Seattle, Washington.
Matt Garman helps you order Starbucks, binge Netflix, browse Pinterest, and more to make your online life possible.
As CEO of Amazon’s cloud computing division, he plays a key role in determining how the computing resources needed to power the Internet are deployed, and his responsibilities may soon include shaping the future of AI.
Few people know this business better than Garman, who spoke to CNN at Amazon’s Seattle headquarters in mid-March. As an intern, he wrote a paper that defined Amazon Web Services’ business strategy. He also joined Amazon full-time in 2006 as AWS’s first product manager, helping businesses pivot to the web during the early days of the internet.
In the 20 years since its launch in March 2006, AWS has become an important part of nearly every business that relies on Internet-based tools. When AWS goes down, parts of society come to a halt.
This is big business for Amazon, which had sales of $128.7 billion last year. But AI has transformed the technology industry. Amazon has already made major changes, increasing spending on AI infrastructure to an expected $200 billion this year and cutting tens of thousands of jobs.
Garman explained why Amazon sees this move as necessary. Downsizing has sped up day-to-day operations, he said, and there is enough pent-up demand for AI that Amazon’s cloud tools will continue to be busy for the next five to 10 years, even if technology advances stop.
When AWS was launched, he said, “we needed to explain what the concept of cloud computing was, why we needed it, and why Amazon would be involved.”
That conversation feels nostalgic again.
“Fast forward to where AI is today, and I think it’s actually pretty much the same challenge,” Garman said. “Many people will need to consider new ways of working.”
AWS was started to provide virtual IT infrastructure and servers to businesses.
The pitch was simple: Amazon would handle back-end technology operations so companies could focus on their products and customers. But it was a risky gamble for a company that made a name for itself by restructuring the retail industry.
“We’re very comfortable with being misunderstood,” Amazon founder Jeff Bezos told Bloomberg Businessweek in 2006.
Jeff Barr, chief evangelist for AWS and a 23-year veteran of the company, told CNN that Amazonians are still receptive to the idea.
As I walked through the reception area on my way to meet Garman, I passed a barista serving a special latte celebrating AWS’s 20th anniversary.
We were greeted by a disembodied voice over the intercom asking who we had come to see on the restricted floor where Garman was waiting. As I was led into one of his private conference rooms, I saw physical signs of AWS’s outsized influence. A shelf displaying autographed football helmets. The NFL is one of AWS’s largest customers, proving how important AWS has become to their lives both online and offline.
This is very different from the Seattle pub conversation about Internet storage that led to the birth of AWS. Garman said that when he explained his job to his parents at the time, they had a hard time understanding what cloud computing was.
“It was very difficult to explain to them,” he said. “And my dad was like, ‘You mean like the guy who comes to my office and fixes my printer?'”
But AWS ended up becoming essential for web entrepreneurs. And now, the future of AWS depends on doing the same for AI companies.
Amazon is deeply involved with some of the biggest players in AI, including OpenAI and Anthropic, investing billions of dollars directly to help distribute their services and provide the technology to train their models. It has also developed a custom chip for AI tasks.
But AWS also wants to be an essential part of every enterprise through its Bedrock platform, which Amazon says more than 100,000 companies use to build their own AI apps and agents. Just as AWS’s early products gave businesses access to storage and computing power without huge infrastructure investments, Amazon’s new tools are making AI models more accessible, the company says.
Amazon is the largest cloud provider and is in a strong position to benefit from the surge in AI-related computing demand. But rivals Microsoft and Google are eager to catch up.
According to market research firm Gartner, Amazon’s share of the cloud market will decline from 39% in 2023 to 37.7% in 2024.
Google’s cloud is attractive to startups right now, said Jacob Kolker, managing director of the Seattle-based A12 incubator, which helps AI startups build their businesses. That’s because it’s easy to use and a little easier to get started. Additionally, Google has a more generous credit program for young businesses.
AWS announced that it has provided more than $8 billion in credits to 350,000 new companies through its startup program AWS Activate. It also said that more than 65% of the $1 billion startups reported in October by Pitchbook, a database that tracks startup funding and other financial data, are built on AWS.
And Kolker said the tide could change at any time.
“The pace of innovation in the technology world is obviously breakneck, and I think that applies to many cloud providers as well,” he said.
Amazon’s monolithic data centers and millions of miles of fiber optic cable power the Internet. But it’s the decisions made inside buildings like the glass-fronted Amazon Reinvent tower in downtown Seattle that will help determine Amazon’s future. It will also determine the future of millions of other businesses if things go Amazon’s way.
Inside the walls on a cloudy afternoon in Seattle, Garman and I discuss one of the biggest questions Wall Street has about Amazon: the company expects to spend $200 billion in capital spending on AI infrastructure this year. This number was more than $50 billion higher than analysts expected and the $131.8 billion Amazon would spend on property, plant and equipment in 2025.
Analyst Nicole Green said spending on AI has become so vast that market research firm Gartner classifies companies like Amazon as “digital nations” because they “control as much land, power, water[and]human resources as practically a country.”
Where will that money go?
“This is no secret. It’s a data center and server issue,” Garman said.
Concerns about an AI bubble are growing due to soaring spending.
Tech giants argue that the demand for AI is so hot that they are competing for computing power. Critics want to know when companies will see a return on those investments.
Amazon’s big spending isn’t the only recent move that has raised eyebrows. The company claims it has cut approximately 30,000 people in two rounds of layoffs, in October and January, in order to respond more quickly to advances in AI.
Amazon has previously said that advances in AI won’t drive most of the cuts, but CEO Andy Jassy said in June that the company would need fewer “headcounts” as the AI changes took effect. Garman said AI is playing a larger role in operations such as supply chain planning, data center resource management and, of course, software development.
AI coding tools that allow programmers to build a personal workforce of AI agents are transforming the software industry. According to Garman, projects that used to take two to three years on AWS can now be completed in a matter of months by small teams.
In fact, Garman said AWS teams are currently “building at a rate not seen in many years.”
Amazon’s move may have struck a nerve because it collides with two of the biggest AI-driven concerns to emerge over the past year: the technology’s impact on jobs and whether it’s creating a bubble.
But some experts, like James Landay, co-founder of the Stanford Institute for Human-Centered AI, previously told CNN that the role AI plays in software development has been overstated in some cases. And with each earnings release, analysts seem to wonder when the billions of dollars being poured into AI will be reflected in new products.
Garman is confident the gamble will pay off.
At a recent meeting with about 150 senior technology leaders, he said, 90% of attendees raised their hands when asked if they saw “definitely positive” returns from AI investments or could expect them within the next six months.
“There are certainly signs of an AI bubble,” Garman said. “But I haven’t seen them yet.”
-CNN’s Rhyannon Bartlett-Imadekawa contributed to this report.
