RBC: 19 cloud service operators spent $63.8B on capital expenditure in 2017, up 22% YoY, and will spend $81B in 2018, a rise of 27%
Dan Gallagher / Wall Street Journal : Tweets: @djtgallagher Tweets: Dan Gallagher / @djtgallagher : Capex by cloud giants accelerated in 2017 and looks likely to rise even more this year. Combined spend by just Amazon, Google and Microsoft up 33%. Still little incentive to slow things down: http://www.wsj.com/...
Context & Ripple Effects
RBC's tally extends an arms race the Journal had already charted: its April 2017 analysis found Amazon, Alphabet and Microsoft spent a combined $31.5B on capital expenses and leases in 2016, up 22% from 2015. The new numbers show the same cohort accelerating rather than pausing — the big three alone grew combined capex 33% in 2017, with RBC projecting $81B across 19 operators for 2018.
The significance is that this spending is not cyclical experimentation but the cost of staying in a market where scale decides share. Subsequent coverage confirmed the trajectory: Amazon's capex hit $54B in 2020 alone after growing 69% YoY, and Synergy's data shows the top three's share of global cloud spending climbing from 52% in 2018 to 65% by Q1 2022.
First-order effects
- Amazon, Google and Microsoft are committing to another year of double-digit capex growth with 'little incentive to slow down,' per Gallagher — meaning larger data center footprints and bigger server orders land on their suppliers' books through 2018.
Second-order effects
- Rivals outside the top tier must match capital intensity they cannot easily fund, reinforcing the consolidation visible in Synergy's later figures, where the big three's share of quarterly cloud spend kept rising toward two-thirds.
Third-order effects
- If the pattern holds, cloud computing structurally resembles an infrastructure utility: only operators able to sustain tens of billions in annual capex compete at scale, and each cycle (2016's $31.5B, 2020's $54B for Amazon alone) resets the entry price higher.
The trend: Cloud computing is consolidating around a handful of hyperscalers whose escalating capital expenditure functions as both the price of admission and the mechanism of market-share capture.