AI cycle is at ‘one of the most expensive’ points: B capital co-founder

And let's bring in Anthropic in this near trillion dollar IPO preo valuation. Open AAI faces a staggering $600 billion capex problem. What are valuations telling us when you look at the balance of risks associated with these frontier LLM developers? >> It's really two very different things that are driving this. One is actual usage. I mean these are these models are fundamental. People are using them to drive real impact. Enterprises have been a little bit behind, not because of the core functionality of the models, but the ability to really mold those models and make them work for enterprises. That's coming. You're seeing that happening more and more. On the other side, they have to raise these large amounts of compute, you know, 3/4 of a trillion dollars this year amongst the hyperscalers to purchase compute. That'll come down in 2027. You'll see the inference costs decrease. You'll see the cost of compute decrease. You'll see the size of data centers decrease. Um we're we're a bit in the peak in the cycle there, but in the next couple of years, the cost will come down. The AI will continue to get better. Um, we uh we just need to make it through this this point in the cycle, which is probably one of the most expensive points in this AI cycle.