119: Nvidia: Inside Jensen Huang’s Brain, Stripe Financials, Discord & Microsoft, John Arnold on Renewables, Boyd Group, NASA, COVID19 Variants, and Torches of Freedom
"He’s basically Oprah... and Steve Jobs"
False opinions are like false money, struck first of all by guilty men and thereafter circulated by honest people who perpetuate the crime without knowing what they are doing
—Joseph de Maistre, Les soirées de Saint-Pétersbourg
🐕 Apparently the expression “eating your own dogfood”, often abridged to just ‘dogfooding”, may come from the fact that “each year the president of Kal Kan Pet Food would eat a can of the company's dog food at the annual share-holders' meeting.”1
It reminds me of the very possibly apocryphal (even if Taleb popularized it — see here) story about roman engineers having to stand under arches/bridges when the scaffolding was removed at the end of construction. Here’s Robert Dewar’s quote form 2004, which seems to be an early version of this:
When Roman engineers built a bridge, they had to stand under it while the first legion marched across.
Anyway, even if it’s made up, the idea of dogfooding/standing behind (or under!) your product is still powerful.
So often when I’m using crappy software, I wonder if anyone who’s designing and making it has to use it at all.. It feels like a lot of obvious issues couldn’t remain for long if engineers were facing them daily.
↬∞↫ Good post by Ava on ‘long feedback loops’:
I’ve been thinking about how the middle of making things often feels like incoherence. You’re far enough away from the beginning that you’ve lost the flush of newness, but you’re also very far from the end. The work feels muddied, unfinished, and possibly very bad.
What everyone tells you: keep going. But it’s not so easy, because the progress is no longer linear. You’re no longer seeing improvement day by day or even week by week. Maybe you’ve plateaued. Maybe you’re in a cul-de-sac and you should quit?
I tweeted recently about the problem with short feedback loops, i.e. things that provide instant validation. The more addicted you are to quick feedback, the harder it is to commit to things that require longer feedback loops. [...]
Mastery requires an ability to stick it out through loooong feedback loops.
It reminds me a bit of what I wrote a few years ago (and ‘remastered’ recently):
In the early part of the learning-curve, what you need to do to improve is well-known and improvement is rapid if you do it, so you get a juicy dose of reward every step along the way.
But as you get up higher on the learning curve, things tend to plateau, next steps are ambiguous or unknown — advice is contradictory, few people are able to provide it, and you have to untangle those from the fakers.
Feedback loops eloooongate and you spend seemingly interminable periods of time not knowing if you’re doing the right thing, going in the right direction, or even maybe getting worse…
I think this “being in the middle” happens both on shorter defined projects, and on lifetime pursuits. Maybe there’s no way to avoid it, but knowing about it probably makes it easier to endure. As Ryan Holiday said:
“The worst thing that could happen is not something going wrong, but something going wrong and catching you by surprise.”🎼 So what are we, music fans of the world, to do if both Spotify and Apple Music start to really suck at some point in the future? Centralizing all that music in a couple big players is neat when it works, but what happens when it doesn’t?
(some would argue they both already suck to various degrees and for different reasons, but what if both turn into the Oracle of music over the years..? That’s a scary thought…).
🤔 Know thyself. That’s a biggie.
I know I’m not very good at “pushing” to things, I’m more “pulled” by them. By that I mean that if someone tells me I have to study “X”, it’ll very likely seem boring to me and it’ll be a real slog.
But if I’m interested by “X” and the motivation comes from inside, I can spend days or weeks digging around and learning stuff.
It works a bit like this with companies. That’s why you’ll see me go through phases where I’ll write about, say, Cloudflare and Nvidia many times in a row. It’s because I’m learning about them lately.
I’ll move on to other things at some point, but I can’t just create an arbitrary schedule of companies to rotate through to I make sure I don’t bore those of you who happen not to be interested by what I’m rummaging through at the time.
Sorry, it’s just not how my brain works ¯\_(ツ)_/¯
🍎 Did you like the 𝕊𝕡𝕖𝕔𝕚𝕒𝕝 𝔼𝕕𝕚𝕥𝕚𝕠𝕟 #𝟛 on the Apple event yesterday? I’m thinking of doing single-topic extras once in a while. No big plan, just when I feel like it.
