266: Buffett Should Buy Texas Instruments Instead of HP, Berkshire + Microsoft Alternate Timeline, UK & South-Korea Nuclear U-Turns, Nvidia, and Bojack
"a lighthouse for like-minded people"
Every act of kindness is a rebellion against a cruel, indifferent world.
–Visakan Veerasamy
🗣🗣🗣 Poll results on Discord vs Slack so far:
I’m letting this ‘community treehouse’ (🏡 ha!) idea swirl around my head a bit…
I think it could be really cool, but it would have to be clear what it is, and what it *isn’t*, because there’s a failure more here where I accidentally create something that takes way too much of my time, and I don’t really have much spare time right now.
The #1 goal would be for readers to find each other — because you guys & gals are fantastic and I think you’d get along really well and enrich each other’s lives — not to create a direct line of communication to me where a kind of 24/7 ‘ask me anything’ would take place, with me expected to see and reply to everything with thoughtful and fully-fleshed out responses…
That’s just too much pressure and I just don’t have the bandwidth for that.
But as long as everybody understood that this would be a lighthouse for like-minded people — to use David Perell’s (✍️) turn of phrase — where I just also happen to hang out, and there are no expectations on anyone…
I think it could be pretty great and lots of fun. 🤔
🐦 Back when Twitter bought Revue and there was huge buzz about it as a Substack competitor I created an account to have a look.
The hot take I kept hearing at the time was “maybe they don’t have quite the features that Substack has, but their distribution through Twitter is a huge advantage and may help them win”.
To be honest, I had kind of forgotten about Revue since. It doesn’t look like Twitter pushed it much, and I almost never find newsletters that I want to read that are published on Revue (I think my only exception is for friend-of-the-show and supporter (💚 🥃) Bill Brewster’s nl).
I’m not sure what happened, but Revue certainly doesn’t seem to be a priority at Twitter. Substack has improved by leaps and bounds in the 20 months that I’ve been on it, and my Revue account seems exactly the same as when I first signed up. I’m sure they added things, but nothing jumps at me.
Probably another example of the power of focus: Substack’s whole purpose is to create a good home for writers. Twitter has 45 different priorities, and newsletters probably aren’t in the top 5.
Too bad, because I want to see Twitter succeed, and I’d like to see more competition for Substack — I think it would push them and increase the pace of innovation in the space generally.
Maybe if Elon Musk starts caring about newsletters something will happen…
¯\_(ツ)_/¯
🔥 🖥 ⌨️ I tried to stress-test the M1 Max in my Mac Studio by running some computational biology workloads on it, and it took anything I could throw at it like a champ, even x86 code ran fine through the Rosetta2 translation software (from x86 → ARM).
The best part is: Even when all 10 CPU cores are pegged to 100%, the fans don’t spin up and the air coming out from the vents in the back is barely warmer than usual.
That thing is ridiculously efficient!
Here’s a good ‘first impression’ review.
💚 🥃 If you make just one good decision per year because of something you learn here (or avoid one bad decision — don’t forget about preventing negatives!), it'll pay for multiple years of subscriptions (or multiple lifetimes).
And the first good decision is to become a supporter 😉
A Word From Our Sponsor: ⭐️ Tegus ⭐️
What makes a great investor isn’t pedigree or diplomas 🎓, it’s quality thinking and information 🧠
Tegus can’t provide the thinking for you, but on the information front, they’ve got you covered! 📚
With 25,000+ primary source expert calls covering almost any industry or company you may want to learn about, and the ability to conduct your own calls with experts, if you so choose, at a much lower cost than you’ll find elsewhere (70% lower than the average cost of ˜$400), this is the service to turbo-charge the depth and breadth of your knowledge 📈
⭐️ Get your FREE trial at Tegus.co/Liberty ⭐️
Investing & Business
Instead of investing in IBM 2.0 HP, Buffett should invest in… Texas Instruments 🤔
With the caveats that Buffett is much smarter and a much better investor than I am, that I’m not following HP (because 🥱), and that this isn’t really a pitch but rather just a random musing:
I’m surprised that Buffett went for HP. Seems like another IBM, an old business that isn’t really riding any strong S-curves anymore and is trying to financialize its way to better stock performance (pulling a Bob Swan, if you will) rather than focusing on innovation and technical leadership to create tomorrow’s products and services that will replace today’s declining ones (in tech, things can change fast, and with mediocre businesses, the surprises tend to be negative and the declines tend to be non-linear and faster than you expect).
