You've gotta spend money to make money
I don't know how many AI startups there are in the world, but I have looked at hundreds of them in the last 18 months. Let's assume I've seen only a small fraction of what actually exists, and the numbers get staggering quickly.
I've written recently about why, despite quick early sales traction, it is pressingly difficult for AI companies to get renewals. I've also written about why most AI projects fail before they start. These are not especially controversial takes. Despite these patterns, investors are still pouring money into new AI startups hand over fist, and AI hype continues to escalate (especially in Silicon Valley).
I'm a founder. Why haven't I started a new AI company yet?
In ancient times
We started Textio, which was the first AI for HR company, in late 2014. Though I'm no longer the CEO, my cofounder/spouse is the CEO. At one basic level, this is why I haven't started any new company, AI or otherwise: we already have one in the family.
But there's more to it than that. The reality is that, even though I have met many impressive and gritty founders with ambitious visions and operating skills, I just haven't seen that many AI startups I believe in.
Look, I might be wrong. Maybe I'm biased from my first-hand burns. I know where too many of the bodies are buried, all the things that can go wrong with pricing and packaging, the eventual pain of living in an outcomes-based world. I also see how much harder it is for new AI companies to differentiate than it was for us at Textio; we invented our own models, but new companies are nearly all using the same commoditized APIs.
Making the math math
On the other hand, the valuations of the hottest AI companies are astounding. What's more, the valuations are often matched by impressive revenue growth. While you can argue that valuations are vapor, revenue growth is legit. The bubble may burst at some point, but it certainly hasn't yet.
At the same time, the cost to serve AI-heavy user experiences is extremely high. Both legacy companies and well-funded AI startups are burning through oodles of money in their race to get to market. They are hemorrhaging cash to try to move out in front, often operating at significant losses. (This doesn't even comment on the environmental impact; Sam Altman says that the average ChatGPT query uses 1/15 of a tsp of water!!!)
To put it bluntly, AI unit economics are currently lousy. Is there any tech coming that will change this, in the way that hardware has changed unit economics in the past for the software industry? Probably. But when?
Unfortunately, right now, it's not clear that moving out in front buys you much. Very few spaces (especially in B2B) look like green fields right now. Name a problem space, and I will show you 5 or 10 or 20 or 100 startups that are pitching identical value props, with solutions built on identical AI that isn't even slightly proprietary.
Don't believe me? Search up any AI market landscape map: for finance, HR, marketing, dev tools, legal, or anything else. The visuals are so dense you often can't even discern individual logos.
As if that wasn't enough, legacy apps are rapidly moving to monetize their user data, and they are charging a premium. This is a notable change from how most productivity apps were positioning themselves a few years ago, when they were all investing in easy APIs so that other apps would build on their platforms and drive up platform value.
To recap, what we have is companies spending piles of money to move out in front with solutions that feel commoditized before they've even launched, relying on data that is becoming increasingly expensive to access. And this doesn't even begin to speak to the fact that many potential customers for these companies are aiming to remove their dependency on AI vendors entirely, by building their own internal tools with the same APIs that the vendors are using.
Yikes.
Believing in the path to win
Look, I'm a glass-half-full person. Despite everything I wrote above, a few AI startups are going to win. I just don't think many of them are going to look like traditional B2B applications. Imo, they're more likely to be companies that make it cheaper for people to build and deploy their own AI tools in various ways.
For the app companies that do make it, AI won't be their main value prop; assuming the cost to serve problem gets solved eventually, AI will just be a standard set of enabling capabilities, like cloud before it.
What do you think?
Kieran
From now until August 31, my Viral Data Stories 101 course is half-price for nerd processor subscribers!
My latest data stories | Build like a founder | nerdprocessor.com