Amazon Q3 Results: CEO Jassy Will Keep Spending To Meet AI Demand

‘It’s still quite early and represents an unusual opportunity for customers in AWS,’ says Andy Jassy, CEO of Amazon Web Services parent Amazon.com.

Even with Amazon Web Services parent Amazon.com adding more than 3.8 gigawatts of capacity in the past 12 months to try to meet growing artificial intelligence demand and plans to add at least another gigawatt in the fourth quarter, the vendor’s executives say to expect continued infrastructure spending, with capacity set to double by the end of 2027.

The Seattle-based vendor’s current power capacity doubles from 2022, with capacity comprising power, data centers and chips–primarily Amazon’s own Trainium chips and ones by Nvidia, Amazon.com CEO Andy Jassy told analysts Thursday on the vendor’s latest quarterly earnings call.

“You’re going to see us continue to be very aggressive investing in capacity because we see the demand,” the CEO said. “It’s still quite early and represents an unusual opportunity for customers in AWS.”

[RELATED: Amazon Confirms 14,000 Layoffs; Says ‘AI’ Innovation Reason For ‘Reducing Roles’]

Amazon Q3 Earnings

Amazon reported results for its third fiscal quarter, which ended Sept. 30.

For any analysts concerned about Amazon spending too much on infrastructure, Jassy said that “as fast as we’re bringing it in, right now, we are monetizing it.”

Spending on property and equipment hit Amazon’s free cash flow, which fell to $14.8 billion, about a third the amount reported for the same period a year prior due to a year-over-year increase of $50.9 billion. Amazon.com reported cash flow of $130.7 billion for the trailing 12 months (TTM), up 16 percent year on year.

The vendor’s capital expenditure (CapEX) in the quarter was $34.2 billion. It has spent $89.9 billion so far this year, most of that going toward AWS for AI and core services and Trainium and custom chips.

Amazon.com expects its full-year CapEx to reach about $125 billion in 2025. The amount should increase in 2026, Amazon.com Chief Financial Officer Brian Olsavsky told analysts on the call.

Amazon Layoffs

Amazon.com revealed that planned role eliminations cost the company $1.8 billion in operating income in the third quarter. Operating income was flat year on year, coming in at $17.4 billion.

The company confirmed Tuesday the layoff of 14,000 corporate employees as part of major organizational changes across the tech giant.

Users of the LinkedIn social network who said they are part of the new wave of layoffs range from a senior internal technical sourcing recruiter with AWS about four years, an AWS technical support engineer with the company about four years and a senior analytics technical writer also with AWS about four years, according to a CRN review of LinkedIn posts.

AWS’ operating income was $11.4 billion in the quarter, up about 10 percent year on year. The parent company said this operating income was hit by estimated severance costs primarily related to planned role eliminations, but Amazon.com didn’t break out the operating income discounting those costs.

The severance charges were recorded primarily in technology and infrastructure, sales and marketing and the general and administrative expense line items, Amazon.com‘s CFO said.

Asked about the layoffs on Thursday’s call, Jassy said the layoffs were “not really financially driven, and it’s not even really AI driven–not right now, at least–really, it’s culture.”

Amazon’s growth in business size, locations and the number of businesses it is in resulted in a lot more employees and a lot more layers that could have held the vendor back from innovating in the AI era, he said.

“When that happens, sometimes, without realizing it, you can weaken the ownership of the people that you have who are doing the actual work and who own most of the two-way door decisions–the ones that should be made quickly, right at the front line–and it can lead to slowing you down,” Jassy said. “As a leadership team, we are committed to operating like the world’s largest startup, and that means removing layers. That means increasing the amount of ownership that people have, and it means inventing and moving quickly.”

Agentic AI In E-Commerce

Amazon.com leveraging AI agents for its commerce business might resonate with solution providers already leveraging the emerging technology with customers in this segment–or serve as a sign of things to come for partners earlier in the AI journey with commerce customers.

Agentic e-commerce can help customers narrow down what they want to buy, with physical stores still presenting advantages compared to online stores for spontaneous purchases.

Amazon.com and its mobile app offers the AI-powered shopping assistant Rufus, which saw 250 million active customers this year, with monthly users more than double year on year. The vendor’s AI-powered Amazon Lens capability that allows shopping through images, screenshots and barcodes has tens of millions of users each month.

