Source Report 11

Hyperscaler Capex and Free Cash Flow: Q3 2026 and 2027 Outlook

Full research prompt

What were combined hyperscaler capex and free cash flow for Q3 2026, versus the $527B 2026 capex figure, and what is guided for 2027?

From Is there an AI Bubble?

Jon Sinclair using Luminix AI
Jon Sinclair using Luminix AI Strategic Research

The four core hyperscalers (Amazon, Alphabet, Microsoft, Meta) spent roughly $202–207 billion combined on capex in calendar Q3 2026 (ended Sept. 30), with five-company totals (adding Oracle) reaching ~$230 billion; this pace already exceeds earlier 2026 consensus estimates and is pressuring free cash flow, with aggregate spending outrunning operating cash flow in the quarter.[1][2]

  • Amazon reported $75.9–77.3 billion in cash/gross purchases of property and equipment (primarily AWS/AI-related).[1]
  • Alphabet: $51.4 billion in purchases of property and equipment.[3]
  • Microsoft: $39.4–45.2 billion (including finance leases; calendar-quarter figure per CFO commentary).[1]
  • Meta: $33.9–34.9 billion (including finance-lease principal).[3]
  • Combined four-company Q3 total: ~$201.9–207.3 billion (up ~84–108% YoY depending on exact measures and inclusions). Five-company (incl. Oracle) reached ~$230 billion.[4]

Free cash flow turned negative or sharply lower for most in Q3 or on a trailing basis, as capex outpaced operating cash flow generation (one report notes five hyperscalers’ Q3 capex at $230.4B vs. $198.6B OCF).[5]

  • Alphabet: –$9.1 billion quarterly FCF (second straight negative quarter; TTM still positive in some reports).[3]
  • Meta: –$2.1–2.12 billion quarterly FCF.[3]
  • Amazon: TTM FCF swung to –$42.4 billion (from +$14.8 billion prior year).[3]
  • Microsoft: +$18.6 billion quarterly FCF (the outlier among the group).[6]
  • Broader trend: Capex consumed well over 100% of OCF for several firms in the quarter or TTM, driving reliance on debt, equity raises, and leases.[7]

The $527 billion 2026 capex figure referenced was Goldman Sachs’ Wall Street consensus estimate as of December 2025; actual company guidance has since been revised sharply higher to ~$720–740 billion (or up to ~$800 billion+ including leases/finance items and broader groups) for the main hyperscalers.[8]

  • Current 2026 guidance (as of late October 2026 reports): Amazon ~$220 billion (cash capex, unchanged); Alphabet $195–205 billion (unchanged); Meta $135–145 billion (narrowed upward at low end); Microsoft ~$170 billion (calendar year, lowered slightly from prior due to third-party capacity delays).[1]
  • YTD (first three quarters) spend for five companies already exceeded $550 billion in one tally, putting the full-year run rate well above the old consensus.[9]
  • This upward revision reflects repeated guidance increases throughout 2026 as AI infrastructure demand (supply-constrained, not demand-constrained) materialized faster than expected.

No precise 2027 dollar guidance has been issued yet, but companies have signaled significant further increases, with analyst consensus and bank forecasts clustering in the $900 billion–$1.6 trillion range (depending on scope and inclusions like leases or additional players).[10]

  • Alphabet: Spending to “increase significantly” in 2027 (details expected on Q4 call).[3]
  • Amazon: Expects to “spend meaningfully more” in 2027.[1]
  • Meta: Plans to “keep investing aggressively” next year.[3]
  • Microsoft: Capex to accelerate in H1 2027 (fiscal); guided >$50 billion for its Dec. 2026 quarter.[1]
  • External forecasts: Consensus for four ~$900–935 billion; Goldman ~$1.2 trillion (five firms); Morgan Stanley $1.6 trillion (broader hyperscalers/neoclouds); other estimates $1T+.[10][11]

For competitors or new entrants, the sustained hyperscaler capex wave (now structurally outpacing internal cash flow and requiring external financing) creates both opportunities in the supply chain (chips, power, networking, construction) and barriers, as the big four/five lock in capacity and raise the capital intensity bar. The mechanism is data-center/AI infrastructure buildout funded partly by debt/equity/leases, with returns dependent on monetizing AI workloads at high incremental margins. Non-hyperscalers face higher relative financing costs and supply constraints; those supplying the build (e.g., power equipment, memory, networking) benefit directly from the multi-year ramp, while pure demand-side players must demonstrate clear ROI to justify continued spending. Watch Q4 2026 reports and early 2027 calls for firmer 2027 numbers and any signs of moderation.

Sources

Get Custom Research Like This

Start Your Research