321: Thoughts on Snowflake Q2, TSMC's Electricity Bill, Good & Bad Kilowatts, Desalination, Directionless Thinking, Status of Actors vs Characters
"this abundance would create abundance in other areas."
The ship wherein Theseus and the youth of Athens returned from Crete had thirty oars, and was preserved by the Athenians down even to the time of Demetrius Phalereus, for they took away the old planks as they decayed, putting in new and stronger timber in their places, insomuch that this ship became a standing example among the philosophers, for the logical question of things that grow; one side holding that the ship remained the same, and the other contending that it was not the same.
—Plutarch, Life of Theseus
🛀 🪧⛔️ While driving to our vacation Airbnb, I thought about how many things have names that I don’t know the origin of — and maybe almost nobody knows either.
Some names have famous origins, like the city of Washington or whatever, but in every area, there are thousands of named things — streets, parks, squares, schools, libraries, rivers, creeks, lakes, mountains, caves, canyons, ravines, deserts, forests, fields, meadows, etc.
Some are even in languages that are not commonly spoken in the area anymore — from the natives who were displaced, or from some group of immigrants that came first but were later outnumbered by other groups of immigrants from elsewhere — there are also plenty of references to other countries (lots of European names can be found in North-America).
I sometimes wonder if I picked one of these names at random (a non-famous one) and tried to figure out its origin, what would I find? A really cool local story? Some bored bureaucrat picking names out of a hat? That nobody even remembers…?
💚 🥃 I'm thinking of it kind of like we're sitting in a pub somewhere, and I'm talking about various things that interest me or that I've learned about recently.
If you like it, once in a while, you send me a scotch to keep me talking and show appreciation. That’s what becoming a paid supporter is. All very casual & civilized. And a jolly g'day to you too, mate. 🍻
Liberty’s Highlights is reader-supported. To support my work, consider becoming a paid supporter. 🦑
A Word From Our Sponsor: 📈 Revealera 📊
Revealera provides data and insights for investors into hiring trends for 3,500+ public/private companies + technology popularity trends for 500+ SaaS/Cloud Products.
We give investors insights into:
Job Openings trends: Insights into a company’s growth prospects.
Technology Popularity Trends: Insights into how widely products like Datadog, AWS, Splunk, etc, are gaining adoption.
Vendor Sign-ups (Currently Alpha) tracks the # of companies, as well as the specific companies, that have signed up for SaaS products such as Zoom in near real-time.
Visit Revealera.com for a ✨free✨ trial/demo.
🏦 💰 Liberty Capital 💳 💴
🇹🇼 TSMC to use 1/8th of Taiwan’s Electricity 🔌⚡️
Extreme Ultra-violet lithography uses a lot of power:
Each machine is rated to consume about 1 megawatt of electricity, about 10 times more than previous generations of equipment.
The chips are getting smaller and more energy efficient, but the machines making them are getting bigger and more power-hungry!
For Taiwan, this isn’t just a marginal change, this is a magnitude that registers on the scale of the whole country:
No one has purchased more EUVs than TSMC, the world’s largest supplier of outsourced chips. It currently has more than 80 and is in the midst of installing a new generation of the machines as part of a $20 billion chip foundry in Tainan, a city in southern Taiwan. Because of the vast amount of power needed to run EUVs, TSMC is expected to soon consume more energy than the entire 21 million-person population of Sri Lanka. In 2020 the company accounted for about 6% of Taiwan’s overall energy consumption. It’s expected to use 12.5% of it by 2025.
Some fear power shortages, new power plants will have to be built. Taiwan is largely powered by coal and gas:
Micron is going EUV too, which will boost its energy use (in Taiwan and elsewhere):
In addition to TSMC, Micron, a major memory chip producer, plans to use at least one EUV at its production site in Taichung, about a two-hour drive south of Taipei, the Taiwanese capital. Within three years, a quarter of all chip fabrication in Taiwan’s foundries will need EUVs to function
❄️☃️❄️ Snowflake Q2 🧊 🥶 🧊
A quick look at the quarter’s numbers:
Product revenue of $466.3 million, representing 83% year-over-year growth
Net revenue retention rate of 171% (and that’s with the recent efficiency optimizations that are a headwind to revenue, 2 quarters ago they were at 176%!)
