Microsoft Office 365 AI Assistant Pricing Exposed: Annual fee of $100,000, 40% more than the regular version, already has 100 customers.

Microsoft Office 365 Copilot has expanded its trial range and is preparing to further explore pricing models.

According to the technology media The Information, more than 600 of Microsoft's largest customers, including Bank of America, Walmart, Ford Motor, and Accenture, are expected to try out the artificial intelligence features in Microsoft Office 365, with at least 100 customers paying an additional annual fee of up to $100,000 for 1,000 subscription accounts each. (On top of the existing Office 365 annual fee, each account pays an additional $100)

Compared to the classic version, the AI version of Office 365 is at least 40% more expensive, with value-added features such as automatic writing in Word documents and automatic creation of PPT.

With 1,000 accounts paid for by each of the 600 customers, Microsoft will receive $60 million in revenue.

In March of this year, Microsoft integrated generative AI technology into its product matrix and launched Microsoft 365 Copilot, conducting closed testing with 29 companies including General Motors and Chevron.

In May, Microsoft expanded the testing range, inviting the first 600 global enterprise customers to experience this feature, and added semantic indexing, integrating OpenAI's text-to-image model DALL-E into PPT, and other intelligent features such as quickly drafting email content.

According to a person familiar with the plan, one of the reasons for the high pricing is that these features typically require servers with special chips that consume more power than traditional servers.

In addition, the source also said that the number of participants in the trial program has exceeded Microsoft's expectations.

Pricing Exploration

The small-scale pilot will provide Microsoft with an exploration of pricing strategies, as well as a reference for other software companies that add AI features.

According to reports, Microsoft has been weighing two potential pricing models for the widespread application of AI features: one is to charge for these features as add-ons to basic Office subscriptions, and the other is to add AI features to all enterprise users' Office and raise the prices of all subscriptions when customers update their subscription plans. **

Last March, Microsoft raised the price of the basic version of Office 365 by 20%, and the most common subscription version now costs $22 per month. This means that customers who purchase 1000 accounts in this version currently have to pay about $264,000.

Of course, larger customers may have corresponding discounts.

High Cost

Cost is the main reason for Copilot's price increase: both the cost of early training models and the cost of server operation in the later period.

Wall Street News previously reported that in Microsoft's early investment in OpenAI, a considerable part of it was realized through the use of Azure cloud platform points. In order to cooperate with OpenAI, Microsoft also spent $1.2 billion to build a supercomputer to develop models, and invested another $10 billion at the beginning of this year, which is equivalent to about one-sixth of Microsoft's annual cash flow.

In the transaction with OpenAI, Microsoft also obtained the right to embed its technology into its own products, which has become one of the sources of Microsoft's monetization of this investment.

According to people familiar with the matter interviewed by The Information, Microsoft tried to introduce AI functions and gave it time to measure the utilization of artificial intelligence functions by ordinary Office users, which affected the energy cost of running models on Microsoft servers. One of them said that in the next few months, Microsoft plans to expand the preview to hundreds of other companies currently on the waiting list.

It is worth noting that earlier reports claimed that when Microsoft publicly launched OpenAI technology, the server capacity allocated for running OpenAI-driven GPUs on its search engine Bing had already been exhausted.