Amazon Finally Finds Its AI Footing After Spending a Year Playing Catch-Up to Big Tech

Turns out AmazonAMZN

wasn’t asleep at the wheel — it was loading the next upgrade. The tech heavyweight spent most of 2025 watching cloud rivals race ahead in AI, but that storyline just snapped. Following months of investor frustration over slow guidance and a fuzzy AI roadmap, Amazon finally unveiled a slate of tools that has the competition sweating.

Building momentum: AWS Chief Executive Matt Garman confirmed what investors had long suspected, saying the company “took half a step back” when generative AI first took off. That pause gave rivals time to grab the spotlight while Amazon focused on infrastructure and scalability, a.k.a. the unglamorous groundwork that’s now paying off. The timing couldn’t be better, as its total trailing 12-month revenue is now neck-and-neck with WalmartWMT

, positioning it to become the world’s largest company by annual sales.

• AWS introduced “frontier agents” that can run complex tasks for hours or even days without user intervention, addressing the biggest weakness in today’s AI agent tech.
• It also introduced Nova Forge for training private Nova models on proprietary data and released its new Trainium3 chip, which delivers roughly four times the speed of the previous generation.

Bending the AI Curve

Amazon’s real leverage in the AI race comes from hardware economics. The Trainium3 chip can cut training and operating costs by as much as 50% compared with similar GPUs, a claim AWS says it can back up. However, AWS vice president Ron Diamant noted, “I don’t see us trying to replace Nvidia.” Instead, AWS is leaning into cost and efficiency — and the results are starting to show. The cloud giant’s Q3 revenue jumped 20% year-over-year, its fastest growth since 2022.

• Anthropic is using more than 1M Trainium2 chips to build and run Claude, and AWS’s custom silicon roster now spans Karakuri, Metagenomi, Neto.ai, Ricoh, and Splash Music.
• The retailer’s capital efficiency is hard to match, with its recent $15B debt sale attracting nearly $80B in demand and pricing only 80 basis points above Treasuries.

The long game pays off: Moor Insight and Strategy’s Jason Andersen noted that AWS’s pervasiveness inside the enterprise gives its AI solutions a structural advantage over competitors (WSJ). That incumbency explains why the company can afford to play its own game rather than chase every AI trend. With roughly 29% cloud market share and an annualized run rate near $130B, Amazon is betting companies will stick with AI tools tied to where their data already lives. It doesn’t need to beat every rival — it just needs to make sure its customers never have a reason to leave.