Fellow regards, I have just had an enlightening conversation with a tech exec that I wanted to share.
Subscription based licensing means predictable high margin revenue. The economics drastically improve with scale as a 200k a year customer usually does not cost more to service than a 50k a year customer. Traditional seat based or org size based licensing models therefore lead to high valuation multiples.
Meanwhile, LLM reliant “AI” use cases require usage based licensing due to the high cost in compute of each workload. But usage based licensing is not predictable, and the costs do rise in tandem with revenue. AI usage based revenue is a bit like services revenue, and even within that category it is particularly unpredictable: You may have a lot of revenue today and very little revenue tomorrow because customers are not generally committed to long term minimum usage contracts.
So the reason ServiceNow and SalesForce stock prices are in the gutter is not just because some of the market thinks you can vibecode a platform in your garage over the weekend. It is also because usage based licensing models simply do not deserve the same revenue multiples in the eye of sophisticated investors, and both platforms have been forced to transition to a usage based model for some of their licensing to enable AI use cases on their platforms.
I am not saying current valuations are fair, I still think ServiceNow in particular is likely to recover and I hold stock myself, I just wanted to share this thought with you all.
AI usage revenue is valued at lower multiples than subscription revenue
byu/Tornagh inwallstreetbets
Posted by Tornagh
4 Comments
It’s called commoditization
But, uh, Salesforce and ServiceNow are seat based. “Forced to transition” isn’t a fair framing, they’re seat based and adding usage based products.
Plus, you’re fundamentally misunderstanding the dynamics behind these models. Obviously seat baed has better and more stable margin – it’s locked in you get paid whether they use it or not. But SaaS has started drifting away from that pricing for two enormous reasons:
**1)** Clients hate it – they strongly prefer “pay to play” where they perceive better bang for their buck (even if it’s not actually a better deal).
**2)** In a leaner, more efficient AI world, seat licenses are already getting cut back dramatically. In fact, if your product implements AI well, then you inherently need less seats to manage it.
Plus plus, good luck vibes coding a Salesforce… They’ve been oversold on a fundamental misunderstanding of their moat.
Not a bad theory. I personally think AI is overrated and it will take a lot longer than people think to replace humans.
one night I got really high and made Google Gemini repeatedly prompt a never ending number of 6’s. eventually when the prompt stopped repeating 6’s (after about 5 minutes) I would give positive feedback then use the “longer answer” feature, then I would go back to playing STS2.
o I also got it to count to 10494 once on voice, I guess if u think this type of usage should be valued higher than it already is be my guest lol. AI isn’t what u think it is going to be