AWS CEO: Billions of AI Agents Will Power Enterprise Future

At this year’s AWS re:Invent conference in Las Vegas, Amazon Web Services (AWS) CEO Matt Garman laid out a vision for a transformative enterprise AI future.

This is centred on “billions of agents” capable of operating autonomously across complex business processes.

Speaking to a packed audience of industry leaders on 3 December, Matt framed this evolution as a continuation of AWS’s long-standing mantra: providing customers the freedom to invent.

“Frontier agents are a step function change, more capable than what we have today,” he shared. “One, they’re autonomous. You direct them towards a goal and they figure out how to achieve it. Two, they have to be massively scalable.”

Matt added: “Three, these agents need to be long running. They may be working for hours, maybe even days in pursuit of ambitious, sometimes frankly amorphous goals without requiring human intervention or direction.”

He emphasised their capabilities stem from an enormous amount of software engineering, infrastructure data and a robust memory architecture that allows agents to handle broad goals over extended periods.

Unlike early AI models that often stalled or repeatedly requested guidance, these agents integrate deeply with business data, core applications and enterprise systems, which Matt says positions AWS back in the middle of enterprise AI.

Matt announced the launch of Nova Forge, a service allowing companies to train private instances of AWS’s Nova models on their own data.

He said: “This allows you to produce a model that deeply understands your information without forgetting the core information that the thing has been trained on.”

Accelerated model training and enterprise modernisation

The CEO also highlighted that with services like SageMaker Hyper Pod, a purpose-built infrastructure for training large machine-learnings models at scale, and with high-performance P5 and P6 GPU instances, customers have been able “to do runs at a third of the time, going from six weeks down to two weeks,” while training pipelines “have become 90% more reliable”.

Matt also emphasised the financial benefits of modernisation tools like AWS Transform, which helps customers move off legacy systems faster.

He shared examples of clients like Thompson Reuters, who are modernising “over 1.5 million lines of code per month”, demonstrating how AWS solutions drive measurable cost and time savings.

Additionally, initiatives such as database savings plans, which “can save you up to 35% across all your usage for our database services”, further underscore AWS’s focus on helping customers reduce costs while expanding their use of the platform.

Massive infrastructure investment

The company announced a US$50bn expansion at the end of November to deploy nearly 1.3GW of AI and HPC capacity across AWS Top Secret, AWS Secret and GovCloud (US) regions.

Matt said in a statement at the time: “Our investment in purpose-built government AI and cloud infrastructure will fundamentally transform how federal agencies leverage computing.”

This effort aims to accelerate critical missions, from cybersecurity to drug discovery, by combining AI, modelling and simulation in secure, high-capacity environments.

Matt framed these initiatives as part of a broader enterprise leadership strategy: building scalable infrastructure, addressing technical debt and creating AI capabilities that integrate seamlessly with business processes.

In AWS’s view, the combination of agentic AI, proprietary data integration and massive computing investments represents the blueprint for enterprises and governments alike to extract real-world values from AI at scale.