Big models are rolling out so fast that even OpenAI is struggling with the "pricing" issue.

The sales pressure on OpenAI is increasing day by day, with challenges from competitors and the "squeeze" from its partner Microsoft.

According to media reports, in order to respond to the ever-changing market, OpenAI is striving to reduce the cost of its models, as more and more potential customers have chosen cheaper model services provided by other companies or customized AI software according to their own needs.

For example, enterprise software service provider Salesforce announced earlier that it is promoting more AI software services through open source models and internal development, although the company is still relying on OpenAI's GPT-4 large model.

Jayesh Govindarajan, Senior Vice President of AI at Salesforce, said:

"We are in the early stages of cost reduction in the field of artificial intelligence. As these AI products scale up, we are starting to focus on cost-effectiveness, which will only become more important."

One of OpenAI's flagship clients, Morgan Stanley, has also started exploring other options. It was one of the first companies to sign a large agreement with OpenAI, but there have been reports that in recent months, it has also been purchasing some OpenAI models through Microsoft Azure.

Eli Brosh, Director of AI Research at web software design company Wix, said that the company has transitioned from exclusively purchasing services from OpenAI to using Microsoft's services to support some of its functions. In order to save costs, it is also testing open source models and Google's Vertex AI models:

"Earlier this year, many things like this could only be obtained from OpenAI. Now things are changing too fast. Our goal is to be able to use any large language model, not just OpenAI."

Microsoft's "interception"

From the "defection" of companies like Morgan Stanley and Wix mentioned above, it is not difficult to see that, compared to companies like Salesforce, OpenAI's greater resistance actually comes from Microsoft, which has supported its growth.

As the largest early investor, Microsoft is providing funding and computing power for OpenAI's operations in exchange for the right to use OpenAI software in its products.

However, at the same time, Microsoft is also selling software developed by OpenAI to customers. This means that customers can purchase software developed by OpenAI from both Microsoft and OpenAI, and Microsoft will also provide customers with some bundled software discounts.

For customers, although the same software can be purchased from both places, Microsoft is a large company while OpenAI is just a startup. The reliability between the two is self-evident, not to mention Microsoft's bundled sales policy.

For OpenAI, where customers purchase their developed software also has a huge impact on them, depending on whether the profits are intercepted by Microsoft. According to reports, OpenAI has signed a revenue-sharing agreement with Microsoft. Under this agreement, OpenAI will retain the majority of its revenue from selling models to customers. However, when customers purchase OpenAI's models through Microsoft's Azure, Microsoft will receive a larger share of the revenue.

In addition, OpenAI and Microsoft have also signed a profit-sharing agreement. According to the agreement, Microsoft will receive 75% of OpenAI's theoretical profits until it recoups its investment of over $1 billion in OpenAI. After that, Microsoft will receive 49% of OpenAI's theoretical profits until OpenAI's profits reach the "theoretical cap".

In the past year, OpenAI's revenue has grown rapidly, with the potential to generate over $1.3 billion in annual revenue. However, the majority of this revenue still comes from subscriptions to ChatGPT, which means that OpenAI still needs to prove its credibility as an enterprise software vendor.

Furthermore, compared to Microsoft, OpenAI's sales team is still in its early stages.

According to Microsoft's salespeople and customers, executives of large companies may be more willing to entrust their enterprise data to Microsoft rather than a younger startup company.