Autodesk wants you to pay for outcomes, not just access

Autodesk held its bi-annual investor event yesterday, and much of it was product-related, a repeat of what we heard at Autodesk University. One thing that became clearer yesterday, though, was how Autodesk will shift its licensing model yet again to include more consumption-based offerings.

Today, Autodesk primarily sells on subscription: pay monthly or annually, use freely until it expires. But its store already shows usage-based models. For example, Civil 3D is $240/month (annual commitment), $360/month (flex), or pay-as-you-go at $300 for 100 tokens. Since Civil 3D consumes nine tokens per day, that’s about 11 days of use before you need to buy more tokens.

In the future (and soon, it sounds), some products will also be sold with outcome-based pricing. This makes sense when we consider the AI products Autodesk is planning to bring to market; for example, describing a building and having a Forma AI assistant create a preliminary design that a human then refines. If Autodesk sold that AI tool as an annual or monthly license, it would likely be prohibitively expensive for a solution that a designer may not need very often. Instead, charging by “preliminary design created”, the price is (we hope) more reasonable.

Outcome-based pricing reflects Autodesk’s thesis that AI will fundamentally change how its users work—and how they expect their tools to perform. This style of pricing enables people to adopt AI as they become comfortable with it and integrate it into an increasing number of their processes. Additionally, as systems automation becomes more prevalent, it will make more sense to charge per outcome, since AI agents may use multiple individual tools to complete their tasks. And of course, all of this AI will utilize a significant amount of cloud computing capability, which is already charged on a usage basis.

Outcome- and consumption-based pricing make a great deal of sense from Autodesk’s perspective. They’re preparing for a world where AI will enable customer to create complex and interconnected design, build/make and operate workflows. And the company can start to get a return on all they have invested in AI research and development while still fine-tuning the specifics of pricing to accelerate adoption  — but we still don’t have actual pricing, so buyers aren’t yet clear on what this will cost them.

As for the cost, we don’t yet know. SAP, whose Joule agents perform various business-process tasks, sells “AI units” that can serve as a proxy for PLMish pricing. SAP’s Business AI Estimator (here: https://discovery-center.cloud.sap/ai-estimator) lets you select a Joule offering (HR, finance, customer experience, etc.) and how many people would be using it, then calculates roughly how many AI credits would be consumed per month. Ten customer service agents, requesting 52,000 actions from Joule AI, would cost roughly €560/month, after paying for the base package, which might be another € 75 and up per user.

The problem that SAP and its users are running into is that this isn’t transparent enough. Enterprises are attempting to estimate their costs based on projected AI scenarios, which are still evolving as workflows adapt to new patterns and people learn to utilize the tools (or not). 

The best solution that SAP customers I’ve spoken with have found is to use free AI trials for as long as possible to determine the real usage patterns, and then negotiate price caps for an initial period to address the uncertainty. 

Back to Autodesk. In its investor materials, Autodesk disclosed that 17% of fiscal 2025 revenue (the year ended 31 January 2025) was already consumption-based, primarily from the enterprise business agreements (EBAs) signed with its largest customers. EBAs are flexible, customized software licensing deals that give each customer enterprise access to the range of Autodesk technologies they need, with flexible licensing that (as I understand it) is based on EBA tokens for cloud services and Flex tokens for products. 

What’s interesting about this is that the consumption-based contribution to revenue was just 3% in fiscal 2015 (even though EBAs were announced in 2007 — but that was a different time, with less cloud capability and other licensing models). Anyway, using Autodesk’s data, increasing from 3% to 17% of revenue from FY15 to FY25 implies a 30% compound annual growth rate in consumption-based revenue so far, which the company clearly hopes will continue as it adds more of this type of pricing.

Finally, one other announcement teased at AU but described more fully yesterday was how Autodesk plans to monetize its platforms. Autodesk intends to charge customers for API (Application Programming Interface) calls made to Forma, Fusion, and Flow. That’s not really new —Autodesk has always charged for API calls— but it sounds like this may be more intrusive since the company said it is including an allowance of API credits with each of its main subscription packages, and, [paraphrasing here] “once usage exceeds the included allowance, customers can buy extra API credits”. 

Autodesk has often used words like “monetizing” and “revenue per customer,” which make investors happy but scare users, who see their prices going up without any obvious increase in value. This time feels different: AI has numerous potential use cases, but it won’t be uniformly adopted. The company said yesterday that the “design a building for me” AI feature will be included in Revit. However, not all Revit users start a project’s lifecycle at this early stage, and so, most won’t need this particular function. Charging all customers for something only a subset will use isn’t fair and could drive customers away. Charging incrementally for that value makes sense and could draw customers to use it — if Autodesk can get the pricing right. 

I think we’re getting to the point where AI’s shiny-new-thing vibe is wearing off. People are tired of hearing promises of a workday that’s radically changed by technology and want to see actual capability, and roadmaps that promise real features rather than vague hand-waving. We’re finally starting to reach the point where vendors can show AI-based offers that can reduce benefits to a buyer’s bottom line. Now we need the commercial models to catch up.

—

You can watch the investor event replay here, no registration required: https://investors.autodesk.com/events/event-details/autodesk-2025-investor-day-0 . The whole thing is 3 hours, but if you focus in on the industry (manufacturing, media or AECO) you care about, those are 15-20 minutes long.