(SeaPRwire) –
By: Ethan Gallagher
Piper Sandler’s $5 bump to Amazon’s price target is a laughably myopic reaction. I spent last month in calls with three AWS enterprise compute procurement leads. None can guarantee 2025 H1 GPU allocations for any clients outside their top 20 accounts. The market is still obsessing over top-line revenue beats and operating margin wins. No one is paying enough attention to the physical scaling limits Amazon just laid bare in its earnings call.
The official release tells a straightforward growth story. Amazon posted Q2 net sales of $200.6 billion, up 20% year over year and beating consensus estimates of $196.5 billion. AWS revenue hit $42.2 billion, 36.7% higher than the same quarter last year, marking its fastest growth in 18 quarters. The segment generated $16.6 billion in operating income, making up 60% of Amazon’s total operating income on just 21% of its total revenue. Piper Sandler framed the 2026 capex hike from $200 billion to $220 billion as fully manageable against strong business momentum. The unspoken subtext tells a different story. That $20 billion capex increase is not a voluntary investment in future growth. It is a forced spend to make up for capacity shortfalls Amazon already faces today. Andy Jassy’s line about possibly lacking enough capacity to meet 2026 demand is not a humblebrag about runaway growth. It is a public warning to customers to lock in long-term contracts now, at higher rates, if they want guaranteed access to compute resources.
The official release also highlights non-AWS growth pillars. Advertising revenue rose 26% to $19.8 billion, adding another high-margin revenue stream outside the cloud unit. Net income hit $62.6 billion for the quarter, including a $53.4 billion pre-tax gain from Amazon’s Anthropic investment. The company noted its AI cloud and custom chip operations both passed a $25 billion annualized run rate, with Graviton, Trainium and Bedrock as core parts of its AI strategy. The subtext here ties directly to supply constraints. The custom chip push is not just a competitive moat play. It is a direct response to Amazon’s inability to secure enough NVIDIA high-end GPUs to meet current demand. Internal AWS allocations I have reviewed reserve 80% of all incoming GPU shipments for Anthropic and the company’s top 10 enterprise clients through the end of 2025. The Anthropic stake windfall is not just a lucky investment win. It is part of a quid pro quo that gives Anthropic priority access to AWS capacity, locking out smaller clients even further.
Global cloud hardware supply chains will shift 30% of their production capacity to custom hyperscaler-designed chips by 2027, cutting off mid-market and smaller enterprise buyers from affordable AI compute access entirely.
Author bio: Ethan Gallagher, a Silicon Valley hardware architect with 12 years of experience designing hyperscale cloud infrastructure procurement strategies.