"The debt financing of the build-out is starting to strain at probably its weakest link right now, which is Oracle." Austin on whether credit, not the labs or regulators, is what finally slows the AI build-out "Is this going to naturally slow itself down? Because I've been watching the markets here, and there are a couple of interesting things moving together." "Treasury Secretary Bessent, to his credit, brushed off some of the AI labs' call to slow frontier development, even as those labs are investigating tens of thousands of self-generated security incidents." "Yields are at multi-decade highs. And my question is, is it gonna be credit that becomes the problem here?" "Bessent is saying, hey, if the frontier development is a problem, they can slow their progress anytime they want to. And he urged Congress not to give AI developers a liability shield." "Dario at Anthropic said we must slow the pace at which we improve the capabilities of AI models. And OpenAI and Anthropic are investigating tens of thousands of incidents of models bypassing those rails." "Though notably, both Meta and Microsoft have come out against both the slowdown and some of the proposed fixes here, and there are other AI labs that don't seem to be having these levels of security incidents. Some of the criticism leveled against it is not that there's any problem with AI, it's that you don't know basic InfoSec and sandboxing."

Ram on the two lessons from Meta's AI turnaround: pick the right leader, and skip the Ivy League "It's impressive what Meta has done. The big players have distribution, they have the data, they know how to code. There are a lot of lessons here. One is around leadership." "Alexandr Wang stepped in after Mark put Yann LeCun who won the Turing Prize out to pasture. And then in a few short months, he delivered." "He's effectively paid Alexandr Wang billions of dollars by acquiring half of Scale AI, which Alexandr founded, put in, like, $6.7 billion into Scale AI. So this is a multibillion-dollar acquihire. And he's added, what, $500 billion in market cap to Meta? A couple hundred billion at least. The return on spend has been there." "The second is Alexandr Wang, he dropped out of MIT at 19 and then went to Y Combinator. And he started building. Y Combinator is a more significant institution now than Harvard and MIT and all these Ivy League schools." "And Y Combinator gave him the capital to succeed. I think people can learn a lot from that. I hope my kids take that kind of path too."

Austin says some bond investors expect Washington to fire-hose money at Social Security instead of cutting it, and that fear is showing up in the 10-year "The US is starting to head to a point where we're gonna need to restructure our spending in a meaningful way. And one of the temptations for politicians in those moments is always to just shoot cash out of a fire hose and cause inflation." "If you've looked at the financial repression of the postwar era, right after World War Two, if you've looked at the 1970s, there's definitely a chance that inflation could run significantly hotter while they don't run rates hotter. And I would imagine, knowing a lot of people in the bond market right now, that's driving some of the ten-year sell-off." "And that's because of things like Social Security. Do we really think when the Social Security trust runs out of money, they're gonna have the fortitude to cut payments? Or are they just gonna spend more money and cause inflation?" "We'll find out. I'm not saying I have a strong belief one way or another. But I am saying there are definitely people out there who are like, you know what, they're just gonna shoot money out a fire hose to solve that problem. Like, they'll pay it in nominal terms."