That is the opinion of Mary Meeker, a former venture investor known as the Queen of the Internet, in a 340-page report called Trends - Artificial Intelligence.
In the report, the phrase unprecedented was repeated up to 51 times, emphasizing the scale and speed of AI-powered transformation. From development, deployment, spending to the use of AI - all are growing at a speed beyond prediction.
Meeker noted that the level of acceptance of AI has surpassed any technology in history. ChatGPT, for example, reached 800 million users in just 17 months, an unprecedented number.
The rate of AI cost reduction is equally impressive. The cost of reasoning that is, using a trained model has fallen as much as 99% over the past two years, which is calculated as the cost of every 1 million token, according to Stanford data.
Meanwhile, the cost of model training could be up to $1 billion. Despite the cost, the speed of performance improvement is still endless. Nvidia's 2024 Blackwell GPU consumes 105,000 times less energy per depreciation than the GPU Kepler in 2014.
Competition between AI rivals is also increasing. Many companies are starting to catch up with each other's features at much lower costs, including Chinese open source models. At the same time, major technology companies such as Google with TPU chips or Amazon with Trainium are developing large-scale AI hardware for cloud infrastructure - these are not sub-projects, but key strategies.
However, AI has not surpassed previous technological revolutions in one aspect: financial profits. Although investors are pouring huge sums of money into this field, AI companies and cloud service providers are also spending huge sums of capital on infrastructure. Consumers and businesses benefit from rapid improvement and price competition, but it is not known which companies will last a long time to become the next generation technology giant.
Only time will tell whether todays AI businesses will become profit machines, Meeker wrote.