The deployment of capital in training models for AI raises concerns about the sustainability of being the first to train models and whether the costs will significantly reduce in the future. The rapid acceleration of model development and training, coupled with cost reductions, poses a challenging investment exercise. The business models must have an advantage of being the first to market and generating new data to maintain a lead. However, it is uncertain when the market and technology will reach a tipping point, making it difficult to invest in technology. Instead, building a sustaining advantage in the market, by being first, building a customer base, and accelerating features, seems more viable. This applies particularly to the application and tools layer, as opposed to competing at the lower stack levels that require significant capital investment. Despite the coordination benefits of cost reductions, making money in this field currently appears unlikely. Open source solutions are also brought into question.