I assume OpenAI and Google haven't shown any new text-to-image models because of the ongoing lawsuits related to the unclear legality of using copyrighted training images. Midjourney was already sued, so they apparently decided to go full steam ahead anyway.