4.5 minutes

For companies building their own agents, Amazon introduced Strands, which CEO Andrew Jassy said makes it “much easier to create agents from any foundation model that builders desire.”

Amazon CEO Andy Jassy used the company’s Q3 call to put agentic commerce squarely on the map, arguing that much of AI’s business value “will be in the form of agents” and that Amazon Web Services (AWS) is assembling the tools to take those agents from proofs of concept to production.

“AWS is heavily investing in this area and well positioned to be a leader,” he said.

Jassy framed the effort as a reprise of AWS’s early cloud playbook: modular infrastructure and developer tooling. For companies building their own agents, Amazon introduced Strands, which he said makes it “much easier to create agents from any foundation model that builders desire.”

For running them at scale, AWS launched AgentCore, a runtime and operations layer Jassy described as “a set of infrastructure building blocks that allow builders to deploy secure, scalable agents.”

Early users include Ericsson, Sony and Cohere Health, which deploys agents that cut medical review times “by up to 30% to 40%.” He added that the AgentCore SDK has already been downloaded more than 1 million times.

Amazon ranks No. 1 in Digital Commerce 360’s Top 2000 Database. The database is how Digital Commerce 360 tracks the largest North American online retailers by their annual ecommerce sales.

Amazon is also No. 3 in Digital Commerce 360’s Global Online Marketplaces Database. That database ranks the 100 largest such marketplaces by third-party gross merchandise value (GMV).

How AWS is investing in agentic AI

AWS is seeding its own agents as well. Jassy pointed to Kiro, an “agentic coding IDE” that drew more than 100,000 developers in its first days of preview and has processed trillions of tokens, and Transform, an agent customers used to save 700,000 hours of manual effort this year, including Thomson Reuters’ migration work. For business users, Quick Suite aims to bring “a consumer AI-like experience to work,” with early users seeing more than 80% time savings on complex tasks, according to Jassy.

The push rides on a fast-expanding compute backbone. Jassy said AWS added 3.8 gigawatts of power capacity in the last year and expects to double overall capacity again by 2027.

He highlighted Project Rainier, a U.S. cluster with 500,000 Trainium2 chips, now training Anthropic’s next Claude models and expected to surpass 1 million Trainium2 chips by year-end. Trainium2, he said, is “fully subscribed,” a multibillion-dollar business that grew 150% quarter-over-quarter, with Trainium3 previewing at the end of this year and broader volumes coming in early 2026.

Amazon will continue to buy “very significant amounts” of Nvidia hardware alongside its own silicon, he added.

Where agentic AI meets commerce

Pressed on what agentic commerce means for Amazon’s retail arm, Jassy said the company will pursue a dual path: its own agents plus partnerships with third-party agents.

But he argued that the broader ecosystem isn’t ready.

“We must find a way that makes the customer experience good,” he said. “Right now, there’s no personalization, there’s no shopping history, the delivery estimates are frequently wrong, the prices are often wrong.”

Amazon’s in-house efforts include:

  • Rufus, its AI shopping assistant.
  • Buy for Me, which can purchase items on other merchants’ sites on a customer’s behalf.

Jassy cast agents to make online discovery better than the guided experience of a physical store.

“If you know what you want to buy, there are few experiences better than coming to Amazon. But if you don’t know what you want… you very often want to ask questions and get help narrowing what you’re going to look for,” he said.

Agents, he argued, will close that gap and expand ecommerce overall.

Amazon backed the thesis with early usage data. The company said Rufus had 250 million active customers this year. Monthly users increased 140% and interactions up 210% year over year. Customers using Rufus during a shopping trip were 60% more likely to complete a purchase. Jassy said those results show agent-style experiences can lift conversion even before third-party ecosystems mature.

AWS’ growth in 2025

Beyond agents, the call underscored AWS momentum. Revenue rose to a $132 billion annually run rate, up 20.2% year over year, with gains across training, inference, Bedrock model hosting and SageMaker. Jassy said enterprises are again moving core production workloads to the cloud and that AWS continues to win “the lion’s share” of large migrations.

Chief financial officer Brian Olsavsky said 2025 cash capex is tracking to about $125 billion, to support AI demand, custom silicon, and data center buildouts.

The bottom line from Jassy: the pieces that enterprises say they need — tools to build agents, a secure and scalable runtime and sufficient computing the pieces — AWS is trying to standardize. The near-term partnership model with outside shopping agents is still forming, he cautioned, but the direction is clear.

“I’m very excited about… the prospect of agentic commerce,” he said. “It has a chance to be good for customers [and] really good for ecommerce.”

Sign up

Sign up for a complimentary subscription to Digital Commerce 360 B2B News. It covers technology and business trends in the growing B2B ecommerce industry. Contact Mark Brohan, senior vice president of B2B and Market Research, at mark@digitalcommerce360.com. Follow him on Twitter @markbrohan. Follow us on LinkedInX (formerly Twitter)Facebook and YouTube.

Favorite