AWS CEO Matt Garman on why enterprise AI is finally delivering real returns

In a new interview on the Platformer podcast, AWS CEO Matt Garman said the enterprise AI landscape has fundamentally shifted from experimentation to real business value. "I was talking to a room full of CIOs just a couple of months ago,” Garman said.

“I asked, 'How many of you are either seeing materially positive ROI today or have a path in the next couple of months to really high ROI?' 90% of hands went up, which is totally different than a year before."

AWS CEO calls AI inference a new building block that transforms what developers can build Task-accomplishing agents deliver more than just content generation, and enterprises will see massive returns in 2026, says AWS CEO Matt Garman. That demand is what underpins Amazon's $200 billion capital expenditure commitment this year.

Garman described the approach as a portfolio strategy grounded in customer data—not speculation. Garman said the shift is happening far faster than cloud adoption, which still hasn't displaced the vast majority of enterprise workloads after 20 years. AI is moving faster, he said, because cloud infrastructure is already in place.

"If you really like the ROIC [Return on Invested Capital] of a business, you want the ‘C’ to be as high as possible. It's not speculative,” Garman said.

He referenced the durability of Amazon's investments in land and power—assets that retain value even if demand shifts—and the fact that server and chip commitments are only made months out, when customer visibility is high. “We have a lot of mitigations in there and we really think intentionally about how we can reduce risk."

How AWS helps customers maximize AI ROI Garman also outlined how AWS is helping enterprises move from experimentation to production and get the most value from their AI investments. Use the right model for the job, not the most expensive one, was a major point.

One of the biggest drivers of AI cost, Garman said, is companies defaulting to the most powerful model for every task.

AWS tools like Kiro, Amazon's agentic development environment, help by routing tasks to the right model for the job—using lighter models for code generation and reserving higher-reasoning models for more complicated or nuanced requests—so customers can reduce costs without sacrificing output quality.

AWS Summit New York 2026: New ways to make AI agents more effective at work AWS Continuum and AWS Context lead a wave of announcements designed to help customers create momentum with agents applied across their organizations.

"One of the things that's driven up a bunch of cost of AI is that people were trying to use the best model for every single thing,” Garman said. “In Kiro, we do a lot of this for customers where we pick the right model and then we appropriately help customers budget to get the results they want faster, and less expensively.”

Measure outcomes, not token consumption. Garman encouraged leaders to let employees behave like owners over their AI usage—focusing on what AI delivers rather than how much compute it burns. Double down on what works; quickly shut down what doesn't.

The fastest path to ROI, he said, is a disciplined approach: scale the use cases showing results and cut the ones that aren't. For more on the future of AI and jobs, infrastructure investment, and enterprise ROI, watch the full episode on the Platformer podcast. FAQs How many enterprise leaders are seeing positive AI ROI?

How can companies avoid overspending on AI? How much is Amazon investing in AI infrastructure in 2026? Is AI going to eliminate jobs? Are software developers still in demand? Can data centers lower energy costs for consumers?

Trending news and stories 125+ of the Best Prime Day 2026 deals in tech, home, beauty, Amazon devices, and more Best Prime Day 2026 Kindle deals: Save on Colorsoft, Paperwhite, Scribe, and more Alexa+ arrives in Brazil as Amazon's next-generation AI assistant expands internationally Amazon Leo mission updates: 350+ satellites now in orbit after record Ariane 6 launch, ULA Atlas V next