The floors of the trade shows are exactly the same as they were three years ago. The same sponsor coffee shops, the same blue lanyards, the same engineers gazing at laptops in between meetings. What is discussed in the side chats has changed. The question of whether to shift workloads to the cloud is no longer up for debate. They are arguing over which of three businesses will ultimately control their AI roadmap for the ensuing ten years.
The figures are not nuanced. According to Gartner, cloud infrastructure spending increased by 22.5% in 2024 to almost $172 billion worldwide, with the top three providers—AWS, Microsoft Azure, and Google Cloud—capturing about 71% of it. Although that share has been gradually increasing for years, the most recent quarter is what altered the discourse. Google Cloud broke $20 billion for the first time, growing 63% annually. 40% was printed in constant currency by Azure. AWS reached its fastest rate in fifteen quarters, reaccelerating to 28%. Three businesses traveling in the same direction at three different speeds.
It was almost like a stress test to watch the latest earnings cycle. Within days of one another, four hyperscalers reported. Each of the four beats. Each of the four pledged enormous capital expenditures. And the market made ruthless decisions rather than rewarding everyone equally. Alphabet saw a 10% increase. Amazon hardly moved at all. Microsoft saw a nearly 4% decline. Meta suffered an 8.6% loss. It seems that investors are no longer debating whether overall AI spending is excessively aggressive. The balance sheets that have something to show for it are being sorted.
Backlog is what separates the winners. In just one quarter, Google’s contracted cloud backlog almost doubled, reaching $460 billion. That isn’t a prediction. That is customer money that has been signed and is awaiting revenue recognition. The stock hardly flinched when Alphabet raised its 2026 capex guidance to $180–$190 billion and hinted that 2027 would be even higher. The receipts were on the table already. Microsoft’s narrative was similar, but it had a twist: management cautioned that capacity would be “constrained” until 2026. This was initially a sign of bullish demand and is now subtly beginning to resemble an early warning about the timing of depreciation.
Amazon’s predicament is unique. Prior to this most recent reacceleration, AWS’s growth had been more consistent and structurally lower, but it is still by far the biggest of the three. Investors are put off by the capex bill. The company’s first-quarter earnings of about $43 billion put it on track for a $200 billion full-year run. Andy Jassy has been stressing that a significant portion of that expenditure is already covered by customer commitments. The size of AWS’s installed base means it doesn’t need to win every new workload to maintain its dominance, though it’s still unclear if the market fully believes him. All it has to do is keep them.
From a distance, the most peculiar aspect of all this is how much of the current infrastructure depends on three campuses, three executive teams, and three sets of capacity decisions made in Redmond, Mountain View, and northern Virginia. Cloud computing was a side project that most CIOs viewed with suspicion fifteen years ago. As Corey Quinn frequently notes, it is currently a top-three line item on the majority of P&Ls, yet when budgets get tighter, nobody seems to discuss reducing it. The dependency is no longer noticeable. That typically indicates the depth of something. The question is whether the same three names will be rewarded in the subsequent leg of the cycle. It continues to do just that so far.
