Friend pendant unsustainable business model

Friend’s hardest problem is that its most engaged users are also its most expensive users. Every extra hour the pendant listens creates more Bluetooth transfer, more clip processing, and more Claude calls, but the customer still pays only once, at $99 to $129. That is the opposite of how durable AI products usually work, where heavier usage either triggers a subscription or sits inside a broader device ecosystem that can absorb the cost.

• The product flow makes the cost problem concrete. The pendant is always listening, sends audio to the iPhone app, the app segments clips, and those clips are analyzed by Claude 3.5. That means inference spend rises with engagement, while revenue stops after the initial hardware sale.
  2 sacra
• Comparable wearables that found stronger economics added recurring software revenue. Oura grew to about $225M in 2023 after adding a $5.99 monthly subscription, which helped push gross margin above pure hardware levels. Friend chose the opposite structure, no subscription and unlimited AI usage bundled in.
  3 sacra 2 sacra
• The category has already shown what happens when standalone wearable economics break. Friend shifted from hardware plus software to a web chatbot in October 2025 after only $348K in sales as of September 2025, while Limitless and Humane also moved away from independent pendant hardware. The market is rewarding software and platform bundling, not open ended per user AI costs on a one time purchase.
  4 sacra 1 sacra

The next durable version of this product will look more like metered software or a bundled feature inside glasses, earbuds, or phones. For standalone wearable AI to persist, revenue will need to scale with usage, either through subscriptions, premium features, or higher value workflows that justify ongoing spend.