The vendor has rolled out an agentic “buy for me” feature for discovering and buying select products from other brands’ sites if the product isn’t sold in Amazon’s store.

The vendor expects more third-party AI agent partnerships in the future, but the customer experience with AI agents has to get better, the CEO said.

“There’s no personalization,” he said. “There’s no shopping history. The delivery estimates are frequently wrong. The prices are often wrong. So we’ve got to find a way to make the customer experience better and have the right exchange of value.”

Jassy also said that building AI agents for any purpose today “is still harder than it should be,” with a lack of building blocks to build secure, scalable agents. He pointed to his company’s Strands Agents open-source software development kit (SDK) for building agents from any model and Bedrock AgentCore platform as ways to improve the agent building process.

Trainium Demand, Nvidia Relationship

Amazon’s multibillion-dollar Trainium2 custom AI chip business more than doubled quarter on quarter, according to the vendor. The chip has attracted attention as a way for the cloud vendor to potentially avoid over-reliance on semiconductor vendors like Nvidia.

Amazon-backed AI upstart Anthropic is training the next version of its Claude chatbot on top of the 500,000 Trainium2 chip cluster called Project Rainier. Anthropic expects to go to 1 million Trainium2 chips by the end of the year, Jassy said.

Amazon has a small number of very large customers on Trainium2, with Jassy saying that the chip offers up to 40 percent more price performance than other options.

“Customers, as they start to contemplate (a) broader scale of their production workloads moving to being AI focused and using inference, they badly care about price performance,” Jassy said. “We have a lot of demand for Trainium. … Project Rainier is something that is specific for Anthropic but we have a lot of other customers who are interested in employing large clusters of Trainium chips that we’ll hopefully give them a chance to do so with Trainium3.”

Speaking of Anthropic, Amazon’s investment in the upstart brought a pre-tax net income gain of $9.5 billion, contributing to third-quarter net income of $21.2 billion.

Amazon cloud rival Microsoft, meanwhile, took a $3.1 billion hit to its net income in its latest quarter from its investment in Anthropic rival and ChatGPT maker OpenAI, the vendor reported Wednesday.

Amazon expects Trainium3 to enter preview at the end of this year with fuller volumes coming in at the beginning of 2026. The vendor has seen large and medium size customers interested in Trainium3. Jassy wants this next iteration of the chip to show another 40 percent better price performance than Trainium2. The company is at work improving the software ecosystem as well.

Jassy said he always wants customers to have multiple chip options, which AWS has strived for with all of its components over the course of its history. Amazon built its Graviton central processing unit (CPU) for better price performance compared to other x86 processor options, as an example.

Asked about Amazon’s relationship with Nvidia, Jassy said that his company has “a very deep relationship with” the chipmaker.

“We have for a very long time and we will for as long as I can foresee,” he said. “We buy a lot of Nvidia. We are not constrained in any way in buying Nvidia. And I expect that we’ll continue to buy more Nvidia, both next year and in the future.”

Amazon also buys chips from the likes of Advanced Micro Devices (AMD) and Intel, Jassy said.

Backlog Battles

Oracle stunned analysts earlier this year with a $500 billion backlog of AI contracts, and its competitors in the AI and cloud space have also reported eye-popping pipeline.

Amazon added to the list of astronomical figures with a $200 billion backlog reached by quarter’s end. Google reported in its quarterly earnings call Wednesday a backlog of $155 billion while Microsoft’s commercial remaining performance obligation (RPO) business backlog reached $392 billion, up 51 percent year on year.

Amazon said its backlog didn’t include several unannounced new deals in October, which together are more than its total deal volume for all of the third quarter.

Amazon Product Milestones, Q3 In Depth

Amazon shared a variety of product usage and business milestones that might interest its solution providers.

Among the updates:

More details from the quarter include AWS sales increasing 20 percent year on year to $33 billion. The business has a $132 billion ARR.

That growth rate was the highest in 11 quarters, marking growth not seen since 2022, and an acceleration of 270 basis points over last quarter.

“Percentage growth is a relative term,” he said. “It’s very different having 20 percent year-over-year growth on a $132 billion annualized run rate than to have a higher percentage growth rate on a meaningfully smaller annual revenue, which is the case with our competitors.”