Anthropic filed to go public on October 26. Four days later, tech stocks lost more than a trillion dollars of value in one day, as investors questioned whether all the spending on AI data centers will pay off. If you lead AI work in your organization, a version of that question may be coming your way too: if this is a bubble, should we slow down?
What can Amazon's 2000 stock drop teach us?
Amazon faced the same kind of question about its own stock during the dot-com crash. Jeff Bezos began his 2000 letter to Amazon shareholders with a single word: "Ouch." That year Amazon's shares had lost most of their value, and over the same year the company went from 14 million to 20 million customers. As Bezos later put it, the business kept improving while the stock fell, and the stock is not the company.
Does the bubble debate tell you whether your AI pays off?
Whether AI is a bubble is mostly about how investors value the companies building it, and nobody can be sure how that plays out. It may say little about whether the AI your teams use is paying off. That you can check for yourself, using your own business metrics.
What should you check before the market turns?
A correction could still reach you through prices, tools and budgets, so it may be worth putting each major AI use case you rely on through four checks:
- Write down the business metrics for the use case. It could be hours saved, the cost of resolving a case, fewer errors or revenue brought in, whichever shows best that the use case pays off. If you struggle to articulate it, that use case may be the first one questioned when budgets tighten.
- Check whether it would still pay off if AI cost twice as much. If it would, a price change is unlikely to stop it. If not, it may be the first use case to improve, or to measure more closely.
- Keep your AI workflows portable. Design solutions and systems that don't rely on a single tool, and every so often run the same workflow on another company's model. The AI ensemble concept from issue #40 can help you work out how you might switch.
- Wait with minimum-spend contracts until your numbers are clearer. Some vendors offer lower prices for annual or multiyear commitments, and Anthropic's IPO filing says a growing number of its business customers sign them. Before signing one, make sure the use cases behind it already show the business value you checked in step 1.
A prompt like this one gets you started on the first three checks:
Here are our AI use cases: [list]. For each one, suggest the metric that would show it pays off, estimate whether it would still be worth it if the price of AI doubled, and list what we would need to move it to a different AI model. If more information is needed, interview me for more details about the use cases.
Go deeper
For a reflection on why "the stock is not the company", watch Jeff Bezos in conversation with John Elkann at Italian Tech Week 2025, on Amazon's 2000 stock drop and why he sees AI as an industrial bubble.
Your action step
Take the three AI use cases your organization relies on most and run the prompt above on them. For each one, write down the business metric that shows it pays off, and whether it would still pay off at double the price.
Which of your AI use cases would still be worth keeping at double the price?
If you are reviewing which AI use cases to keep, grow or move, that is the kind of question I work through in AI strategy advisory engagements and in sessions as an AI keynote speaker.