💚 🥃 Thank you to the new supporters — it really means a lot to me!
Don’t get me wrong, just giving me some of your precious time by reading is already a gift. But if you enjoy the show and want a stake in keeping it going, now’s the time to join:
Investing & Business
Nvidia: Inside Jensen Huang’s Brain
Huang recently did an interview with VentureBeat and hosted an investor day (with other execs). Here are some highlights:
Game development is one of the most complex design pipelines in the world today. I predict that more things will be designed in the virtual world, many of them for games, than there will be designed in the physical world. They will be every bit as high quality and high fidelity, every bit as exquisite, but there will be more buildings, more cars, more boats, more coins, and all of them — there will be so much stuff designed in there. And it’s not designed to be a game prop. It’s designed to be a real product. For a lot of people, they’ll feel that it’s as real to them in the digital world as it is in the physical world.
On the Grace CPU (well, more than a CPU — a whole system tying together ARM CPUs with GPUs and memory via lots of NVlink fat pipes, which are the magic dust making it all work — I wrote about this in edition #116):
Grace is designed for applications, software that is data-driven. AI is software that writes software. To write that software, you need a lot of experience. It’s just like human intelligence. We need experience. The best way to get that experience is through a lot of data. You can also get it through simulation. For example, the Omniverse simulation system will run on Grace incredibly well. You could simulate — simulation is a form of imagination. You could learn from data. That’s a form of experience. Studying data to infer, to generalize that understanding and turn it into knowledge. That’s what Grace is designed for, these large systems for very important new forms of software, data-driven software.
“Simulation is a form of imagination”. What a poet.
as a philosophy, we tend not to do anything unless the world needs us to do it and it doesn’t exist. When you look at the Grace architecture, it’s unique. It doesn’t look like anything out there. It solves a problem that didn’t used to exist. [...]
our primary preference is that we don’t build something. Our primary preference is that if somebody else is building it, we’re delighted to use it. That allows us to spare our critical resources in the company and focus on advancing the industry in a way that’s rather unique.
Above is a very important point that too many companies forget once they have huge resources, thousands of engineers and billions in the bank:
Focus. On what matters most.
Their approach to both integration & modularity (unlike what many think, it’s not a 1-dimensional axis — many dials you can tweak):
Our platforms are open. When we build our platforms, we create one version of it. For example, DGX. DGX is fully integrated. [...]
We take all of the building blocks, and we open it. We open our computing platform in three layers: the hardware layer, chips and systems; the middleware layer, which is Nvidia AI, Nvidia Omniverse, and it’s open; and the top layer, which is pretrained models, AI skills, like driving skills, speaking skills, recommendation skills, pick and play skills, and so on. We create it vertically, but we architect it and think about it and build it in a way that’s intended for the entire industry to be able to use however they see fit. Grace will be commercial in the same way, just like Nvidia GPUs are commercial. [...]
acquiring ARM is a very similar approach to the way we think about all of computing. It’s an open platform. We sell our chips. We license our software. We put everything out there for the ecosystem to be able to build bespoke, their own versions of it, differentiated versions of it. We love the open platform approach.
They created Grace not because they decided to check a box and have a “datacenter CPU in the lineup”, but because they saw important problems to solve that weren’t being addressed by existing products:
The world would like to train 100 trillion parameters on multi-modal data, looking at video and text at the same time.
For reference, GPT-3 has 175 billion parameters.
The journey there is not going to happen by using today’s architecture and making it bigger. It’s just too inefficient. We created something that is designed from the ground up to solve this class of interesting problems. Now this class of interesting problems didn’t exist 20 years ago, as I mentioned, or even 10 or five years ago. And yet this class of problems is important to the future. AI that’s conversational, that understands language, that can be adapted and pretrained to different domains, what could be more important? It could be the ultimate AI. We came to the conclusion that hundreds of companies are going to need giant systems to pretrain these models and adapt them. It could be thousands of companies. But it wasn’t solvable before. When you have to do computing for three years to find a solution, you’ll never have that solution. If you can do that in weeks, that changes everything.
