With basically a blank check from VC, they’ll instead invest in making their models and infra more efficient/better instead of raising prices. They can run a large loss for a very long time.
Why though? They have a capped profit model (theoretically) so there’s less value in this strategy, and their biggest investor would probably prefer that people use Bing instead.
General AI services is a natural monopoly. It has a large fixed cost to develop a competitive model, and lower marginal costs to deliver.
The best* model will have the most paying customers. It’s a monopoly flywheel, the monopoly niche occupant reinvests in the most compute and the best engineers for model improvement, and the N+1 model is even more dominant and so on.
There is second network effect involved in hosting platforms for AI services. This can be an even strongest monopoly. Assuming the “app store” has some common copyrighted APIs for intercommunication between AI tools, it could make it impractical for companies offering models on the store to sell their wares anywhere else. This sends revenue to the monopoly platform owner even after they no longer offer the best model.
*Refusing certain services, and refusing to offer long term guarantees, such as forever access to a frozen weight model, means openAI is leaving the door open to be evicted from this market niche.
With basically a blank check from VC, they’ll instead invest in making their models and infra more efficient/better instead of raising prices. They can run a large loss for a very long time.
Why though? They have a capped profit model (theoretically) so there’s less value in this strategy, and their biggest investor would probably prefer that people use Bing instead.
General AI services is a natural monopoly. It has a large fixed cost to develop a competitive model, and lower marginal costs to deliver.
The best* model will have the most paying customers. It’s a monopoly flywheel, the monopoly niche occupant reinvests in the most compute and the best engineers for model improvement, and the N+1 model is even more dominant and so on.
There is second network effect involved in hosting platforms for AI services. This can be an even strongest monopoly. Assuming the “app store” has some common copyrighted APIs for intercommunication between AI tools, it could make it impractical for companies offering models on the store to sell their wares anywhere else. This sends revenue to the monopoly platform owner even after they no longer offer the best model.
OpenAI seems to be pursuing both avenues like any for profit startup would. Their board has recently voted to lift the profit cap by 20 percent per year. ( https://www.economist.com/business/2023/11/21/inside-openais-weird-governance-structure )
*Refusing certain services, and refusing to offer long term guarantees, such as forever access to a frozen weight model, means openAI is leaving the door open to be evicted from this market niche.
News is the cap grows 20% a year so it will really last until AGI