Why consumption-based AI pricing doesn't work for manufacturers
The Economist ran a piece last month on soaring AI usage costs. The numbers are hard to ignore: Uber burned through their entire annual AI budget in four months, while another company spent $500 million on tokens in a single month. Sam Altman of OpenAI has described the mounting customer costs as "a huge issue."
Now that the bills are piling up, companies are rethinking not just how much AI they use but what they're willing to pay for. This is a shift from just a short time ago when companies were encouraging employees to use AI in their day-to-day work.
What does this mean for the industry?
The conversation is now turning to how vendors should be pricing AI in the first place. In a recent IndustryWeek piece, author Stephan Liozu states that industrial leaders are about to repeat the same pricing mistakes with AI that they've made before. He suggests that organizations should not focus on consumption based pricing but rather on pricing based on the value AI is meant to deliver, such as avoided downtime, scrap reduced, yield improved, and warranty claims prevented.
These are metrics plant managers are tracking today so how do you know your vendors are focused on the same intended outcomes as you?
How should manufacturers evaluate AI vendor pricing?
Rising costs can force organizations to put guardrails around AI usage. But the more practical move is to use this opportunity to evaluate your technology strategy based on the following criteria:
- Data usage. The amount of data you have is only going up. Every manufacturer we work with is constantly collecting more sensor data, more manufacturing data, and more people who could benefit from access to it. You shouldn't have to pay more every time those numbers increase. Instead partner with a vendor whose pricing doesn't impact you for scaling adoption.
- Purpose-built AI. AI deployed at the application layer is built for a specific job such as reducing defects. However, a general-purpose copilot can be used or something that has nothing to do with your operation, like planning a vacation, and you're still paying for that usage.
- Build vs. partner. Focus your team's time and resources on what they're best at, and leave AI development to a partner who can do it more efficiently at scale.
Building AI capability in-house isn't a one-time cost, it means keeping a team on staff indefinitely to keep pace with a technology that constantly changes or risk being left with yesterday’s technology. A vendor focused on optimizing for the latest developments in model efficiency can provide the most value at the lowest cost.
What value-based AI pricing looks like
Viaduct has been focused on driving customer value and outcomes from day 1. Our solutions are purpose-built for connected assets, combining our patented AI technology with deep industry expertise to help teams make faster decisions, increase throughput, prevent failures, and eliminate unnecessary costs.
Our pricing reflects that. It is based on unlimited data and unlimited users so there is no per-seat pricing or charges for additional data usage, tokens, or compute.
We believe it's only a matter of time before all vendors follow suit.
Share your thoughts and comments by emailing me at bora@viaduct.ai.
