Exclusive: Amazon Web Services CFO says AI spending is sustainable

Zoom in: AWS CFO John Felton says the company's AI capital expenditures are paying off, adding that its experience shows the payback of spending aggressively on AI.

• He argued that the sophisticated way to approach AI is by viewing it as a productivity booster more than a cost-saving tool.
• It's about using AI to drive "more interesting customer experiences, more interesting products, more revenue-driving features," Felton says.
• On his team, "AI is an amazing muck reduction tool," he added, saying it allows his professionals to "reinvest their time into more value-added activities."

The big picture: His comments came after Gartner research analysts rattled investors with a report estimating that more than $3.5 trillion in AI infrastructure spending "could be exposed to impairment" by the end of 2030.

• Gartner noted the real risk of revenue growth failing to pay the costs of the AI buildout, leading to potentially massive write-downs.
• Gartner projected that more than 80% of specialized AI cloud providers will cease operating independently by 2030, Axios' Madison Mills noted.

Between the lines: While AWS is exposed to the risks Gartner identifies, it's also among the companies best positioned to withstand them, as a hyperscaler with massive and highly diversified revenue streams.

Felton says he's "comfortable about what we're doing" with capex based on a nearly $500 billion order backlog as of Q2, driven by a mix of "big and small customers, large enterprises, large AI labs, but also small and medium-sized businesses."

• "The demand is real," he says.
• He also says AWS' AI server investments are proving fruitful: "We're seeing paybacks of under three years on those servers."

By the numbers: AWS posted revenue of $42.2 billion in the second quarter, up 36.7% from a year earlier.

• The division's AI revenue run rate was over $25 billion and growing at a triple digit percentage rate year over year.

Friction point: Backlash to data center construction poses a threat to the AI infrastructure buildout of companies like AWS as communities fear the impact on the environment, energy rates and jobs.

• "I've been surprised with some of the anti-data center movement," Felton says. "For me, we're investing in local communities. We are investing to create jobs. I feel like we're doing the right thing."
• He added that it "does confuse me as to where it's coming from. It makes me feel like we need to get better with our messaging, with what we're doing, why we're doing it, how we're doing it," saying AI companies need to "be much more proactive" and "tell that story, which I think is a very positive story."
• He said communities don't need to be concerned about data centers being abandoned and becoming a modern form of industrial blight: "They've got 30-year useful lives. You can put multiple generations of servers and chips into those."

Felton's comments about AI spending come after the business community's approach to LLM usage has evolved rapidly.

• Some companies have paired back on AI token spending, putting limits on users to control their costs.
• But Felton says he studied AWS' own token usage and found that, in general, more is better.
• "What we saw is a very linear line between productivity and token usage," he says. AWS is experiencing "a lot of ROI, a lot of productivity, and we're planning for more."