At the World Economic Forum in Davos, Amazon CEO Andy Jassy warned that U.S. import duties are beginning to show up in consumer prices as third-party vendors deplete inventory stockpiled before the Trump administration’s levies. Jassy noted that sellers face a roughly 10% increase in input costs and have three primary strategies: absorb the extra expense to maintain market share, pass it on to customers in the form of higher sticker prices, or pursue a hybrid approach. He emphasized that retail’s typical operating margin sits in the mid-single digits, leaving limited scope to absorb substantial cost hikes. As a result, some categories—particularly electronics and small household appliances—have registered price increases of 3% to 5% in recent weeks according to internal Amazon marketplace data, a marked shift from last year when pre-tariff build-outs largely buffered end users from any noticeable impact.

In a move to defend its position in conversational AI, Amazon has begun rolling out a web-based version of its Alexa+ assistant at alexa.com to a select cohort of early-access customers. The interface blends traditional smart-home controls with a chat-driven layout, offering users tools to plan events, conduct research, shop across Amazon’s catalog, and troubleshoot devices. This launch follows a year that saw Amazon commit $8 billion to its partnership with Anthropic and boost its 2025 capital expenditure guidance to $125 billion, with a significant share earmarked for AI infrastructure. Those outlays complement a trailing-12-month net income of $76.5 billion, giving Amazon ample financial firepower to refine Alexa+ and expand its agent-style commerce capabilities beyond the 200 million active Echo devices in global markets.

Speaking to Squawk Box last week, Jassy described current AI workloads as “unprecedented,” citing exponential growth in demand for high-performance compute and power capacity within Amazon Web Services data centers. He disclosed that AWS has added more than 30% additional GPU rack space year-over-year and is on track to double its AI training cluster footprint by mid-2026. This expansion is driven by enterprise customers running large-scale generative AI experiments, which now account for over 20% of AWS’s overall consumed machine-learning compute hours. To support this, Amazon is deploying its next-generation Graviton3E processors alongside third-party accelerators, ensuring availability and performance SLAs for mission-critical AI applications.