# [the cost of AI inference compute will reduce by 100x in the next 2 years due to the scaling law](https://twitter.com/tsarnick/status/1802570392571466002)  *** <iframe title="The Futurist Summit: The Age of AI - YouTube" src="https://www.youtube.com/embed/bR9pCJYDdq4?feature=oembed" height="113" width="200" allowfullscreen="" allow="fullscreen" style="aspect-ratio: 1.76991 / 1; width: 100%; height: 100%;"></iframe> #### [[👤Kai-Fu Lee]]: Looking ahead, those of us working on GenAI believe in the scaling law, which means that by simply having more data and more GPU power, GenAI will automatically get smarter with some adjustments. Every year, we can expect GenAI to become significantly smarter than the previous year. For example, the jump from GPT-3.5 to GPT-4 was huge, and the release of GPT-5 is expected to be another big leap. As technology rapidly improves, the costs are also dramatically decreasing. The cost of using the GPT-3.5 and GPT-4 APIs has decreased about tenfold each year. Increased competition and smarter infrastructure technologies are driving these costs even lower. Our recently launched E-Large model performs roughly on par with GPT-4 but at only a quarter of the cost. With these technological advancements and cost reductions due to the scaling law, we anticipate many great applications emerging in the next year, and even more in two years as inference costs continue to drop—ten times in one year and a hundred times in two years. Consider this: if someone needed GPT-4 to build an application last year but found the APIs too expensive, they now have access to those APIs at over ten times lower cost. This makes developing applications much more affordable. We rarely see such rapid technological improvement and cost reduction happening simultaneously and compounding each other. This is what we're looking forward to—an era where AI capabilities grow exponentially while becoming increasingly accessible and affordable. #🤖Kai-Fu_Lee