On security:
security is a real concern today. The reason for that is because of virtualization and cloud computing. Security has become a real challenge for companies because every computer inside your datacenter is now exposed to the outside. In the past, the doors to the datacenter were exposed, but once you came into the company, you were an employee, or you could only get in through VPN. Now, with cloud computing, everything is exposed. [...]
you have to do security at the node. [...] There’s no way to do this on the CPU because it takes too much energy, too much compute load. People don’t do it. I don’t think people are confused about what needs to be done. They just don’t do it because it’s not practical. But now, with BlueField and EGX, it’s practical and doable.
Accelerators everywhere. He’s basically Oprah going: “You get an accelerator, you get an accelerator, you get an accelerator!”
Computer graphics, machine learning code, encryption and the inspection of large volumes of traffic flying through datacenters…
All that stuff has been offloaded from general-purpose CPUs to Nvidia’s more specialized chips (usually mixes of massively more parallel, with very fast I/O, with some custom silicon for specialized tasks, running on their own software/APIs that are optimized for the tasks..).
What used to be a whole bunch of electronic devices — we can now virtualize it, put it in the cloud, and remotely do computing. All the dynamics we were just talking about that have created a security challenge for datacenters, that’s also the reason why these chips are so large. When you can put computing in the datacenter, the chips can be as large as you want. The datacenter is big, a lot bigger than your pocket. Because it can be aggregated and shared with so many people, it’s driving the adoption, driving the pendulum toward very large chips that are very advanced, versus a lot of small chips that are less advanced. All of a sudden, the world’s balance of semiconductor consumption tipped toward the most advanced of computing.
Interesting insight about the dynamic pushing for these ever-larger chips (even as the etching nodes are shrinking, they’re cramming more transistors in them much faster than they shrink).
Huang adds that the fabs are adapting to this, and that he “doubts it will be a real issue in two years”. So don’t expect a quick fix for the supply issues next month…
On deep learning:
Ten years ago someone asked me, “Why are you doing all this stuff in deep learning? Who cares about detecting cats?” But it’s not about detecting cats. At the time I was trying to detect red Ferraris, as well. It did it fairly well. But anyway, it wasn’t about detecting things. This was a fundamentally new way of developing software. By developing software this way, using networks that are deep, which allows you to capture very high dimensionality, it’s the universal function approximator. If you gave me that, I could use it to predict Newton’s law. I could use it to predict anything you wanted to predict, given enough data. We invested tens of billions behind that intuition, and I think that intuition has proven right.
That’s some powerful stuff.
I believe that there’s a new scale of computer that needs to be built, that needs to learn from basically Earth-scale amounts of data. You’ll have sensors that will be connected to everywhere on the planet, and we’ll use them to predict climate, to create a digital twin of Earth.
🌏 🌏
GPUs were once a zero billion dollar market. That’s Nvidia’s style. We tend to go after zero billion dollar markets, because that’s how we make a contribution to the industry. That’s how we invent the future.
Man, he just keeps spitting those out. He’s probably the closest thing we have to Steve Jobs right now when it comes to clarity of vision of what this industry is actually all about (watch Jobs’ ‘bicycle for the mind’ analogy , it’s from an interview in 1990, it’s just 1-minute long and you won’t be disappointed):
So basically, I guess I’m saying he’s both Oprah and… Steve Jobs?
¯\_(ツ)_/¯
Huang also provides some of the high-level thinking behind the DPU/Bluefield. It’s a bit long and nerdy, but it’s a powerful vision:
Number one, every single datacenter in the world will have an infrastructure computing platform that is isolated from the application platform in five years. Whether it’s five or 10, hard to say, but anyway, it’s going to be complete, and for very logical reasons. The application that’s where the intruder is, you don’t want the intruder to be in a control mode. You want the two to be isolated. By doing this, by creating something like BlueField, we have the ability to isolate.
Second, the processing necessary for the infrastructure stack that is software-defined — the networking, as I mentioned, the east-west traffic in the datacenter, is off the charts. You’re going to have to inspect every single packet now. The east-west traffic in the data center, the packet inspection, is going to be off the charts. You can’t put that on the CPU because it’s been isolated onto a BlueField. You want to do that on BlueField. The amount of computation you’ll have to accelerate onto an infrastructure computing platform is quite significant, and it’s going to get done. It’s going to get done because it’s the best way to achieve zero trust. [...] That journey requires a reinvention of the datacenter, and that’s what BlueField does. Every datacenter will be outfitted with something like BlueField.
