If you've read enough lesswrong, this won't be new to you. The strategy always was "become the most capable model lab so that you have leverage to control AI development."