I wonder where Matthew Price is rolling this profits now. Recent form 4 show he sells all shares he vests instantly, keeping $NET shares bag at 0. Cognitive dissonance.
I don't know, you'd have to ask him. But considering he owns over 36 million shares worth 2.75 billion, it's not exactly like he has no skin in the game. Michelle owns another billion worth...
Re Cloudflare. While the capital light nature of building out equipment in an ISPs datacenter is a benefit to shareholders, do you think there comes a point in the future where customers view this as a negative? It feels a little like paying a guy to keep your belongings safe and you find out he is storing them at his friends house, but "don't worry he's a good guy".
I might be overthinking this, perhaps physical security is not as important as I think.
Life is trade-offs, so it's not 100% positive, especially in certain countries, but at the same time, regular data centers in some countries are still under the laws of the land, so it's not like these wouldn't be accessible to the state there.
I feel like ISPs are generally aligned in wanting to protect their own stuff from most threats, so that should work.
Watching the physical infrastructure presentations from this years re:Invent gave me an appreciation for how complicated a world class datacenter is to build. AWS is manufacturing custom components in favor of off the shelf solutions because at that scale small improvements matter. I've come to the conclusion (and perhaps incorrectly) that at true global scale the atoms matter almost as much as the bits.
Considering their ambitious plans, it feels like Cloudflare is doing all the right things in terms of bits but not having ultimate control over the atoms is eventually going to be a problem. Maybe I am wrong, either way, it doesn't prevent them from pivoting to a more capital intensive model where they own their datacenters.
Is this something you have previously thought through?
Hosting your hardware in someone else's building doesn't mean you have no control over the hardware. They can send custom hardware there if they want, and they've been partnering with Nvidia for GPU/AI at the Edge, and have been deploying ARM-bases servers to lower their costs/energy use... They can do all kinds of things.
Whatever disadvantage they have from not designing the building is probably small compared to the advantage of being literally next to ISP equipment.
I truly appreciate your responses and hope I'm not wasting your time. I value your opinion. It is the reason I read your work and choose to be a paid supporter.
This is the video I was referencing earlier. If you haven't already seen it, you might appreciate it. Here AWS discusses the thought and engineering that go into simple things like back-up generators, switchgear, and UPSs. So much so that AWS brought the design of these components in house, not to save money but because a world class solution didn't exist. At the scale of AWS, Azure, GCP these little things matter.
I typically when you co-locate in another's datacenter you do not have control over this level of detail. I also suspect that Cloudflare has ambitions to be pretty darn big. Maybe at some point this level of control becomes important to them or maybe it never does. It is something I have flagged as a risk and appreciate you engaging me. Perhaps it is not as big a problem as I first suspected.
The way I look at it -- and again, I'm just a guy in pyjamas, I don't know these things *for sure* -- there's like 500 different variables you can tweak and optimize. Nobody is going to be a "perfect 10" on every single one of those.
Cloudflare so far has heavily optimized on latency and bandwidth costs. Being inside ISPs all around the world gives them a really high score there, and they can leverage that for all kinds of stuff (ie. if you build an Edge network, it helps to be everywhere close to people -- core cloud may have great datacenters, but they're not as close to people in many places and can't match latency).
They've also optimized a lot at the software level to differentiate, so Workers is basically a dev platform too and a lot of the value comes from the software, not just commodity hardware.
The big core clouds often will compete against each other at the infrastructure layer, and there saving pennies matter. But at different levels of the stack, other factors can have a much bigger impact (ie. Snowflake's software is so good that it competes with the core clouds *while being built on top of core cloud and paying them for instrastructure*).
So right now AWS is at 50bn+ run rate and Cloudflare is at under a billion. I suspect that if they even become much much bigger, they'll start to find it worth their time to invest in optimizing some of the other levers, but right now their best bang for the buck is to use their advantages and build on top of them higher-value products like security products and edge development platforms that create huge ROI for what they invest in R&D to create them rather than spend huge sums on custom hardware and capex that wouldn't return nearly as much.
That's just how I'm seeing it, but I could be totally wrong ¯\_(ツ)_/¯
Thanks for reading, for supporting, and for the good questions! 💚 🥃
I wonder where Matthew Price is rolling this profits now. Recent form 4 show he sells all shares he vests instantly, keeping $NET shares bag at 0. Cognitive dissonance.
