On July 16, Coinbase CEO Brian Armstrong, while participating in the podcast show "Nikhil Kamath," stated that the gap between open-source models and cutting-edge models may be only about six months, yet the inference costs could be reduced by over 99%. In the future, a significant amount of workloads may shift towards more cost-effective models. Ultimately, the entities that may truly capture most of the value could be not just large model companies but also providers of infrastructure such as chips and energy. He compared the current AI craze to the crypto, internet, and even railway eras: the valuation may experience a bubble and correction initially, but as long as there is real underlying value, the industry will still experience growth after the adjustment.
