Highly leveraged growth stocks: compounding mistakes
I think I have told you before, but I am just a kid. And as a kid, I observe markets and try to recognize patterns. And because I am a kid that just joined the market in 2020, lots of patterns I took notice are pretty dumb.
One pattern that I recognized is that: companies that report some form of pre-revenue number are worth more.
The canonical example is Shopify: they disclose GMV and GMV goes up every year. Therefore, a possible model for Shopify in 2021 was: they grow GMV 20% per year for a decade while growing take-rates from 2.5% to 5% through logistics and payments. Therefore the 20% GMV CAGR becomes a 28% Revenue CAGR. And because back in the day we were optimistic, we would say that Shopify would double its margins from 12% to 25%. That meant that profits would grow at a 38% CAGR.
GMV: 6.2x
Revenue: 12.4x (double the take-rate)
Profits: 25x (double the margins)
That meant that Shopify would be making $9.5B in EBIT in 2031, with the company $180B enterprise value at the end of 2021 at 19x 2031 EBIT. Obviously not cheap, but not out of this world.
Of course, the market did mathematics like this to every stock it could find.
At certain point during the bull market, Bessemer Venture Partners did this study where they identified that during the limited period where we have data for public SaaS companies, they usually decelerate just by 80%: if you grew 20% in the year 1, you would grow, on average, 16% in the year 2. They called it growth endurance.
Let’s pretend that’s how the market prices companies, to try to gauge how much a deceleration impacts the business net present value. Using a 15 year timeline, in which the average softwareCo have 0% FCF-SBC margins from year until year 7 and then 40% FCF-SBC margin from thereafter, using a 3% perpetual growth rate and a 10% discount rate, we achieve to the following “fair” NTM revenue multiples:
From this table, you can therefore estimate the change in the expected net present value as you go from expecting x% to x-3% growth in the next twelve months.
If you believe (like I do) that markets are doing math like this one1: extrapolating current growth to infer long term growth runway, you can make sense of 10%+ movements in growth companies after 2% beats.
I decided to pen down this one after rereading Hayden Capital letter on Coinbase. Unlike my dunk on Twitter may sound, I read at the time and thought that Fred was doing good research there. I ain’t smarter than the next guy. But I try to make mistakes (preferably other people’s mistakes) not go in vain and learn something from them.
Anyway, I tweeted:
You can see my full thread here, but to summarize the Coinbase thesis was:
Total Crypto market cap goes up.
Coinbase MTUs (Monthly Trading Users) go up according to total crypto market cap linearly.
Coinbase revenue per MTU goes up according to log(total crypto market cap)
It’s an app! They will show some operating leverage in that scenario.
You can see where I am going: there’s a TON of leverage in the total crypto market cap number (as in the case of Shopify GMV) in the model. But actually, total crypto market cap when south, bringing down with it MTUs, Revenue per MTUs and margins.
At least in Coinbase, one would make a lot of money if we had $4T total crypto market cap in 2025. But there are lot of times where:
There are two variables, say users and ARPU.
Both are growing fast, and analysts model a gentle decel over the years
Revenue is expected to CAGR the product of both, which means fast
People don’t realize that the two variables are actually dependent.2
The market reprices the stock, now discounting a gentle decel from a smaller growth for both, which means even slower growth for topline.
The stock price drops bad.
In summary: due to compounding effects in companies’ growth models, you can have sharp price swings from small changes in assumptions. Even without financial leverage, you’re playing with fire. Beware and play it accordingly.
It’s not because it’s a software company with predictable revenue and even predictable growth rates, that the net present value is predictable.
Some nerds will use Net New ARR and variations. I prefer NNARR too, because of the elephant.
If you have an online video streaming service, if you launch good shows, both new users will sign and current users will accept higher prices. But if you launch bad shows, users won’t join and current users won’t accept higher prices.