The relationship between Microsoft and OpenAI is looking increasingly strained as Microsoft plans new offerings on its Azure cloud computing platform.
Microsoft is rumored to be planning to make new software from a company called Databricks available on Azure. Databricks provides analytics and AI solutions to manage and analyze large data for enterprises.
Once deployed on Azure, companies will be able to use it to create AI apps using their own models or repurposed open-source ones.
This sounds great but it gets a little awkward when you remember that Microsoft has invested over $10B in OpenAI. The AI models that OpenAI provides are proprietary and you’ve got to pay to use them. So with this move, Microsoft is essentially competing with the company it invested in.
Ironically, Microsoft will be using OpenAI’s technology to create a chatbot to help users navigate the Databricks software.
The Azure-Databricks solution further erodes the potential for OpenAI to monetize its models. OpenAI had a head start on most other AI companies and makes money by charging for API calls to its models. But it’s struggling to find a way to be profitable.
And why would you keep paying to use an LLM when there are so many free ones becoming available? With Microsoft making the new Databricks software available on Azure it’s going to be easier than ever to bypass OpenAI and just deploy one of the many free models available.
— The Information (@theinformation) August 17, 2023
It looks like Microsoft has realized that the money doesn’t lie in the AI models but in the processing of AI apps using the models. It’s pushing hard to get companies on board Azure and isn’t going to force them to use the AI it has invested in so heavily.
The ongoing data security concerns surrounding ChatGPT are also making more and more companies look for secure, private AI solutions for their staff to use. Big tech companies like Apple, Samsung, and Amazon have already banned or restricted their staff’s use of ChatGPT.
OpenAI continues to insist that its products don’t pose any threat to private data but corporates aren’t buying that. With Microsoft very focused on enterprise solutions it’s having to find ways to mitigate these concerns.
Just last week it released a private, secure AI chat model on Github and highlighted the security issues with OpenAI’s ChatGPT. Microsoft promptly removed the code a few days later and denied that it had any such offering.
With Databricks it might be that Microsoft is trying to add one layer of separation between it and free AI models to try to pacify OpenAI. The reality is that it’s becoming difficult to see how a mutually beneficial relationship between Microsoft and OpenAI will work with these conflicting interests.
OpenAI has to find a way to make money and the options of charging for its models or maybe monetizing ChatGPT just don’t seem viable with so many other options on the table. Meanwhile, Microsoft is happy to sell its Azure processing power to all comers and is becoming increasingly AI model agnostic.
This isn’t going to end well for OpenAI.