That said, there’s a price for everything, and maybe Buffett will do great on HP and it’ll turn out how he expected IBM to turn out originally (HP stock certainly has done better than IBM in the past decade)…
…but it made me think that there’s a business in the tech sector that seems a perfect fit for Buffett: Texas Instruments (disclosure: I own some).
It’s large enough to deploy plenty of capital into (one of Buffett’s biggest constraints these days — his universe of investable companies isn’t that large), has impeccable capital allocation and is a very durable business that is highly diversified by end-market and geography, has all kinds of tailwinds and competitive advantages, and a clear line of sight to highly profitable growth as far as the eye can see (they’re making big manufacturing investments right now that should pay off for a long time).
I’m sure Buffett could understand it if he wanted to, at least at a higher level, like he did with Apple (don’t need to know all the details of the M1’s on-die unified memory architecture to get it…).
Sure, it sells for a higher multiple than HP, but it deserves a much higher multiple than HP and it isn’t that expensive for its quality (around a market multiple).
If you’re not familiar with TI and what makes it different from most other semiconductor companies, this Chit Chat Money episode is a good 101:
But I guess if we’re discussing this, we can probably discuss why he never pulled the trigger on Google and Amazon (there were times in recent years when they got pretty cheap). In fact, this brings me to…
Speculation: What if Buffett hadn’t made Microsoft ‘off-limits’ because of his friendship with Bill Gates 🤔💭
It’s interesting to speculate that if Warren Buffett didn’t care about the appearance of conflict from buying Microsoft when his friend Bill Gates ran it, Berkshire would probably have bought 10%+ of Microsoft sometime in the 90s and held it until today, possibly buying more along the way during dislocations.
Who knows, maybe Berkshire would own 20%+ of Microsoft and that alone (including decades of dividends re-invested at Buffett’s long-term CAGR) would be worth a large fraction of today’s market cap, if not more than today’s market cap.
Interesting “what if”.
BP’s predictions for the cost of wind & Solar to 2050
This will probably turn out to be just as wrong and conservative as previous forecasts (see below), but if we put that aside for a moment, it’s still quite something to see that energy sources that are already quite cheap and have already fallen by 95%+ over the past few decades are projected to fall another 20-40% in the next 5 years.
If I had to bet, I’ll take the over on these numbers.
The question will be more about other things, like scaling up storage (as I previously posted about in edition #264, the cost of batteries is on a similar curve), building more nuclear for baseload, investing in energy efficiency so we stop wasting so much of the energy we produce, etc.
Just because it’s hilarious, here’s a chart showing actual solar PV additions (in black) vs the IEA predictions at various times:
Interview: Rob Koyfman, Koyfin CEO & Co-Founder
Supporters (💚 🥃) can listen to the new podcast here:
It’ll go out to everyone tomorrow.
Science & Technology
🇬🇧 ‘UK to build 8 nuclear reactors amid new energy strategy’ ☢️
Britain plans to build eight new nuclear reactors and expand production of wind energy as it seeks to reduce dependence on oil and natural gas from Russia [...]
The government said it wants to almost triple nuclear power generation capacity to 24 gigawatts by 2050.
Better late than never, I guess…
I’m seeing criticism of the plan, saying it has a bunch of long-term targets, but doesn’t do enough in the short-term to really make a difference, and doesn’t attack low-hanging fruits like efficiency (the country’s aging housing stock could use an upgrade).
🇰🇷 South Korea’s incoming government will reverse the country’s nuclear phase-out plan ☢️
South Korea has the highest density of nuclear reactors in the world, with most of its 24 reactors located at two complexes in the country’s industrial south-east.
In 2019, its reactors produced 139 terawatt-hours of electricity, making South Korea the fifth-largest nuclear power producer in the world. Nuclear accounted for 26 per cent of the country’s total electricity generation, according to the US Energy Information Administration.
Under Moon’s nuclear phaseout plan, construction of plants was to be suspended while older plants were to be retired, with the aim of reducing the number of operational reactors to 17 by 2034.
