6 Hyperscalers Driving the AI Revolution Are Projected to Spend $1.3 Trillion on Capex in 2027. Only 1 Is Forecast to Have Positive Free Cash Flow.

S&P Global projects that major hyperscaler AI capex will rise to ~$1.3T next year (from $870B this year and $470B in 2025), increasing AI capex by ~50% in 2027. Under its model, only Microsoft is projected to generate positive 2027 free cash flow, while peers show negative FCF (e.g., Alphabet -$82.7B, Amazon -$60.1B, Meta -$3.5B; SpaceX -$114.4B). The article frames weakening FCF from heavy AI spending as a balance-sheet headwind, offset by Microsoft’s AAA credit rating and potential lease-accounting effects.

Analysis

The market implication is less about "AI is expensive" and more about which balance sheets can absorb a multi-year capital cycle without sacrificing buybacks or credit quality. That favors the names with the deepest internal funding pools and the cleanest financing optionality, while making the higher-lease / lower-visibility operators look increasingly like duration trades rather than compounders. In practice, the first-order winners are the infrastructure suppliers and the strongest cloud franchises; the second-order loser is equity multiple expansion in the spend-heavy cohort as FCF stops being a reliable valuation anchor.

The more interesting spillover is in credit. When capex is funded through leases and other off-balance-sheet structures, the pain migrates from headline net debt to fixed-charge coverage, which equity investors often ignore until spreads move. That means the next 1-3 months of catalysts are not just earnings calls, but bond market pricing, lease disclosure, and whether management teams start narrowing the gap between capex growth and revenue growth; if they do not, the market will likely punish any incremental "investing for the future" narrative.

Contrarianly, consensus may be underestimating how much of this spending is defensive rather than optional: if AI usage continues compounding, the current FCF trough could prove temporary and the market is too early calling peak capex. But if monetization lags the infrastructure build by even a few quarters, the valuation hit can be severe because investors will re-rate these as low-ROIC utility-like assets. The key falsifier is simple: evidence that incremental spend is converting into higher-margin revenue growth by the next earnings season; absent that, the trade stays toward relative underperformance for the most capex-intensive names.

Key Decisions for Investors

• Long MSFT / short ORCL for 1-3 months: MSFT has the best funding flexibility and should hold up better if the market starts distinguishing "self-funded" AI buildouts from balance-sheet-stretching ones; target 8-12% relative outperformance, exit if ORCL shows faster-than-expected FCF conversion or tighter credit spreads.
• Long NVDA vs a basket short of AMZN, GOOG/GOOGL, META over the next earnings cycle: express the view that the pick-and-shovel supplier benefits while hyperscaler equity returns get diluted by capex drag; risk/reward improves if AI spend remains elevated but monetization timing stays vague.
• Avoid chasing the broad hyperscaler basket on strength; use rallies in AMZN/GOOG/GOOGL/META to trim exposure until managements show capex intensity is flattening and incremental revenue is inflecting. The setup is vulnerable to multiple compression if FCF guidance is revised lower again.
• Watch ORCL and SPCX financing conditions as an alert, not a full trade: if spreads widen or lease-heavy financing gets more expensive, that is an early signal the market is repricing AI capex from growth story to leverage story.