I don't know, you'd have to ask him. But considering he owns over 36 million shares worth 2.75 billion, it's not exactly like he has no skin in the game. Michelle owns another billion worth...
I have misread forms 4 (Few years experience in US stock market, esp. reading proxy/form 4s). Sorry to intrude with false claims like this!
Public mistake admission:
https://twitter.com/Generalist_Lab/status/1425379385868275713
Go $NET!
No worries, I hate these confusing forms!
Re Cloudflare. While the capital light nature of building out equipment in an ISPs datacenter is a benefit to shareholders, do you think there comes a point in the future where customers view this as a negative? It feels a little like paying a guy to keep your belongings safe and you find out he is storing them at his friends house, but "don't worry he's a good guy".
I might be overthinking this, perhaps physical security is not as important as I think.
Life is trade-offs, so it's not 100% positive, especially in certain countries, but at the same time, regular data centers in some countries are still under the laws of the land, so it's not like these wouldn't be accessible to the state there.
I feel like ISPs are generally aligned in wanting to protect their own stuff from most threats, so that should work.
Watching the physical infrastructure presentations from this years re:Invent gave me an appreciation for how complicated a world class datacenter is to build. AWS is manufacturing custom components in favor of off the shelf solutions because at that scale small improvements matter. I've come to the conclusion (and perhaps incorrectly) that at true global scale the atoms matter almost as much as the bits.
Considering their ambitious plans, it feels like Cloudflare is doing all the right things in terms of bits but not having ultimate control over the atoms is eventually going to be a problem. Maybe I am wrong, either way, it doesn't prevent them from pivoting to a more capital intensive model where they own their datacenters.
Is this something you have previously thought through?
Hosting your hardware in someone else's building doesn't mean you have no control over the hardware. They can send custom hardware there if they want, and they've been partnering with Nvidia for GPU/AI at the Edge, and have been deploying ARM-bases servers to lower their costs/energy use... They can do all kinds of things.
Whatever disadvantage they have from not designing the building is probably small compared to the advantage of being literally next to ISP equipment.
I truly appreciate your responses and hope I'm not wasting your time. I value your opinion. It is the reason I read your work and choose to be a paid supporter.
https://www.youtube.com/watch?v=AaYNwOh90Pg
This is the video I was referencing earlier. If you haven't already seen it, you might appreciate it. Here AWS discusses the thought and engineering that go into simple things like back-up generators, switchgear, and UPSs. So much so that AWS brought the design of these components in house, not to save money but because a world class solution didn't exist. At the scale of AWS, Azure, GCP these little things matter.
I typically when you co-locate in another's datacenter you do not have control over this level of detail. I also suspect that Cloudflare has ambitions to be pretty darn big. Maybe at some point this level of control becomes important to them or maybe it never does. It is something I have flagged as a risk and appreciate you engaging me. Perhaps it is not as big a problem as I first suspected.
The way I look at it -- and again, I'm just a guy in pyjamas, I don't know these things *for sure* -- there's like 500 different variables you can tweak and optimize. Nobody is going to be a "perfect 10" on every single one of those.
Cloudflare so far has heavily optimized on latency and bandwidth costs. Being inside ISPs all around the world gives them a really high score there, and they can leverage that for all kinds of stuff (ie. if you build an Edge network, it helps to be everywhere close to people -- core cloud may have great datacenters, but they're not as close to people in many places and can't match latency).
They've also optimized a lot at the software level to differentiate, so Workers is basically a dev platform too and a lot of the value comes from the software, not just commodity hardware.
The big core clouds often will compete against each other at the infrastructure layer, and there saving pennies matter. But at different levels of the stack, other factors can have a much bigger impact (ie. Snowflake's software is so good that it competes with the core clouds *while being built on top of core cloud and paying them for instrastructure*).
So right now AWS is at 50bn+ run rate and Cloudflare is at under a billion. I suspect that if they even become much much bigger, they'll start to find it worth their time to invest in optimizing some of the other levers, but right now their best bang for the buck is to use their advantages and build on top of them higher-value products like security products and edge development platforms that create huge ROI for what they invest in R&D to create them rather than spend huge sums on custom hardware and capex that wouldn't return nearly as much.
That's just how I'm seeing it, but I could be totally wrong ¯\_(ツ)_/¯
Thanks for reading, for supporting, and for the good questions! 💚 🥃