Cloud profits at a cyclical peak

It's been a 10yr growth boom for cloud services, aka companies selling server space so their customers can host apps, databases, and store files, among other things. Revenues of the big three tech leads here last quarter were massive (Microsoft Azure: $25B,  AWS: $20B, Google $6B) and it's not surprising. These services offer a lot of benefits by letting companies: 

Prior to the cloud, companies typically bought their servers and ran their own networks. This is hard and requires shelling out cash upfront. 

So will the boomtimes continue? Long-term: Providing cloud services is a good business and will become more important over the next 20-30yrs. Short-term: things will likely get ugly. I think cloud provider's growth and margins will fall sharply in '23 and '24 for two reasons: customers will spend less and cloud pricing will drop.

Cloud spending likely to fall
Why would cloud spending fall? First, it's been driven up sharply by a historical increase in venture funding. According to CB Insights, VC funding has jumped from less than $10B/quarter in early 2014 to frequently logging over $40B/quarter since 2018. Meanwhile, the "batch" of companies this has produced may not be sustainable and are rapildy going from "grow at all costs" to "survive at all costs", and that means cost-cutting.

The fate of the unicorns has mostly been to get creamed in the public markets. If you take a basket of these companies the stock performance has been grim. Just glance at the earnings and stock charts of Gitlab, Asana, and WeWork for examples.

Investors are more skeptical now, so are pressuring companies to cut costs. But what do you cut? Well the top costs for the new wave of venture companies tends to be sales and marketing, expensive engineers, and cloud services. Most have already started cutting staff so why not tackle cloud costs too? After all, some of these company's survival depends on it. But how will they save on their cloud services? I'll get to that at the end. 

Cloud leaders likely to get into a price war
The second factor that could drive cloud profits down is a price war. In the 90's, lots of companies like Level 3 laid a ton of fiber optic cable and licensed it out. It was boomtimes and tons of corporations and telcos were leasing it. Until 2001. Then, post-bubble burst, the fiber players were left with lots of capacity and sharply lower demand. So what did they do? They could leave their pipes idle and lose potential business, or they could price just above their marginal cost of service to get as much money from that excess capacity as possible. They did the latter. All it takes is one big player to start cutting prices and everyone else has to follow suit or risk losing customers.

Cloud switching costs may be a pain, but given the cost pressure companies are facing, plenty will be willing to bear it and move to Azure if their Amazon account rep won't play ball and cut the rate.

What have cloud customers been doing so far?
Customers moving from the cloud back to owned servers altogether has plenty of precedent. This week DHH surprised me by announcing that they're moving Basecamp and Hey off the cloud entirely to save potentially millions by owning their services vs. renting. Dropbox shifted from AWS to their own infrastructure in 2017. Basecamp is private, but founders say it makes "tens of millions a year" in profits. Dropbox is public and made $342M in profit over the last 12mo. 

Gitlab, meanwhile, decided to stick with cloud services back in 2017 and lost $172M over the prior 12mo. Cloud costs aren't the only profit driver here, but for a storage and bandwidth intensive product like theirs, it's material. In their article, which amalgamates experiences from their user community about costs savings, several mention lowering costs by 50 - 90% by moving servers in-house, so Gitlab should probably revisit that '17 decision. 

What should cloud customers do going forward? 

Almost everyone should care more about their cloud costs over the next 5yrs. Small companies should use Heroku, Render, or someone similar to get off the ground fast. If they get big, then they can try a few other things to keep server costs low: 

Finally, if these aren't enough, go fully in-house with your servers. I like to think of this as "yoloing" it and only recommend it if you've had success doing #3 above already. I'm not an expert in devops or back-end infrastructure, so reader beware, but these all seem like reasonable options given the state of the market. 

 
---

Disclosure: As of October '21, I have financial short positions against Amazon, Google, Microsoft, and Gitlab, pretty much for the reasons above. I didn't write this to "talk the stocks down" since I don't have a large audience, but rather because this topic and the dynamics around it interest me.