Analysis: IBM and Oracle are tens of billions behind Amazon, Google, and Microsoft in CAPEX, and grow much slower in cloud infrastructure from a much lower base
Charles Fitzgerald / Platformonomics : Tweets: @jonathansorum , @timsneath , @charlesfitz , @jaypipes , @timoreilly , and @marypcbuk Tweets: Jonathan Sorum / @jonathansorum : CAPEX is a leading indicator of 1) building scale that competes 2) customer demand “Amazon, Google, and Microsoft each spent more on CAPEX in 2017 than Oracle has in its entire history.” IBM is even worse off, nowhere near. Little doubt that #IaaS & #cloud is a 3 player race http://twitter.com/... Tim Sneath / @timsneath : Favorite line: “That red line you may mistake for the x-axis is Oracle's CAPEX spending.” Great, snarky analysis by the master of platformonomics. #Cloud http://twitter.com/... Charles Fitzgerald / @charlesfitz : “In the cloud era, IBM's CAPEX is shrinking in both absolute and relative terms.” http://twitter.com/... Jay Pipes / @jaypipes : “Because, unlike marketing, CAPEX doesn't lie.” http://www.platformonomics.com/ ... Tim O'Reilly / @timoreilly : Great analysis of what the numbers show about IBM and Oracle's miserable showing in the cloud market. Excellent analysis! http://twitter.com/... Mary Branscombe / @marypcbuk : Follow the CAPEX: Separating the Clowns from the Clouds - Amazon, Google, and Microsoft each spent more on CAPEX in 2017 than Oracle has in its entire history. http://www.platformonomics.com/ ...
Context & Ripple Effects
Charles Fitzgerald's Platformonomics analysis lands mid-arms-race: RBC had just tallied $63.8B of 2017 capex across 19 cloud operators, up 22% YoY, and the top three had already spent $31.5B combined in 2016. Against that backdrop, the finding that Amazon, Google, and Microsoft each outspent Oracle's entire historical capex in 2017 alone frames IaaS as a three-player contest by sheer capital weight.
The gap did not close afterward — the same analyst's follow-up showed Amazon alone at $54B of capex in 2020, up 69% YoY, while Google spent $22.3B and Microsoft $20.6B (2020 capex tracking), so the 2018 snapshot reads as an early marker of a compounding divide rather than a one-year anomaly.
First-order effects
- IBM and Oracle enter enterprise cloud pitches carrying a visible infrastructure deficit: with capex tens of billions behind and IBM's shrinking in absolute terms, their hybrid and licensed-software narratives must substitute for owned hyperscale capacity.
- For buyers, the practical IaaS shortlist narrows to Amazon, Google, and Microsoft, since capacity commitments at RBC's projected $81B industry run-rate are only credible from balance sheets already spending at that scale.
Second-order effects
- Laggards are pushed toward partnership and acquisition strategies rather than organic data-center builds, because matching the leaders' capex curve would consume cash flows their legacy licensing businesses cannot spare.
- Rising capital intensity raises the floor for every aspiring cloud operator, pressuring smaller providers to specialize or exit commodity compute and pushing pricing power toward the three firms whose depreciation schedules they can absorb.
Third-order effects
- If the pattern holds, IaaS structurally consolidates into a capital-gated oligopoly where annual capex, not software IP, is the entry barrier — a dynamic the later repatriation debate over cloud costs for public software companies shows cuts both ways for customers locked into those platforms.
- New demand layers like edge computing — IDC's $240.6B pace through 2024 (edge spending forecast) — become the plausible reopening through which excluded vendors could re-enter infrastructure without matching hyperscale central-region builds.
The trend: Cloud infrastructure is consolidating around a capital-arms race in which only hyperscale balance sheets can compete, converting annual capex into the industry's primary moat and marginalizing legacy enterprise vendors.