Non-GAAP gross margin of 75%
That was 69% two years ago! That’s quite the leverage considering they’re building on top of hyperscaler infrastructure.
During the same 2-year period, FCF margin went from -12% to +26% (YTD).
Let’s look at the call to see if anything interesting was said:
The core idea behind the Snowflake Data Cloud is to enable work to come to the data and prevent data from having to be moved to the work. Previously, data was copied, transferred and replicated to be used wherever it was processed.
This is a simple, but powerful idea.
It’s a lot easier to move the relatively tiny amount of code that needs to be run to do compute than to bring gigantic quantities of data over to that code (this also has positive security implications, since data is easier to secure if you don’t copy is around too much 🔐).
A metaphor: If you want a photo of the Taj Mahal, it’s easier for you to go to it with your camera than to stay where you are and have the Taj Mahal moved to you. 📸
In Q2, the number of Snowflake data sharing relationships measured with what we call stable edges grew 112% year-on-year. 21% of our growing customer base has at least 1 stable edge, up from 15% a year ago. [...]
data sharing is incredibly important. Every industry and sub-industry has its own unique data networks and their own reasons and use cases where we know why they need to share data.
In financial services, for example, which is, of course, an industry that you're all very familiar with, I mean, the need for sharing is extremely pronounced. And it's a daily preoccupation between institutions that need to share data. So Snowflake has really become a de facto platform for financial institutions on how to share data. So that becomes a very, very powerful thing. We feel the network effect from data sharing in certain verticals that are really more advanced, more mature in terms of the adoption of data sharing than some others that are taking more time to get into that.
These ‘stable connections’ between Snowflake customers are a powerful thing. Some are selling access to data and others are paying for it, or companies that have existing partnerships and are operationalizing them on Snowflake.
They make the service stickier and more valuable to customers, creating a network effect where, over time, the more buyers and sellers there are, the more benefits an average member of the data exchange can reap.
Today, we have 590 Powered by Snowflake registrants, representing 35% quarter-over-quarter growth. [...]
People are building and deploying applications on Snowflake. I mean we have customers like Western Union who were starting whole businesses on top of Snowflake.
This is another way that Snowflake is making itself stickier than it would be as just a data lake.
If you actually build your application on top of it as a platform and unify transactional and analytical workloads (Unistore), Snowflake becomes at the very core, the engine of watever it is you’re doing.
Can you say mission-critical?
We continue to pursue a vertical sales strategy. Our core verticals are financial services, advertising, media and entertainment, retail and CPG, technology and health care and life sciences. While all verticals are growing rapidly, financial services drove the most product revenue growth sequentially. Advertising, media and entertainment and technology verticals grew in line with the overall company.
This vertical approach seems to be one of Slootman’s big changes over previous management, who positioned Snowflake more horizontally.
It seems to be paying off so far, and makes sense in the context of the ‘Data Cloud’ model where customers can buy & sell data.
In any two-sided marketplace, the hard thing is to reach critical mass, and by targeting verticals it becomes easier to reach critical mass by industry, in a more focused way, rather than try to attack the economy as a whole.
I found it surprisingly how little non-USD exposure they had 💵:
less than 5% of our revenue is invoiced in currencies other than the U.S. dollar. So at the moment, we do not evaluate our business on a constant currency basis given the immateriality.
Asked how Snowflake would fare in a recession/bad economic environment:
In general, I would say that Snowflake gets prioritized fairly high inside the enterprise. And the reason is we're sitting right on the intersection of cloud computing, artificial intelligence, machine learning, advanced analytics. [...]
They're hiring. They're buying because these are things that are going to be big changes for the industry. The whole industries are going to get redefined in health care, in pharma and financial services.
