.png)
3 weeks ago
3
Yesterday we hosted our annual general meeting with our ≈25 largest investors in NY. It’s great to get together as a team IRL and obviously very nice to get backslaps and high fives from the investors congratulating me .
I’m sure we’ll have more thoughts to share coming out of that but there were a couple slides in particular that I wrote/spoke to that I think bear repeating here now.
Every year for the last several AGMs we’ve presented lots of reasons why we’re not just chasing the latest, hottest AI companies and are not bulled up to just ape in with the crowd.
The reasonable pushback/line of questioning has been “but AI companies grow so fast and get all the markups!”
There’s four reasons AI companies are growing so quickly:
Attacking bigger line items with bigger ACVs and clearer value props (selling work, roughly) than their pre-AI counterparts. When you move from selling a tool to a person to do a task to simply doing the task, you have a much clearer value to your customer so it’s faster to adopt AND they’re willing to pay a lot more for it. That is a hugely significant change from a prior generation of (particularly vertical) software companies. You can much more confidently underwrite a bigger purchase as a customer and the markets get meaningfully bigger. That the products can actually work is of course a big part of doing that well.
Low/no gross margins (giving away product). There is a whole class of companies that are effectively reselling inference/compute/models/chips/data center capacity from other providers at a loss or with razor thin margins. Selling a dollar for 90 cents is a great way to grow and you can do it if you have the capital to subsidize it which...
Unlimited money to invest in S&M and unlimited experimental budget to buy AI. All the headline growth props up fundraising which injects capital to invest in more growth. In the main this is normal. At the limit is not. When the sellers have unlimited money to get their products into market and the customers have unlimited money to “AI-ify” themselves, you start seeing a lot of roundtrip venture dollars and growth without signal. Remember, revenue should be a lagging not leading indicator of value.
Funny accounting. This is becoming a meme but it’s now clear that ARR is often not annual recurring revenue (experimental, usage based), annual recurring revenue is often not annual (high churn), and revenue may not even be revenue (GMV, credits, LOIs, etc). It’s starting to feel like “Who’s Line”
The first of those (expanding product surface area ➡️ expanding ACVs ➡️ increasing velocity) is genuinely important and interesting. We are tracking that quite closely and reflecting it in our investing. While it won’t always mean investing in lightweight vertical AI/software companies (Metropolis is arguably among the biggest and fastest growing applied AI startups in the world), it is a real thing that really matters.
Will some market sizes still be constrained? Yes. Is competition is a huge concern, of course. But this is a moment in time when people can do more new things and do them faster than in a long time (maybe ever). But…
The other three are bubblicious at best and will end in tears for all but a few companies. Anything that can go 0 to 500 overnight can go back to 0 just as fast, all the moreso if that 500 is really 100. Founders are under immense pressure to hit those benchmarks, even if they have to manufacture progress “artificially.”
By this point, the meme of “$0 to $XM in ARR in Y days” is so common and widespread it’s become a competition to see how can say the most ridiculous and bombastic number. We are going to see some genuinely great products, ideas, companies get killed by VCs demanding they hit made up benchmarks for growth which ultimately preclude the potential for durable value.
It’s a 15 year game. It’s a sprint AND a marathon.
The end of pricing power - We basically take for granted that good margins = good businesses and bad margins = bad (or at least not-yet-good) businesses. But now negative-gross-margin-but-extremely-fast-growing AI businesses are the hottest game in town - so hot that they even have a name “AI Supernovas.”
Revenue Comes Second. - Historically revenue is a lagging indicator of value creation. Do something novel and important for your customers first, and the revenue (value capture) should follow second. But now, revenue is often a leading indicator of value, hypothesis validation, and PMF - sometimes even a false flag. These companies are (implicitly or explicitly) being run themselves as “number go up” machines instead of “true, false” experiments.
