Author: an investment professional (guest contributor)
When should investors exit a rapidly rising market segment, and how can one tell if prices are in a bubble? This analysis assesses the AI-driven stock rally through the lenses of investment cycles, valuation measures, market concentration and the Buffett indicator. The picture is mixed: corporate profitability remains strong, but the hyperscalers’ free cash flow is collapsing, raising notable risks.
Reflexivity and the timing problem
George Soros’s reflexivity theory highlights that market participants can reinforce a price trend by acting on expectations: buying into rising markets can itself push prices higher. The momentum effect—where recent winners continue to outperform—is an expression of this dynamic. The harder question is when to sell: after 100%, 350% or 1,000% gains? Or do improving fundamentals justify continued holdings?
Capex as an indicator: a chicken-and-egg dilemma
Planned investment (Capex) related to AI—data centers and supporting infrastructure—is a key datapoint because without such buildout expected demand cannot be satisfied. Expectations for these investments followed an exponential path, and if realized spending persistently undershoots plans, profit forecasts will be revised down and can trigger selling.
Historical experience shows, however, that by the time investment data reflect a slowdown it may already be too late: markets often turn before the capex statistics reveal the change. Correlation therefore does not prove causation, but markets price the future, so even signals that capex might stall can provoke market declines.
What’s different this time? Big expectations, limited macro evidence so far
AI-related expectations are large, yet broad-based macro-level productivity gains have not yet materialized. Some investors expect a repeat of past technology-driven booms and busts; others argue AI can deliver unprecedented efficiency gains, potentially justifying continued high valuations.
On forward-looking P/E metrics the MSCI USA Technology index is not as expensive as it was at the 2000 peak. Still, forward metrics reflect market expectations, and reflexivity implies that market beliefs can shape fundamentals rather than merely reflect them.
Technology’s outperformance and a changing environment
US tech companies have delivered much stronger profit growth than the rest of the market over the past decade. The MSCI global technology index’s trailing 12-month EPS has outperformed broad global indices that have seen protracted stagnation.
The environment, however, has shifted: yields have risen, and large tech firms—Meta, Microsoft, Google (Alphabet), Amazon and Oracle among them—are planning and spending hundreds of billions on concrete, cooling and rapidly depreciating chips. These investments are increasingly financed via off-balance-sheet commitments and bond issuance, while free cash flow is shrinking sharply.
Financing, bond markets and falling free cash flow
Hyperscalers are active in non-dollar bond markets: this year they issued 22% of investment-grade Swiss franc bonds, 10% of sterling issues and 3% of euro-denominated investment-grade bonds. Financing needs are crowding out share buybacks and dividends.
Analysts expect the combined free cash flow of Microsoft, Amazon, Google, Meta and Oracle to fall from a near-$300 billion peak in 2025 to roughly $20 billion by Q1 2027—a dramatic decline.
Concentration and the potential to manipulate competition
The success of a handful of firms has produced heavy concentration in profits and market capitalization—greater than during the dot-com bubble or the Japanese equity peak of the 1990s. Historically, technological innovation attracts many entrants and capital, leading eventually to overcapacity and lower returns for incumbents; a small group of market leaders can then fall together.
The spread of Chinese open-weight models is one manifestation of this risk: underestimating competitive pressure and overestimating returns are recurring themes in past innovation-driven bubbles. Competition can also be limited by regulation: geopolitical and national-security concerns may lead jurisdictions to restrict foreign models, and that fragmentation could create artificial oligopolies among the remaining providers.
Notably, Anthropic, OpenAI and xAI executives have called—unusually in concert—for self-restraint on large language model development, citing safety risks; that stance could presage regulatory involvement.
What does the Buffett indicator say?
The Buffett indicator—the ratio of equity market capitalization to GDP—is useful for assessing long-term valuation relationships. Using the Wilshire 5000, the indicator stands near 240%, versus about 150% at the 2000 peak. Mason’s decomposition breaks this ratio into four factors: corporate value added/GDP, profits/value added, payouts (net of issuance), and the valuation multiple.
Since 2020, corporate profitability—largely driven by the technology sector—has been a principal contributor to the high ratio. If companies meet profit expectations, the market rise can be fundamentally supported despite high concentration. But payout prospects look less favorable: the very firms with the largest profits are tying up free cash flow in AI investments, limiting buybacks and dividends.
Google has even financed some of its capex through equity issuance during the summer, and further large issuances from OpenAI and Anthropic are anticipated.
On valuation multiples alone, current measures and the Mason decomposition do not indicate the same degree of overvaluation that preceded the major market collapses of the 1970s or 2000s.
Possible outcomes and practical takeaways
Pinpointing a bubble top is difficult, and recognizing a bubble before a crash is even harder. Historical episodes—such as the UK railway boom of the 1840s—show that rising short-term dividends and prices can mask overvaluation until the collapse begins. First-mover advantages can quickly become crowded and then eroded.
From an investor’s standpoint, perfect timing may be unattainable, but prudence is warranted when expectations and capital are heavily concentrated in a single trend and a small number of firms. That said, technological progress often outlives the companies that pioneered it: later entrants may commercialize innovations successfully and deliver durable products and services.
Note: guest contributors are independent experts and their views do not necessarily reflect those of Forbes’s editorial staff.