So this is not a business-as-usual let's try to hit the brakes. There is a very, very high urgency around advancing towards cloud computing environments and then having an opportunity to really pursue the promise that it brings. We're looking at very, very exciting times where things are becoming possible that we couldn't even dream of just a few short years ago. So this is why we feel that this is not one of those expenses that people are going to casually cut back on because it's strategically compelling and important.
Of course, Frank’s not exactly a neutral observer in all this, but there’s probably some truth there.
The CFO also gave an important reminder about the model, which is different from SaaS:
this is a consumption model, not a SaaS model. So even if we sign up a customer, it takes some months before they go into production. [...]
So you can't just do a quarter-over-quarter. And because it is a consumption model, we do see patterns. Some customers do go down consumption based upon specific projects.
But by and large, most of our customers are still ramping, moving workloads to us. And we think that is going to continue on average with our customers. [...]
I don't even look at billings because it's a meaningless thing for our business because we're not a SaaS company. We're really focused more on cash flow and revenue consumption by our customers.
They’ve mentioned a few times before that it can take many many months (9+) for any new customers to get ramped up, so a lot of the revenue that is coming in today is from customers they signed in 2021 and 2020.
most SaaS companies, when they land an account, they typically license most of the users in an account. When we land, we land small. And they go workload by workload, and they just keep moving stuff over to Snowflake. That drives that.
And it's a multiyear journey within our customers. And I don't see any of our customers that are fully saturated, where I think some SaaS companies may be saturated.
This helps explain the best-in-class net revenue retention, which is likely to remain very high for a long time, at least until the business is much more mature.
the majority of our customers, 80-plus percent run in AWS, and about 18% is Azure and 2% is GCP.
To put this in context, they started only on AWS, so it would seem that Azure has been growing nicely for them.
From what I’ve heard, GCP seems less interested in partnering with Snowflake and more focused on competing with them directly, so they probably don’t offer them pricing that is as attractive in this particular instance.
We will continue to do optimizations. They're good for our customers because when they see if we can help make them use Snowflake more efficiently, we become cheaper, they move more workloads to us.
It’s a tangent, but I’m fascinated by compression algorithms.
I grew up using my father’s 386 DX/25mhz with 4 megs of RAM and a hard drive of something like 200 megabytes, so efficiency and compression was very much on my mind.
From zipping files and then RAR’ing them, discovering 7z later on and utilities that allow you to vary dictionary size and change the trade-offs… and then lossy algos for images, sound, and video (GIF, JPEG, PNG, HEIC, MP3, OGG Vorbis, AAC, FLAC, xvid, h.264, HEVC, AV1 etc).
I’m amazed that decades later, we’re still finding new ways to squeeze data ever smaller… though I guess we’re mostly trading upfront compute and RAM for file size. But still, I like the elegance of doing more with less. SNOW 0.00
💀 Lethal Combo 🪦
Friend-of-the-show and Extra-Deluxe supporter (💚💚💚💚💚 🥃) David Kim wrote:
extensive diligence + poor judgment is a lethal combo
Words of wisdom.
🧪🔬 Liberty Labs 🧬 🔭
⚡️ Good Kilowatts, Bad Kilowatts? 🤔 🤨
This sub-title is a play on ‘Good Calories, Bad Calories’, and the whole debate about whether all calories are the same. From one angle, yes they are, by definition. But from another, in the real-world, if you get all your calories from sugar, *all else may not be equal* and the huge spikes in insulin and low satiety probably won’t have the same result as if you get your calories from a more balanced mix of protein/carbs/fats… There are parallels with the grid.
There’s a crew of people in energy that love to compare all sources of energy as if they were purely commodities, totally fungible. A kWh is a kWh, right?
“How can we get the kWhs we need at the lowest cost? Let’s do that!”
But I think it’s *incredibly* important to take other factors into account. A power grid isn’t a casino game where the system is closed, the variables known, and the stakes limited to what players voluntarily bet.
