135: Cloudflare’s Cornerstone Decision & Downstream Effects, Nvidia Q1, Jeff Bezos Buckets, Snowflake's HQ, Ford's EV Plans, Twitter's Privatized Protocol, and Jim Keller
"cutting-edge cultural commentary"
Contrarianism is overrated.
True contrarians never try to be contrarian.
They merely try to be congruent and, only as a side effect, appear contrarian to the dissonant mob.
I realized that there’s this weird balancing act when you write as he does — or even when you just read this type of analysis… I found myself most excited to read what he has to say about the companies that I already know well, even if I know that I’ll probably get the most value from reading about companies that I know almost nothing about (Peter Thiel’s ‘zero to one’ framework also applies to knowledge, not just companies).
So when you write, do you focus on well-known and well-loved companies because it plays into this phenomenon, and you’ll get more and happier readers, or do you focus on the broccoli and go for lesser-known companies where your writing is going to make a bigger difference to someone’s knowledge base, but they may not be nearly as excited about it... ¯\_(ツ)_/¯
It’s a tough gig!
🛀 There's lots of advantages to youth, but being constantly terrified of social faux pas, obsessed with what others may think of you, and judging others based on superficial attributes/in-group vs out-group status aren't the ones people should try to emulate…
What truly keeps you young at heart is not caring what other people think (well, most people — you should care deeply about what people who are close to you or that you respect think).
👩🏫🎨 This tweet by David Perell (and I the only one who never remembers if it’s two “R”s and one “L” or vice versa?) made me think about what I wish I had back in school:
(I first misread his tweet and only saw “museum” but my brain didn’t parse that David was talking about “history class”, so my mind went to the idea of art museums instead… Still seemed like a good vein, so I went with it)
I think what I’d have loved to see, and what would’ve made me learn a lot more and appreciate art a lot more, is a class where you watch an artist make a painting or sculpture from raw materials — maybe a reproduction of a masterpiece to avoid having kids not like it just because the end result is mediocre — to see how much work, skill, and attention-to-detail goes into it.
Even just watching little Tiktok videos by painters and artists is making me appreciate the craft more…
That’s certainly a big benefit of the internet, this ability to go peek inside studios and see what goes on behind the curtains. You learn a lot more than you could by just looking at the finished product — or maybe even are inspired to try it for yourself, since seeing someone else do it makes it inherently seem more achievable than to see a perfect masterpiece that seems impossible for anyone to have made.
✍ Two tricks I know for better writing: Assume the reader is smart, and be secure enough not to have to show-off how smart you are.
This way, you can write about interesting/complex things in simple ways, and everybody’s better off.
📜 There’s a habit I have while researching stuff online that some of you may be interested in duplicating:
When I read an interesting article, interview, or blog post, I’ll often save the highlights I want to remember/find most useful in my notes (Obsidian FTW) along with the URL of the source.
But because this is all just ephemeral bits flying around, hosted on any random server, you never know if that original source will still be there in a few years when you’re reviewing your notes and may want to go back to the original.
So I have an Internet Archive browser extension (y’know, the Wayback Machine at Archive.org), and I check to see if a copy of the page is saved in the archive. If it isn’t, I can hit a couple buttons and submit that link so that it is indexed.
I must’ve added thousands of pages that weren’t archived over the years. It’s good for me because I may need them later, but it’s also a public service since I act as a human webcrawler 🦀
💚 🥃 Been a while since I posted a stats update on this project, so for the charts-obsessed (you know you are):
Recently passed 4,300 total subs (3.7% paid/96.3% free — I’m still aiming for 5%/95%, I think I’d be happy with that, it looks like a sustainable win-win-win configuration).
The line is still pretty straight, and going up, so I can’t complain. Not seeing any reverse cup handles or double head & shoulders-with-dandruff patterns, so that’s probably good…
Oh, and a warm ‘welcome’ to recent members, welcome to the club! 👋
There’s a big pile of stuff in the archive over here if you’re curious to get more context on some of the running-jokes (ie. what this emoji 🛀 means around here).
