Amazon dropped its Q1 earnings yesterday, and the headline numbers are predictably huge. AWS brought in $28.4 billion in revenue, beating analyst estimates by a comfortable margin. That’s a 17% year-over-year jump, and it signals that the cloud slowdown everyone worried about in 2023 is firmly in the rearview mirror.
But here’s the thing that caught my attention: capital expenditure hit $18.5 billion in the quarter, and Amazon expects that number to keep climbing through 2026. Andy Jassy was blunt about it on the earnings call — this spending is mostly going to AI infrastructure, and they’re not about to tap the brakes.
I’ve been watching these hyperscaler earnings cycles for years, and the pattern is becoming almost comical. Every quarter, Wall Street analysts ask when the AI capex spigot will get turned down. Every quarter, the answer is “not yet.” Jassy specifically called out that AWS’s AI business is growing at a triple-digit percentage rate, which is a polite way of saying “we’re not stupid enough to stop investing in the thing that’s working.”
The real story here isn’t just the revenue beat — it’s the margin story. AWS operating margins came in at 37.6%, up from 35.5% a year ago. That’s impressive given the massive infrastructure buildout. Amazon is somehow getting more efficient while spending more. That doesn’t happen by accident.
What’s interesting to me is how Amazon is positioning itself differently from Microsoft and Google in the AI cloud race. Jassy spent a lot of time talking about custom silicon — their Trainium and Inferentia chips. He’s essentially saying “we don’t need to buy all our AI hardware from Nvidia.” That’s a bet on vertical integration that could pay off huge if it works, or leave them with stranded assets if the AI landscape shifts.
The elephant in the room is whether this spending pace is sustainable. Amazon’s total capex for 2026 is on track to hit $75-80 billion, most of it going to AWS data centers and networking gear. That’s more than what most countries spend on their entire defense budgets. If AI demand softens — and I’ve seen enough hype cycles to know that’s always a possibility — Amazon will have a lot of expensive servers sitting around.
But right now, the demand is real. AWS signed several massive new deals this quarter, including a multi-year agreement with an unnamed Fortune 50 company that Jassy described as “one of the largest in AWS history.” Enterprise customers who were cautious about AI in 2024 are now going all-in.
The bottom line: Amazon is making more money from AWS than expected, but they’re plowing every dollar — and then some — back into infrastructure. For now, that bet is paying off. But I’ll be watching the utilization rates on all those new data centers closely. If they start to dip, that’s when the tone will change.
Comments (0)
Login Log in to comment.
Be the first to comment!