🤘
⛽️ It turns out that gas pump IT security isn’t very high… 🏴☠️
Thieves in Paris in 2019 took 26,000 gallons – yes, thousand – from multiple outlets and were believed to have made roughly $168,000 from their spree. In that case, the thieves brought in vans that carried tanks that could carry roughly 660 gallons of fuel. [...]
In June of 2017, a BP employee in New Jersey was arrested and charged with manipulating gas pump computers to steal more than $300,000.
Unlike the culprits in High Point, these thieves knew the access code for a specific brand of fuel pump, which allowed them to set the price at zero. (Source)
😬
some gas pumps were vulnerable to takeover by hackers because there was an embedded controller in that gas pump. “The device we investigated was not just a tiny web interface. It was an embedded box running a Linux-based controller unit that was installed with a tiny httpd server,” Ido Naor wrote for Kaspersky.
“According to its manufacturer, the controller’s software is a site automation device that is responsible for managing every component of the station, including dispensers, payment terminals and more. [...]
According to Naor’s work for Kaspersky, an intruder accessing a gas system could:
Shut down all fueling.
Cause fuel leakage and risk of casualties.
Change price.
Circumvent payment to steal money.
Scrape vehicle license plates and driver identities.
Halt the station’s operation, demanding a ransom in exchange.
Execute code on the controller unit.
Move freely within the gas station network.
Internet-connected gas pumps… What could go wrong?
This presentation from 2012 shows a bunch of other techniques to steal gas.
The IoT future is just getting started. We’re going to see all kinds of weird things going forward.
I hope we’ll start to take securing the more critical pieces more seriously…
Nvidia’s past 3 generations: Volta → Ampere → Hopper
AnandTech compiled the data above, which provides a nice way to compare the improvements made with each generation of recent Nvidia GPUs.
The piece goes into much more detail.
Here’s a highlight on the new in-silicon transformer engines that are now part of the ML Tensor cores:
In keeping with NVIDIA’s focus on machine learning, for the Hopper architecture the company has taken a fresh look at the makeup of the ML market, and what workloads are popular and/or the most demanding on existing hardware. The winner, in this regard, has been transformers, a type of deep learning model that have risen in popularity rather quickly due to their utility in natural language processing and computer vision. Recent advancements in transformer technology, such as the GPT-3 model, along with demand from service operators for better natural language processing, have made transformers the latest big breakthrough in ML. [...]
Combined with the additional memory on H100 and the faster NVLink 4 I/O, and NVIDIA claims that a large cluster of GPUs can train a transformer up to 9x faster [...]
NVIDIA is claiming anywhere between a 16x and 30x increase in transformer inference performance on H100 versus A100. Like their training claims, this is an H100 cluster versus an A100 cluster, so memory and I/O improvements are also playing a part here, but it none the less underscores that H100’s transformer engines aren’t just for speeding up training.
The Arts & History
🎶💥 Exploding some more songs… Billie & Bojack
While I’m on a song exploder kick, an episode that I liked was Billie Eilish and her brother discussing how “Everything I wanted” was created, from a suicidal fantasy/nightmare to a more hopeful second half written many months later:
I also greatly enjoyed this one on the creation of the very distinctive theme from one of my favorite shows in recent years, Bojack Horseman (you can listen/watch it at the top).
This was written and performed by Mark Carney and his uncle, Ralph Carney. Mark is the drummer of the Black Keys, and it’s really interesting to hear about how his uncle (who play with Tom Waits, among others) influenced his love of music, and how they have a long-distance, multi-decade collaboration, sending music back & forth.
I’d love to have a creative collaborator like that…
(btw, if you haven’t seen Bojack and decide to try, know that there’s a few episodes at the beginning of season 1 that aren’t representative of the heights that the show reaches.. I think it takes them about 3 episodes to find their feet)
I have seen communities built in Slack in the past (lucky they migrated away from it) and Discord. Slack has a 10k message limit on free communities, and messages are invisible (do not appear in search or you cannot navigate through them) after. It seems 10k limit is a lot, but in fact it is not. This is quite a huge technical limitation which makes Slack objectively worse IMHO.
This prediction chart really made me think; why do we even bother to do long term projections? It’s just fake comfort of knowing and thinking about where things are headed. Plus, I imagine an employee just amending his excel sheet a bit. ;)