A kWh that pollutes the air or where you have to pay a tyrant for fuel *isn’t* equivalent to a kWh that doesn’t pollute and where you can store years of fuel that you got locally or got from a friendly democracy. It’s also very different to have a kWh that is dispatchable whenever you need it vs a kWh that may not be there when you need it.
Let’s be smart about this, because everything else is built on top of energy.
Study on the value of ‘directionless contemplation or free-floating thinking’ 🤔💭🧠
We’re so afraid of what we think will be “boredom”, when in fact directionless thinking is often just as fulfilling and enjoyable as whatever BS we occupy ourselves with:
The ability to engage in internal thoughts without external stimulation is a unique characteristic in humans. The current research tested the hypothesis that people metacognitively underestimate their capability to enjoy this process of “just thinking.”
Participants (university students; total N = 259) were asked to sit and wait in a quiet room without doing anything. Across six experiments, we consistently found that participants’ predicted enjoyment and engagement for the waiting task were significantly less than what they actually experienced. This underappreciation of just thinking also led participants to proactively avoid the waiting task in favor of an alternative task (i.e., Internet news checking), despite their experiences not being statistically different.
These results suggest an inherent difficulty in accurately appreciating how engaging just thinking can be, and could explain why people prefer keeping themselves busy, rather than taking a moment for reflection and imagination in our daily life. (PsycInfo Database Record (c) 2022 APA, all rights reserved)
‘Joy for environmentalists as California blocks bid for $1.4bn desalination plant’
One of the reasons why I think it’s so important to have really stable power grids with *lots* of reliable, clean power is because this abundance would create abundance in other areas. Not only for all kinds of economic activity, but for things like massive desalination projects in areas where fresh water is scarce.
You’d start by alleviating shortages, but over time, with enough supply, you can irrigate large areas and prevent some bad crop years when a drought would’ve destroyed lots of food (high food prices always affect the poor severely), and even eventually turn formerly fertile land that was lost to desertification back to its lush state.
Literally: Make the desert bloom. 🌵🏜💐🌿
But that’ll be hard if we can’t even take the first steps on this:
A California coastal panel on Thursday rejected a longstanding proposal to build a $1.4bn seawater desalination plant to turn Pacific Ocean water into drinking water as the state grapples with persistent drought that is expected to worsen in coming years with climate change.
The state’s Coastal Commission voted unanimously to deny a permit for Poseidon Water to build a plant to produce 50m gallons of water a day in Huntington Beach, south-east of Los Angeles. [...]
Poseidon’s long-running proposal was supported by Governor Gavin Newsom but faced ardent opposition from environmentalists who said drawing in large amounts of ocean water and releasing salty discharge back into the ocean would kill billions of tiny marine organisms that make up the base of the food chain along a large swath of the coast.
🎨 🎭 Liberty Studio 👩🎨 🎥
🎥 Actors vs Character, Status Games Edition 🎬
I’ve long found it interesting how actors who play high-status characters tend to get some of that status rubbing off on them, while actors who may be just as good at their craft, if not even better, but playing low-status characters will often get less recognition and won’t be seen as high-status.
In fact, it seems even worse if they are so good at their role that they fool the audience into thinking that they actually are as screwed up as their characters. Our brains just haven’t evolved to fully distinguish between acting and just being.
I’m sure there are exceptions, but it seems generally true to me.
I don’t know if there’s a name to this effect..? It sounds like something that would’ve been figured out by theater folks a long time ago.
Excellent point on the actors! Relatedly, I think ppl discuss sometimes how certain actors get typecast into certain roles, but it really does seem pretty extreme. How good of actors are they if they can't just show up and be told what role to play? I remember watching a bunch of Joaquin Phoenix movies in a row and realizing he seems pretty similar in each.
Maybe there need to be more movies where the actors don't even act that human at all. Instead of "getting in the head" of some kind of character, maybe just create a new imaginary "animal" and act that out? Maybe that's what star wars basically is?
I'm not in the movie industry at all and by no means a qualified critic. Just talking out loud..