I believe that every single edge device will be a datacenter. For example, the 5G edge will be a datacenter. Every cell tower will be a datacenter. It’ll run applications, AI applications. [...]
Every single car will be a datacenter. Every car, truck, shuttle will be a datacenter. Every one of those datacenters, the application plane, which is the self-driving car plane, and the control plane, that will be isolated. It’ll be secure.
How’s that for a comprehensive view of where things are going?
Anyway, I guess I accidentally lied in the intro of this section. I think I’ll keep the investor day stuff for another edition, this interview has taken up all the space.
Of course I could go back and edit the intro, but it’s more fun to leave that as a teaser for next time! I know, I’m baaaaaad.
‘Discord Ends Deal Talks With Microsoft’
Microsoft had been in advanced talks to acquire Discord for at least $10 billion… Those talks ended without a deal, though it is possible they could be rekindled in the future [...]
In a move that could facilitate an IPO, Discord last month hired its first finance chief, Tomasz Marcinkowski, a former Pinterest Inc. executive. (Source)
Some Stripe Financials 💰
Everybody’s waiting for Stripe to go public, it seems. We don’t know that much about what’s going on under the hood, but this recent WSJ piece had a few numbers:
Stripe’s revenue last year rose nearly 70%, to about $7.4 billion, according to people with knowledge of the company’s finances. [...]
Stripe is laying the groundwork for an initial public offering it could pursue in late 2021 or early 2022, according to people familiar with the matter. [...]
Stripe operates in 44 countries and multiple currencies [...]
Excluding what it pays other financial partners to facilitate payments, Stripe generated $1.6 billion in 2020 revenue
h/t Tanay Jaipuria
John Arnold on ‘Seismic shift’ in Energy Industry
The man knows a thing or two about this space:
There’s been seismic shift in the Houston energy industry of late. A year ago, there was a lot of defending the oil & gas sector and denouncing renewables. Anecdotally, about 75% of the talk was O&G and 25% clean energy. It feels like those numbers have reversed.
Three-quarters of discussions now are about wind, solar, batteries, transmission, lithium, cleantech, etc. Even those who are not ideological believers are taking the cues from the financial markets, which have no interest in oil production growth anymore.
The markets are rewarding those in a growth industry (zero carbon energy) vs one in secular decline. Plus, the capital available to O&G has dried up. Every energy PE firm in town is raising $ for clean energy (good luck pitching an oil fund to a university endowment now).
The shift has made me more optimistic about future of Houston and speed of decarbonization. The latter requires enormous scale and financial resources that large companies possess. The fossil fuel industry has that expertise and is now focusing on a low carbon future. (Source)
Boyd Group 101 (screeeeech… !BANG!)
Good overview of the Canadian collision repair company by friend-of-the-show and supporter (💚 🥃) YoungHamilton:
It’s an interesting company that has figured out what works for them — a mix of operational excellence and disciplined and repeatable M&A — and just keep cranking on that process (to great long-term returns).
Similarly to Constellation software, they have figured that if you can get really good at M&A, it’s possible to take a bunch of small businesses that are pretty good on their own, and make them much better as an aggregate group because they go from having no real re-investment opportunity for the cash they generate to having a high-ROIC one. This changes everything for both VMS cos and collision repair shops (I’m not equating the two verticals here, btw, just saying there’s a similarity in the model).
High Quality Classifieds, Analyst at NYC Investment Manager Edition
Here’s a message from a friend-of-the-show and supporter (💚 🥃), Daniel Abood:
Kai Management Group is a growing investment manager based out of NYC
We are seeking a generalist analyst to perform research related activities on businesses and sectors in addition to pitches and other general analytical work
The analyst will work closely with the portfolio manager and other analysts on the team
Competitive compensation offered
Requirements include:
Candidate must have a passion for analyzing businesses (don’t fake it, we’ll know!)
Investing principles must be intrinsic value based
Been in a research role, preferably buy side for at least 5 years
Excel skills (the less you use the mouse, the better, right?)
