Are business models of current AI companies unsustainable with pervasive economic incentives?

We can analyze the current AI investment environtment as an economic system made up of multiple players (OpenAI, Anthropic, xAI, NVIDIA, ASML, Oracle, ...), each with their own incentives.

For this system to be sustainable long-term incentives need to be aligned between players. Disclosure: right now they aren't.

The entire AI investment hype relies on OpenAI mainly. OpenAI is deeply unprofitable, they have around 13B$ annual losses, this is a basic economic equation:

**Profit = price \* volume - Capex cost - Opex cost**

The core problem lies in Capex cost, the enormous computing power required per prompt (GPUs), data centers, electricity, refrigeration and cloud. And here is where it gets concerning for me:

1. Nvidia is a major investor of OpenAI.
2. If OpenAI made their business sustainable by reducing compute per prompt (optimizing models by requiring less GPU), then Nvidia revenue (or its expectations thereof) would fell because demand chips would shrunk.
3. Former point means Nvidia has no incentive to support a path where OpenAI turns out profitable by reducing Capex cost variable.

The other path for profitability is increasing the price \* volume component, and this is not a path forward either:

1. B2C OpenAI users are highly price-elastic, if OpenAI rises monthly suscription price to let's say +100$/month mosts users will leave for free or cheaper alternatives like Gemini or DeepSeek. So its a race to the bottom in terms of prices for these business model as competition increases.
2. B2B OpenAI users can be a bit more flexible, but even many SME won't be able to justify the cost, especially with open-source LLMs catching up fast. For it to be sustainable there should be massive ROI for AI implementation in these companies (cutting cost or increasing revenue).

Even B2C users that pay $200/m subscription are unprofitable for OpenAI, they have a problem called adverse selection and moral hazard ([https://en.wikipedia.org/wiki/Adverse\_selection](https://en.wikipedia.org/wiki/Adverse_selection), https://en.wikipedia.org/wiki/Moral\_hazard). Users that will intensively use OpenAI are those that are willing to pay subscription, and if these are of unlimited use (even with subscription), they incur an adverse selection and moral hazard problem which is why these users are still unprofitable for OpenAI. However, if they cap the usage even with subscription they risk not capturing enough market size for them to recoup Capex costs.

As a consequence If OpenAI increases price then price \* volume stays flat or even decreases, not solving the profitability issue. AI usage is currently being subsidized by unprofitable business like OpenAI, Anthropic, and xAI, and if these businesses wanted to turned profitable again their business model would collapse, consumer demand would be killed instantly, they have no way to pass on the Capex cost onto the end consumer.

This seems structurally unsustainable with the current Capex structure