The cloud division head of Amazon.com Inc. has observed that clients are beginning to transition from utilizing the company's services for training artificial intelligence models to embedding them into their operations, which is driving a surge in demand for a practice known as inference.

"We still see large training clusters being used by a number of companies, but as these models get really popular and really powerful, more and more companies are integrating that inference into their workloads," said Matt Garman, Chief Executive Officer of Amazon Web Services, in a Monday interview.

Training involves the computing-heavy process of building an AI model, while inference refers to the application of that model to address user queries.

Amazon, the foremost vendor of rented computing power and data services, announced last week that it plans to invest $220 billion in capital expenditures by 2026, an increase from its earlier projection of $200 billion. This higher spending reflects the growing cost of memory and other components essential for data centers.

Garman declined to provide a forecast for the company's spending next year but described the prospective artificial intelligence business for the firm as "just massive."

He noted that Amazon is confident in its demand projections and is securing customers with a focus on long-term agreements, which assist AWS in anticipating utilization for new data centers.

"As we're investing, we're getting five-year commitments from customers," Garman told Bloomberg Television. "Today, demand still significantly outstrips supply and we're trying to build and invest to keep up with what customers are asking for."

The company's shares have surged in recent days, pushing Amazon's market capitalization beyond $3 trillion, following last week's report that its cloud-computing revenue accelerated for the fifth consecutive quarter.