If that's you, contact Daniel here. If that’s not you, but it may be someone you know, please send them this info! Thanks
Science & Technology
Media: Are SARS-CoV-2 Variants Evading Vaccines?!? (PANIC PANIC PANIC EVERYONE, BUT FIRST, CLICK ON THIS HEADLINE AND LOAD OUR ADS)
Once again, Zeynep Tufekci provides some much needed.. intelligence to this coverage:
For background, vaccines mostly have what’s called “breakthrough” infections: people who get infected or sick despite vaccination. Some of the plausible mechanisms for this are weak immune systems (something that happens especially in the elderly), waning of immunity over time (that’s why it’s recommended that adults repeat, for example, tetanus vaccines every ten years), or, obviously of great interest right now, because the pathogen mutated and was able to circumvent some immune system defenses to some level. But as discussed before … the immune system is not a seawall that, once breached, is gone. Even when a breakthrough does happen and cause symptoms, the case might be much milder, not progressing to severe disease as other parts of the immune system are called in.
you know what the paper does not say, imply, or hint at? There is nothing in the paper to suggest the “South African variant may evade protection from the Pfizer vaccine,” as Reuters or many other news organizations claim, beyond what we already know: there will be a few breakthroughs even with the mRNA vaccines, because that’s just reality with almost all vaccines, and it’s possible more of those breakthroughs will be the South African variant because of its effect on neutralizing antibodies. [...]
what researchers mean with “evade” when they use that term is not “leave without any protection,” which is what all the headlines imply and what caused the panic. Full vaccine escape, where vaccinated individuals show no difference from people for whom this virus is completely novel, would certainly be alarming, and necessitate vaccinating everyone, again, with a new vaccine. [...]
Here’s a good rule. The first question, always, should be: did the reporter read and understand the methods section? If the answer is not yes, maybe it’s not necessary to write that piece or that headline, yet. (Source)
From the Wright Brothers to a Helicopter on Mars
I’m sure you’ve heard about the helicopter that they flew on Mars.
Commenters on Reddit put it in ways that drove home the speed of progress and the unique challenges of the task:
Mans first flight was a little over 100 years ago.
Now we’re flying on Mars.
Ingenuity contains fabric from the first successful plane the Wright brothers flew.
At 1% of the earth's atmosphere! It's amazing this could even been done!
Only 1% the atmosphere, but also only 1/3 the gravity
How we know (for sure) that bees perceive time
Still can’t embed Tiktok videos, so you’ll have to click to see it:
It’s just a couple minutes, on all the successive experiments that were required to confirm that bees can perceive time (cool example of how you need to isolate variables to make sure you’re really measuring what you think you’re measuring).
The Arts & History
Locorotondo, Italy
population of about 14,000. The city is known for its wines and for its circular structure which is now a historical center, from which derives its name, which means "Round place" [...]
The site has been settled since ancient times, as testified by archaeological finds dating between the 3rd and the 7th century BC. The foundation of the town dates back to around 1000 AD as an unfortified hamlet under the jurisdiction of the Benedictine monastery of St. Stephen in Monopoli. The estate of various feudal lords for 500 years, it saw an increase in population, housing development, and the construction of the walls and castle. (Source)
‘Torches of Freedom’ (Advertising Cigarettes to Women)
Teachable historical moment, showing how much framing and marketing matters, and how almost anything can be hijacked by something else for its own benefit:
Before the twentieth century smoking was seen as a habit that was corrupt and inappropriate for women. [...] Women's smoking was seen as immoral and some states tried to prevent women from smoking by enforcing laws. In 1904 a woman named Jennie Lasher was sentenced to thirty days in jail for putting her children's morals at risk by smoking in their presence and in 1908 the New York City Board of Aldermen unanimously passed an ordinance that prohibited smoking by women in public. [...]
"Torches of Freedom" was a phrase used to encourage women's smoking by exploiting women's aspirations for a better life during the early twentieth century first-wave feminism in the United States. Cigarettes were described as symbols of emancipation and equality with men. The term was first used by psychoanalyst A. A. Brill when describing the natural desire for women to smoke and was used by Edward Bernays to encourage women to smoke in public despite social taboos. Bernays hired women to march while smoking their "torches of freedom" in the Easter Sunday Parade of 31 March 1929, which was a significant moment for fighting social barriers for women smokers. (Source)
Very Don Drapper.
h/t Gregory Doyle
Oh yes, I’ve got a footnote for this! Check it out. h/t Tanay Jaipuria on this vital piece of information.