And of course, if you like this project, feel like you’re getting value from it and want it to keep going, you’ll achieve that goal + a warm feeling if you buy me a virtual scotch here, it takes 20 seconds:
Investing & Business
Bezos’ Two Buckets
Jeff Bezos placed prospective business opportunities into one of two buckets. There were land rushes, when the moment was ripe, rivals were circling, and Amazon had to move quickly or else it would lose out. Then there was everything else, when the company could bide its time and patiently experiment.
Source: Brad Stone in Amazon Unbound.
Nvidia is a 28-years-old company that grew revenue 84% and GAAP earnings 106%.
Oh, did I forget to mention? This is while they’re supply-constrained…
Cloudflare’s Ambition, Cornerstone Decision + Downstream Effects
Lack of thinking big has never been a problem at Cloudflare:
Q: can you kind of briefly explain why your business is not a content delivery network company like Akamai or at least the part of Akamai that's not -- that's CDN.
Matthew Prince: from the beginning, almost 12 years ago now, when Michelle and I started working at Cloudflare, what we saw as the opportunity was to really replace all of the network security hardware that used to exist, so from companies like Cisco and F5 and Riverbed and Checkpoint. And the same way that AWS is the world we're replacing, what EMC and Dell and HP sold and the same way that Salesforce was replacing what Oracle and SAP and Microsoft sold, we thought there was just a huge opportunity to replace those boxes.
Prince then explains how the flat-rate pricing that they’re known for came about:
if you got a large cybersecurity attack, you didn't get a bigger bill from Cisco. And so from the beginning, we -- if we were replacing boxes, we couldn't bill based on use because it would just be gross. If it sort of starts to be something almost extortionary that you get a better attack and you get a big bill from your security vendor. [...]
so we started with the basic idea that if we're replacing the hardware, we needed to bill for it in a very predictable and reliable way. And I think that then led to a million other decisions.
I love this kind of stuff where a big cornerstone decision upstream means that a zillion other things are happening differently downstream.
This is the stuff where founder-quality makes all the difference, because these early decisions are very hard to undo, and their first-second-third-etc-order effects get embedded in every aspect of a company’s organization, physical and cultural.
Case in point:
It meant that we had to — instead of just buying bandwidth for x and selling it for x plus some percent, we instead needed to find ways to drive the cost of bandwidth to close to 0 over time.
Instead of deploying different hardware to service different functions, we needed to make it so that our software-defined network could allow any server anywhere in our network to run any different function that we did and to be able to shift traffic around to deliver a level of efficiency that no one else has.
I wrote a bit about this stuff in edition #133, about Cloudflare’s culture of frugality and innovation. This is what it’s driven by.
Prince on the recent Biden executive order on cybersecurity:
the Biden administration's executive order saying that if you're doing business with the federal government, you have to adopt a zero trust approach to security. And reading it was like reading the Cloudflare product catalog
Good travel analogy, helpful to visualize the role that various companies play:
we actually think about ourselves as a networking company. So the reason that we picked the ticker symbol NET was because, fundamentally, what it is that we sell is the network that you plug into and then you don't have to worry about anything else. [...]
maybe the best analogy would be if you imagine in our current pandemic world that you fly into a country and you go to show up at the border, and at the border, you have to present your passport, and that might be presented -- the passport might be issued by Okta or Ping or Microsoft Active Directory. You have to do a COVID test where they take your temperature, maybe even swab nose. And that's sort of like -- that's the device posture and that's end point security. So that might be provided by CrowdStrike or VMware Carbon Black or a Symantec or others. But there's a border agent there that then takes the data from Okta and takes the data from CrowdStrike and maybe does some other things to assess whether you're going to be led in or not, and that's the role that we play. We're the border agent that is connecting you from where you were to where you're going. And that, I think, is what we are really good at.
I also liked this overview by Prince of what makes for a great subscription business and how Cloudflare ranks on each:
great subscription businesses do four things well:
They have good unit economics. And I think our gross margins are -- we really designed the business to have good unit economics.
They have low cost of customer acquisition, and we've been in top decile in that.
They have low gross customer churn, and we've been very, very strong on that.
And then they have really great expansion rates. And historically, if we had a weakness, I think it was actually that fourth bucket where expansion wasn't our strength, and part of that was because we didn't have usage-based pricing. So there wasn't sort of just a natural expansion lever that was for that.
Ok, I don’t want to keep going on this forever, but this bit on competition from Amazon was great:
Q: what could Amazon do that could become a competitive threat to Cloudflare?
