The most important thing that fund managers get wrong is not having appropriate reserves for your winners. So take 20 billion off the table of that 20 billion, I would probably take two thirds of it. And I would go to all the hyperscalers and anybody that's providing cloud compute and essentially buy out all of the compute credits on GPUs so that I could tell any startup in the world, you will do a safe with me, I will give you free compute on pick your cloud provider, pick your model, I don't care.

In return, we get 7% up front and you have to tell us some benchmark of how this model is improving. Now you've scoped the problem to 20 billion is in reserves, 15 billion of it is tied up in credits. And then the 5 billion, a billion and a half to the hardware, a billion and a half to infrastructure, a billion and a half to some of these discrete ideas, you probably need a 30 or 40 person team total no more, but that business could make a trillion dollars if it was set up that way.