Amazon CEO Andy Jassy emphasized that the current expenses associated with AI inference are excessively high. To encourage the widespread integration of AI across the less-developed middle market of the various software applications businesses rely on daily, these costs must be brought down to roughly one-tenth of their current levels. Presently, there exists a robust demand for AI computing capabilities. Despite Amazon's intention to make substantial investments in capital expenditures, constructing enough infrastructure to satisfy this demand proves challenging, as a considerable portion of computing power supply has already been reserved in recent times.

Jassy noted that in the past, Amazon sought more cost-effective computing solutions by developing its in-house Graviton processor. A parallel scenario is unfolding in the AI domain, prompting Amazon to unveil its self-designed Trainium chip. This chip boasts a cost-effectiveness advantage of approximately 30% to 40% over comparable processors and is set to be instrumental in driving down AI inference costs. At present, the chip has garnered future revenue pledges surpassing $225 billion.