Amazon.com Inc (NASDAQ:AMZN) used its annual re:Invent conference in Las Vegas to outline a series of AI-focused updates, including new Trainium chips, expanded Nvidia integration, autonomous agents, and a new on-premises “AI Factories” model.
The event, which drew roughly 60,000 attendees, offered incremental product progress rather than a headline-making partnership, according to Bank of America analysts.
“The keynote lacked a big new hardware or LLM partnership announcement to materially change sentiment,” the Bank of America analysts wrote.
At the same time, they noted that AWS delivered “lots of AI substance,” pointing to continued improvements across infrastructure and advanced model-building tools.
AWS CEO Matt Garman highlighted a strategy focused on autonomous AI agents, predicting enterprises will eventually create “billions” of them as they become more scalable and independent.
AWS introduced three new frontier agents designed to operate collaboratively: the Kiro Autonomous Agent, which maintains context across sessions; the AWS Security Agent, which scans code for vulnerabilities; and the AWS DevOps Agent, which runs continuously to triage incidents and recommend code improvements.
Bank of America analysts believe that “AI demand is robust,” and that AWS’s focus on agent technologies aligns with broader enterprise adoption trends.
In infrastructure, AWS announced the general availability of the Trainium3 chip, which Amazon says delivers 4.4 times more compute and four times greater energy efficiency compared with the prior generation.
Looking ahead, AWS is developing Trainium4 chips that will use Nvidia’s NVLink Fusion racks to improve system speed and efficiency.
According to the analysts, “the biggest ‘change’ at AWS, in our view, was more collaboration with Nvidia in future Trainium hardware,” with Trainium4 expected to offer greater interoperability with Nvidia systems.
AWS also highlighted EC2 instances powered by Nvidia’s GB300 chips, which provide higher FP4 compute performance than GB200-based instances.
Another major reveal was “AI Factories,” a deployment model that embeds dedicated AWS AI infrastructure, including Nvidia and Trainium hardware, directly into customers’ data centers.
Amazon said these environments operate as private AWS Regions, offering low-latency access to compute, storage, and AI services and targeting organizations with strict regulatory and data-sovereignty requirements.
Bank of America analysts said AI Factories could help AWS expand capacity “while leveraging partners’ ability to provide power.”
Bank of America also noted that discussions with AWS executives and ecosystem partners suggest AI is unlocking new cloud demand and intensifying competition for AI workloads.
Although AWS did not announce a marquee LLM partnership, the analysts wrote that its latest roadmap reflects “growing hardware, model, and agent capabilities” that should support increased adoption. “We expect revenue acceleration,” they wrote, reiterating a ‘Buy’ rating with a price objective of $303.
They forecast AWS growth could move toward 25% in 2026, up from about 20% in the third quarter of 2025, as Amazon adds incremental capacity.
Shares of Amazon traded hands at $232 on Wednesday afternoon.