A: I think Amazon is the one company that really keeps me up at night. And if there's any CEO speaking at this conference, and they don't say that Amazon is the company keeping them up at night, then they're not paying attention because Amazon's ambitions are boundless.
I think that we go into everything we do with the assumption that whether it's Amazon or Microsoft or Google or any of the sort of hyperscale public cloud, there's no line of code that we can write that is so clever that it gives us a long-term durable advantage. [...]
as you sort of think about what are the things that we can do that Amazon can't do, the list is -- at some level, it's not very long. But I do think that it's -- there are some things that are really important. [...]
Foundationally, what Cloudflare is, is the network that can sit in front of everything. And that's different than what the hyperscale public cloud providers are providing. And so I actually think that if you play this out over the long term, that it could be that we end up being the network that carries a lot of traffic for the big public clouds.
And it could be that we increasingly are almost a vendor to those large public clouds as they're trying to figure out how do I connect these various pieces together.
And so I think that neutral network that can sit in front of anything is very different than the model of what the big public clouds are providing. And I think it is the reason why we do have strong partnerships with Microsoft and Google. But we have a little bit of a tense relationship with Amazon.
They even have a Bezos rule for APIs:
what I'm the most excited about for Workers is that Workers is the engine that allows Cloudflare's own development team to run as fast as we run because we build our products on workers. [...] we also have a rule internally that any API or any development tool we built for ourselves, we also release to our customers, unless both our CTO and I sign off on us making them private, which we almost never do.
🍪 if you’ve read all this. Gotta do that positive reinforcement.
EVs: Ford is Stepping on the Gas.. errr
We’re going to need new metaphors that produce less CO2. Anyway, Ford is accelerating its transition to EVs:
Ford Motor expects electric vehicles to make up almost half of its global sales by the end of this decade under the company’s latest turnaround plan that includes increasing its investment in EVs to more than $30 billion through 2025, the company said Wednesday. [...]
The increased investment in EVs is up from $22 billion that the company announced in February. Of which, about $7 billion had already been invested since 2016. (Source)
Well, hopefully next year they increase it by another $10-15 and just keep going until they are a leader instead of a late follower ¯\_(ツ)_/¯
Snowflake’s new HQ is in Montana?! Oh wait…
The company isn’t kidding when they say that they’re cloud-native! Even the offices and HQ are in the cloud!
The earnings press-release yesterday started with the words:
They explained in a SEC filing:
We are a Delaware corporation with a globally distributed workforce and no corporate headquarters. Under the Securities and Exchange Commission’s rules, we are required to designate a ‘principal executive office.’ For purposes of this report, we have designated our office in Bozeman, Montana as our principal executive office, as that is where our Chief Executive Officer and Chief Financial Officer are based.
So basically, from a legal point of view, Snowflake’s HQ is basically Slootman’s house. 🏠 He’ll probably get more junk mail because of this…
❄️ Compression and Chips
Also interesting, from the Q1 transcript:
this is something the company has historically done about every 2 years. There's a big focus on new compression technology for storage. And the impact of it was bigger than we would have thought. And we only knew that once we actually got real live examples from customers. And our philosophy has always been to pass that on to customers. But there's other performance improvements as well. For instance, we're working on a new chip technology that will dramatically increase performance or improve performance. [...]
what I will say is it does improve margins. And the way it improves margin is because storage becomes more efficient. Storage is a smaller component of the overall mix of the revenue, and compute is the real value of our software that drives more margin.
This is interesting, I wish I had the technical details on what kind of compression they use and how it’s implemented, and how they price the compute used for the compression/decompression itself (which I guess is still less expensive than the storage space would be, so customers save, but Snowflake gets more high margin revenue even if it gets fewer total absolute dollars of revenue). And since customers are saving some money on storage, many will just use those dollars from their budget to run extra compute, also beneficial to Snowflake.
On the chip stuff, this phrasing makes it sound like they’re designing their own chip in-house, but I don’t believe that’s what it is. It’s just not their niche.
I suspect they mean that they’ve been porting the platform to use some of what the hypercloud vendors have, like AWS’ ARM Graviton chips, which should be less expensive per FLOPS than the more common x86 server chips from Intel and AMD.
