“Our customers, the company and our shareholders will feel good in the long run,” said Amazon CEO Andy Jassy.

Amazon CEO Andy Jassy said during the company’s latest quarterly earnings call that the provider plans to spend more than $75 billion in capital in 2025, with most of it going toward services. Amazon Web and generative artificial intelligence.

The Seattle-based cloud giant has spent $51.9 billion in capital expenditures since the start of the year and expects about $75 billion in 2024, focusing on infrastructure, AWS and GenAI. But the CEO highlighted the long life cycle of this infrastructure – he estimates the lifespan of data centers at 20 to 30 years – and said Amazon’s history of spending should be trusted responsible.

“We have proven over time that we can generate enough operating profit and free cash flow to make this business a very successful return on invested capital, and we expect the same to happen here with the Generative AI,” he said. “This is a truly exceptionally great, perhaps once-in-a-lifetime opportunity. And I think our customers, the company and our shareholders will feel good about this long-term strategy that we are aggressively pursuing.

(RELATED: Presidio signs ‘huge’ AWS deal to boost GenAI, sales and innovation)

AWS Q3 Results

Jassy shared his views on his company’s spending and the AI ​​opportunity during Amazon’s call Thursday to report results for its fiscal third quarter, which ended September 30.

During the call, Jassy highlighted AWS’s “deepest partner community” as a competitive differentiator, along with expanded functionality and strong security. He also detailed the supplier’s continued investment in custom chips to help meet AI demand and achieve better AI pricing for customers scaling these workloads.

Growing activity in AI

Over the past 18 months, AWS has released approximately “twice as many machine learning and GenAI features as other major cloud providers combined,” Jassy said on the call.

“AWS’s AI business is a multibillion-dollar revenue business that continues to grow at triple-digit percentages, year over year, and is growing more than three times faster at this point in its evolution, while AWS itself was growing,” he says. “We felt like AWS grew pretty quickly. »

Jassy told listeners on the call that Amazon Q continues to see “strong adoption,” with the AI ​​chatbot saving Amazon teams $260 million in annual savings, 4,500 years of development work for more than 1,000 developers and the migration of more than 30,000 applications to new applications. versions of the Java Development Kit (JDK).

“Expect more practical AI game changes from Q,” Jassy said.

Jassy also said Amazon is revamping its Alexa virtual assistants with GenAI with next-generation foundational models, seeking to go beyond today’s GenAI applications focused on “costless productivity,” customer experience and “taking into account large bodies of data and the ability to summarize”. and grouping and answering questions” – but not yet “really good” at taking action for customers.

“The next generation of these assistants and generative AI applications will be more efficient not only at answering questions and summarizing, indexing and aggregating data, but also at taking action,” Jassy said. “And you can imagine we’re pretty good at this with Alexa.”

AI chips, infrastructure

Jassy said Amazon’s cloud business, like other providers, has been hit by “lower capacity than they asked for,” primarily with chips.

Amazon’s Trainium2 machine learning (ML) training chip “starts to ramp up in the coming weeks and will be very compelling for customers and value for money,” Jasy said. “We are seeing considerable interest in these chips and have gone back to our manufacturing partners several times to produce much more than we initially planned.”

Growing demand for AI has led to increased spending on data centers, networking equipment and chips that are more expensive than central processing unit (CPU) hardware, he said. “We invest in all of this up front before we can monetize it to customers using the resources.”

Jassy said its AI business is “more fluid and dynamic than our non-AI part of AWS,” but that the vendor has “very strong demand signals” to avoid too much or too little investment in capacity and hub. of data.

“People don’t show up for 30,000 tokens a day,” he said. “As the market matures over time, there will be very healthy margins here in the generative AI space.”

Jassy also touted Amazon’s partnership with chipmaker Nvidia, saying that “we tend to be their primary partner for most of their new chips.”

Q3 in detail

AWS sales reached $27.5 billion, up 19% year-over-year. The cloud giant has an annualized run rate of $110 billion.

Operating profit was $10.4 billion, up $3.5 billion year-over-year. Amazon Chief Financial Officer Brian Olsavsky said the supplier continues to “focus on cost control, including a measured pace of hiring,” and to “focus on the efficiency of our infrastructure and the cost reduction across the business.

Office staffing is “down slightly year-over-year, and it remained stable until the end of last year,” he said.

Amazon’s stock rose about 5% after the market closed Thursday, to about $196 per share.