Amazon CEO Andy Jassy Calls for Heavy Investment in AI by Companies

Amazon CEO Andy Jassy is urging companies to dive headfirst into AI investments if they want to see big financial payoffs down the road. In his annual letter to shareholders, released on Thursday, Jassy emphasized that keeping up with the whirlwind pace of AI innovation and meeting customer demand for AI products requires pouring in "substantial capital." He made it clear that Amazon itself must make these hefty investments now to ensure lucrative returns in the future.

Jassy's push for aggressive AI investment follows Amazon's announcement during its February fourth-quarter earnings call, where they revealed plans to shell out over $100 billion on capital expenditures in 2025. A "vast majority" of this hefty sum, Jassy noted at the time, will be funneled into enhancing AWS's AI capabilities.

"We continue to believe AI is a once-in-a-lifetime reinvention of everything we know," Jassy wrote passionately in his shareholder letter. "The demand is unlike anything we’ve seen before, and our customers, shareholders, and business will be well-served by our investing aggressively now."

Jassy highlighted that the biggest chunks of AI spending are currently on data centers and chips. However, he reassured investors that over time, these costs will decrease. "In AWS, the faster demand grows, the more data centers, chips, and hardware we need to procure (and AI chips are much more expensive than CPU chips)," Jassy explained. "We spend this capital upfront, even though these assets are useful for many years."

To illustrate his point about declining AI infrastructure costs, Jassy mentioned Amazon's own Trainium2 chips, released in late 2024, which provide a 30%-40% better price-performance ratio compared to current GPU-powered computing instances.

Looking ahead, Jassy also predicted a shift in AI price dynamics. He believes that as training costs for AI decrease, more money will be directed towards inference—the actual serving of AI models. "We feel strong urgency to make inference less expensive for customers," Jassy noted. "More price-performant chips will help. But, inference will also get meaningfully more efficient in the next couple of years with improvements in model distillation, prompt caching, computing infrastructure, and model architectures."

Currently, Amazon is in the process of developing over 1,000 generative AI applications, according to Jassy. He boasted that Amazon's AI revenue is growing at "triple-digit" year-over-year percentages, representing a "multi-billion-dollar annual revenue run rate."

Amazon declined to provide further comments on the matter.