We’ll probably learn more at their big conference that is being held (remotely) June 8-10.
Twitter’s Private Protocol
Who else but friend-of-the-show and Extra-Deluxe supporter (💚💚💚💚💚 🥃) Byrne Hobart, again:
Privatizing protocols is lucrative—Twitter could have been an RFC or an open-source product instead of a company, but the for-profit version of the protocol had access to more financial resources. But those benefits come at a cost; governments feel entitled to regulate the commons, and those regulations rarely reflect the norms of social media companies' home countries.
This is something that I remember John Gruber saying a few years ago (I may be misremembering, though), how Twitter is a bit like if the email protocol had been owned by a private company instead of being just an open-standard, and all the weirdness that results from this.
Science & Technology
Solar Power Forecasts Over Time, Great Chart Edition
Kudos to Jordan Schneider of ChinaTalk for finding an updated version of one of my favorite charts from a few years ago.
If you’re more from the finance side, you may be familiar with the “hairy” charts that show predictions of interest rates going forward at the time vs what actually happened. For example:
There’s a similar one for solar power, which shows the various forecasts made by the International Energy Agency (IEA) vs what actually happened. Here’s the updated version:
(here’s the previous one, which went to about 2016)
It’s the whole “you have to compare your theory with reality once in a while” thing.
Maybe revise your methodology after a while of getting it this wrong, y’know?
Valve Building a Switch-like Handheld Gaming Device?
Video game and hardware studio Valve has been secretly building a Switch-like portable PC designed to run a large number of games on the Steam PC platform via Linux—and it could launch, supply chain willing, by year's end. [...]
The "SteamPal," whose name we're putting in scare quotes because we do not have confirmation of the device's final name, is an all-in-one PC with gamepad controls and a touchscreen. In other words, it looks and functions like a Nintendo Switch (albeit without removable "Joy-Con" controller functionality). [...]
with a system on a chip likely coming from either Intel or AMD, not Nvidia. (Source)
Meanwhile, Nintendo appears to be about to come out with an upgraded Switch that has a nicer OLED screen and a faster Nvidia SoC that can output 4K in TV-mode.
Microsoft on the latest cyberattacks from Russian group
This wave of attacks targeted approximately 3,000 email accounts at more than 150 different organizations. While organizations in the United States received the largest share of attacks, targeted victims span at least 24 countries. At least a quarter of the targeted organizations were involved in international development, humanitarian, and human rights work. Nobelium, originating from Russia, is the same actor behind the attacks on SolarWinds customers in 2020. These attacks appear to be a continuation of multiple efforts by Nobelium to target government agencies involved in foreign policy as part of intelligence gathering efforts. [...]
nation-state cyberattacks aren’t slowing. We need clear rules governing nation-state conduct in cyberspace and clear expectations of the consequences for violation of those rules. (Source)
Jim Keller Talk on Moore’s Law, Machine Learning (April 2020)
I don’t know why, but I guess I wasn’t paying as much attention to Jim Keller talks as I should’ve back in April of last year… ¯\_(ツ)_/¯
As Dylan Patel says:
Imagine being such a badass in your industry that you can use the Comic Sans font in a conference presentation and noone would bat an eye.
This tells you about half of what you need to know.
For the rest, you’ll have to watch!
I like what he says about the self-fulfilling nature of technological optimism and pessimism: “If everybody believes [the tech] wasn’t going to move [forward], it won’t”.
P.S. I thought this slide was funny.
The Arts & History
‘Kids playing Minecraft for the first time’ by Jesse Martin
The Office 📺
I hope you’re following this newsletter for cutting-edge cultural commentary on the most recent shows like Mad Men, Deadwood, and The Office.
Maybe next I’ll do MASH…
Anyway, I started watching ‘The Office’ for the first time ever (with my wife).
It felt like a hole in my culture, and smart pieces based on the show like VGR’s Gervais Principle and Alex Danco’s Michael Scott Theory of Social Class made me want to watch it more so I could better understand the references...
So far we’ve seen 4 episodes of season 1 (of the US version — I know the UK version is also highly regarded). I don’t really have much of an opinion yet, taking my time and letting it reveal itself a bit more. I’ve also been told by many that S1 is kind of like a long pilot, and that I should wait until S2 to get a real feel for it.