- Michael Allison, CFA
- Oct 5
- 2 min read
By Michael Allison, CFA
The AI Investment Boom: Gauges for Investors
As you may recall, I’ve recently become increasingly focused on looking for signs that the current AI spending boom is starting to come to an end - looking for canaries emerging from the coalmine of Hyperscaler financial statements.
As a veteran of the 1990s internet boom and bust, having served as an buy-side equity research analyst for a major asset manager in Boston, I’ve tried to consider in what ways this cycle is like that one, and in what ways it is not. I ask myself, “As we sit here in 2025, is it more like 1996 or 1999? I’m increasingly inclined to believe it is more like the former and this spending cycle can potentially run for several more years.
In my research seeking to support or refute that view, I was provided with a “Eureka!” moment when I ran across an article entitled, “Is AI a Bubble?” on the Exponential View Substack. Authors Azeem Azhar and Nathan Warren offer a thoughtful and practical framework containing five gauges that help answer the question posed in the title of their article. I think it can be a very helpful way for us to think about the progression of the current AI cycle. The article is behind a paywall ($$$) and is quite lengthy. If you’re interested in reading the whole thing, email or message me and I’ll send you a PDF version.
Here’s my perspective on the article and the five gauges for AI spending depicted in this week’s Chart:
Economic strain: AI data‑center capex is climbing fast. Morgan Stanley projects AI infrastructure spending could reach $800 billion by 2030, nudging U.S. investment toward 1.6% of GDP. That’s still green by historical standards, but edging into amber. Unlike railways, AI hardware depreciates in dog years, compressing the window to earn a return.
Industry strain: the ratio of capex to revenues is high. Global data‑center spending around $370 billion supported by roughly $60 billion in AI revenues—a six‑to‑one mismatch. That lands in the amber zone. Yet demand is voracious; enterprises book capacity before it’s built and revenues are doubling yearly.
Revenue growth: this gauge is solidly green. GenAI revenues could double this year and continue compounding; some analysts forecast a trillion‑dollar market by 2028. Adoption is still nascent—only about 9% of U.S. firms have a useful AI use case today—leaving ample runway.
Valuation heat: tech valuations are elevated but far from dot‑com extremes. The Nasdaq trades around 32× earnings versus 72× at the 2000 peak.
Funding quality: hyperscalers like Microsoft and Amazon can largely self‑fund, but future spend will require trillions in private credit and securitizations. CoreWeave’s debt troubles illustrate the fragility creeping in.
Conclusion: these gauges suggest a demand‑led, capital‑intensive boom, not a bubble—yet. But if investment tops 2% of GDP, P/E ratios expand toward 60×, or funding quality deteriorates, two needles could flash red. For investors, that means enjoy the ride for now, but keep one foot near the brake.
Interested in reading more of Mike's weekly newsletters? Click below to view The Sunday Drive.
