Roth/MKM reiterates Buy on Amazon stock, keeps $285 target By Investing.com

Amazon’s expanded Anthropic partnership adds another $5 billion now and $20 billion later, with Anthropic committing more than $100 billion over 10 years on AWS. Roth/MKM kept a Buy rating and $285 target, while other firms lifted targets to as high as $325 on accelerating AWS demand from AI workloads. The article points to roughly $12 billion to $14 billion of incremental annual recurring revenue from added AWS capacity by year-end 2026, reinforcing a positive outlook for Amazon’s cloud growth.

Analysis

The real implication here is not just incremental AWS share gain; it is that Amazon is effectively converting AI demand into a multi-year capacity pre-sell, which de-risks utilization for the next several capex cycles. That matters because the market still tends to value AWS on current growth, while the partnership increases the visibility of future revenue and makes the Trainium stack more credible as a margin defense versus an Nvidia-dominated buildout.

Second-order winners are Amazon’s custom silicon, networking, and power infrastructure ecosystems. If Anthropic is the template, the bottleneck shifts from model demand to physical deployment speed, which should support suppliers of datacenter power, cooling, and fiber interconnect rather than just semiconductor names. The likely loser is the market’s assumption that cloud AI spend will remain easily portable across providers; once workloads are optimized around a specific silicon/cloud stack, switching costs rise sharply.

The main risk is that consensus may be over-earning the near-term earnings impact while underpricing the capex drag and execution risk. The revenue signal is strong, but the margin path depends on whether AWS can keep supply ramping without compressing returns on invested capital; any delay in 2026 capacity or weaker AI demand from other frontier labs would hit sentiment faster than revenue. Over the next 2-6 weeks, the stock may remain momentum-driven into earnings, but over 6-18 months the debate shifts to whether AWS can monetize AI infrastructure at a better ROIC than peers.

The contrarian angle is that the market may be underestimating how concentrated AI cloud demand is becoming: a small number of buyers can create outsized headline growth, but that also raises customer concentration and negotiation risk later. If AWS is increasingly the venue for the biggest model labs, pricing power is real now, but so is the risk that hyperscaler economics become more utility-like once the buildout normalizes.

Key Decisions for Investors

• Stay long AMZN into earnings with a 2-6 week horizon; use call spreads or a tight risk budget rather than outright stock if the position is already crowded, because the setup is good but the post-earnings reaction could compress on capex concerns.
• Pair trade: long AMZN vs short a basket of cloud names with less visible AI workload capture over the next 1-3 months; the thesis is that visible pre-committed AI revenue should support relative multiple expansion at AWS versus less differentiated platforms.
• Add exposure to datacenter infrastructure beneficiaries on a 3-12 month view, especially power/cooling/networking supply chain names; the trade is on physical capacity bottlenecks, not just model training demand.
• Consider a tactical hedge by selling short-dated upside against existing AMZN longs into earnings if the stock has already run sharply; upside surprise is less interesting than guidance on 2026 capacity and ROIC.