Microsoft AI pricing

Microsoft Copilot has entered the AI chat, but they have multiple pricing models depending on where you use it. And I think the differences are interesting ??

First: Microsoft Copilot for Security ??

Microsoft’s security cloud is a more emerging product – and use case – and this is reflected in the pricing model.?

Toward our goal of enabling security for all, Microsoft is also introducing a provisioned pay-as-you-go licensing model that makes Copilot for Security accessible to a wider range of organizations than any other solution on the market. With this flexible, consumption-based pricing model, you can get started quickly, then scale your usage and costs according to your needs and budget. - Microsoft product announcement blog

This pricing model enables Microsoft to:

1?? Allow customers to try before they commit to buying, and also use as much of the product as they need with relatively few barriers to entry.?

With Microsoft selling a lot through partners, it will also be easier for those partners to bring new users onto the platform without having to 'pre-sell' the benefits of this new product.

This is a far more niche product than others like Microsoft 365, meaning two things:

• Maximizing the number of users is very important, as the total population is limited.
• Users of the security product are likely to be working intently on security topics = the average user is much more likely to be a power user with a lot of usage.

2?? Monetize the growth of customer usage if the product hits the mark.

If the product doesn’t hit the mark, pivot to a different model as needed, with limited customer commitments to sort out.

3?? Monitor how much the product is used and cover the variable cost (e.g. of computation, etc.).?

If a product is priced as a flat or fixed subscription, then power users (i.e. customers using the product a lot) are likely to be expensive to serve. As a newer product, the algorithms may also not be optimized to minimize the amount the product costs to run.?

Note that a big variable here is the arrangement that Microsoft has with OpenAI in terms of the cost of computation; that's likely to reduce the cost compared to other developers building on top of LLMs.

4?? Achieve a big payoff.

The break even amount for this pricing model is under eight hours, or one day per month.?

Compare that to a $30/month flat fee like Copilot for 365 (below), and it’s likely to pay off for Microsoft. If security professionals start relying on this functionality in their day-to-day activities, it is likely to cost them more than if they were paying an up-front flat fee.

Second: Copilot for Microsoft 365 ??

On the other side, Microsoft has priced Copilot for Microsoft 365 as a flat $30 per month.

The use cases and demand for this product are far better known, as are the expected costs to run the service (since it has been on the market longer).

The flat rate pricing matches the pricing for the 365 suite in general, so it becomes an easy and logical add-on item.

The product caters to a much more general audience. This means that:

• There is more value in getting as many people as possible to commit to a subscription, as it will be hugely valuable to MSFT's balance sheet.
• There is likely to be a big mix of usage, with lots of lighter/less frequent users. These users will be highly profitable and will subsidize the lower profits made on heavy users.

Microsoft is just one of the countless companies navigating the uncertain road of pricing new AI products. We’ve seen a whole range of different pricing models pop up, with many companies gravitating toward pricing AI tools with outcome-based metrics or usage-based metrics and anything in between. (I discussed this spectrum in-depth in my last blog.)

What remains to be seen is how it will play out and whose pricing models will actually work for their user base and their bottom line.