanley (Lichtenberg, 2025), reflecting the immense concentration of capital expenditure growth in generative AI infrastructure. The Implied AI Revenue (Green) is modeled at 15% of that spend.
Section II: The Technological Cycle (The Reality & Survival)
This section provides evidence that the technology itself is deeply entrenched and will survive a financial correction, concluding the thesis.
Can a financial bubble burst while the technology survives? Drag the slider. At the start, Amazon's 90% stock collapse looks like a terminal disaster. But as you slide to the present, that "massive" crash shrinks into a blip. This proves the core thesis: the Financial Cycle creates volatile bubbles that burst, but the Technological Cycle drives exponential utility that eventually dwarfs the initial hype.
While the financial cycle remains unstable, the technological cycle is solidifying on the ground. This map, based on job interest and business intent, shows that AI adoption demand is not universal but is highly concentrated. The darkest regions—the Northeast and West Coast knowledge hubs—are the real human capital pools driving the transition. This geographic entrenchment confirms the technology is profoundly real, suggesting it will endure even if the financial bubble bursts.
I test this entrenchment against the real economy. This scatter plot reveals a statistically significant link (p=0.002) between state wealth (GDP per Capita) and AI adoption. Unlike the financial bubble, which relies on speculation, the technological cycle is supported by measurable economic activity. The transition is capital-intensive, favouring states with established economic power.
Finally, the labour market confirms the structural shift. While the regression proves the trend, this box plot reveals the certainty. I observe distinct, non-overlapping salary bands, where the "floor" for Senior AI roles frequently exceeds the "ceiling" for Mid-level positions. The Technological Cycle has fundamentally re-priced human capital, creating a stable, high-value asset class of labour that exists independently of stock market volatility.
Project Challenges
The most significant challenge was the scarcity of granular, open-source data; much of the desired financial and labour market information was paywalled. Additionally, technical restrictions within the analysis environment frequently blocked external data fetching. These hurdles necessitated the cleaning and self-hosting of datasets to ensure the regression analysis remained statistically valid and reproducible.
Concluding Remarks
The data confirms the thesis: while the financial valuations of the sector remain precarious and historically stretched, the underlying technology is geographically and economically entrenched. The bubble may burst, but the utility is permanent.
Data Sources & References
Yahoo Finance API: Historical stock prices, quarterly Capital Expenditure (Capex) data for Hyperscalers. Accessed via Python yfinance library.
Robert Shiller Online Data: Cyclically Adjusted Price-to-Earnings (CAPE) ratio historical dataset. Yale University.
ai-jobs.net (via Kaggle): Global AI, Machine Learning, and Data Science Salaries (2020–2025). Used for salary band regression analysis.
Anthropic Economic Index: US GDP per Capita proxies used for the economic validation scatter plot. Claude.
Lichtenberg, N. (2025). "75% of gains, 80% of profits, 90% of capex—AI's grip on the S&P is total and Morgan Stanley's top analyst is 'very concerned'." Morgan Stanley Research. Yahoo Finance.