Amazon’s CEO, Andy Jassy, advocates for substantial investment in AI, suggesting that companies should “aggressively” allocate resources toward AI development to achieve greater financial benefits in the future. In his annual shareholder letter, Jassy highlighted the need for “substantial capital” to meet the rapid advancements in AI and rising customer demand. He emphasized that Amazon must make these expenditures now to realize strong returns on its investments in the long run.
Jassy’s remarks follow Amazon’s announcement in February during its fourth-quarter earnings call that it plans to allocate over $100 billion toward capital expenditures in 2025. The majority of this investment will focus on enhancing AWS AI capabilities, as stated by Jassy at that time.
He described AI as a once-in-a-lifetime transformative innovation, noting the unprecedented demand. Jassy believes that the interests of customers, shareholders, and the business will be well-served by proactive investment in AI.
Currently, the largest AI-related expenses are in data centers and chips; however, Jassy noted that infrastructure costs will decrease over time. He cited the need to procure more data centers, chips, and hardware as AWS demand grows, pointing out that AI chips are significantly more expensive than CPU chips. Despite the upfront capital required, these assets are expected to be beneficial for many years.
Jassy referred to Amazon’s own Trainium2 chips as an illustration of how AI infrastructure costs may decrease. These chips reportedly offer a 30%-40% improvement in price-performance compared to existing GPU-powered computing instances. Trainium2 became available in late 2024.
Additionally, Jassy mentioned that AI pricing dynamics will shift in the future, as training costs for AI models decrease and spending transitions toward inference, or the application of AI models. He expressed a strong urgency to make inference more affordable for customers, highlighting that more cost-effective chips and advancements in model distillation, prompt caching, computing infrastructure, and model architecture will contribute to greater efficiency in the coming years.
According to Jassy’s shareholder letter, Amazon is developing over 1,000 generative AI applications. He also noted that Amazon’s AI revenue is experiencing “triple-digit” year-over-year growth and represents a “multi-billion-dollar annual revenue run rate.”
TechCrunch has reached out to Amazon for further comments.
