We’ve constantly been sharing a list of our recent reads in our weekly emails for The Good Investors.
Do subscribe for our weekly updates through the orange box in the blog (it’s on the side if you’re using a computer, and all the way at the bottom if you’re using mobile) – it’s free!
But since our readership-audience for The Good Investors is wider than our subscriber base, we think sharing the reading list regularly on the blog itself can benefit even more people. The articles we share touch on a wide range of topics, including investing, business, and the world in general.
Here are the articles for the week ending 10 September 2023:
1. Rediscovering Berkshire Hathaway’s 1985 Annual Meeting – Kingswell
In the mid-1980s, Berkshire Hathaway’s annual meeting was an entirely different beast than today’s weekend-long “Woodstock for Capitalists”. Attendees didn’t have to book their hotel rooms months in advance or wake up before dawn just to get in line outside of an arena. There was no mad rush for seats once the doors opened.
It was a quieter, simpler chapter in Berkshire’s history.
So quiet, in fact, that 1985’s annual meeting was held on a Tuesday. And, instead of a cavernous arena, Warren Buffett and Charlie Munger opted for the Red Lion Inn in downtown Omaha. Approximately 250 shareholders attended the meeting and the ensuing Q&A session lasted only — only? — two hours…
…HOW TO VALUE A BUSINESS: “Do a lot of reading,” replied Buffett.
Generally speaking, he recommended the writings of Benjamin Graham and Philip Fisher for those trying to sharpen their investment mindset — and annual reports and trade magazines for evaluating particular businesses and industries.
Reading, he insisted, is more important than speaking with company executives or other investors. In fact, Buffett admitted that he had recently purchased a substantial amount of Exxon stock before talking to any of that company’s executives. “You’re not going to get any brilliant insights walking into the [Exxon] building,” he said.
And, at least in the money game, size matters.
It’s easier to determine the value of a large business than a small one, Buffett said. If someone buys a gas station, for example, another station opening across the street could have a major effect on the value of the first station…
…A COUPLE OF LAUGHS & A ROUND OF APPLAUSE: No annual meeting is ever complete without some of that trademark Warren Buffett wit.
- Will the federal deficit be substantially reduced? “I’ll believe it when I see it.”
- What about so-called “junk” bonds? “I think they’ll live up to their name,” Buffett quipped.
2. Germany Is Losing Its Mojo. Finding It Again Won’t Be Easy – Bojan Pancevski, Paul Hannon, and William Boston
Two decades ago, Germany revived its moribund economy and became a manufacturing powerhouse of an era of globalization.
Times changed. Germany didn’t keep up. Now Europe’s biggest economy has to reinvent itself again. But its fractured political class is struggling to find answers to a dizzying conjunction of long-term headaches and short-term crises, leading to a growing sense of malaise.
Germany will be the world’s only major economy to contract in 2023, with even sanctioned Russia experiencing growth, according to the International Monetary Fund…
…At Germany’s biggest carmaker Volkswagen, top executives shared a dire assessment on an internal conference call in July, according to people familiar with the event. Exploding costs, falling demand and new rivals such as Tesla and Chinese electric-car makers are making for a “perfect storm,” a divisional chief told his colleagues, adding: “The roof is on fire.”
The problems aren’t new. Germany’s manufacturing output and its gross domestic product have stagnated since 2018, suggesting that its long-successful model has lost its mojo.
China was for years a major driver of Germany’s export boom. A rapidly industrializing China bought up all the capital goods that Germany could make. But China’s investment-heavy growth model has been approaching its limits for years. Growth and demand for imports have faltered…
…Germany’s long industrial boom led to complacency about its domestic weaknesses, from an aging labor force to sclerotic services sectors and mounting bureaucracy. The country was doing better at supporting old industries such as cars, machinery and chemicals than at fostering new ones, such as digital technology. Germany’s only major software company, SAP, was founded in 1975.
Years of skimping on public investment have led to fraying infrastructure, an increasingly mediocre education system and poor high-speed internet and mobile-phone connectivity compared with other advanced economies.
Germany’s once-efficient trains have become a byword for lateness. The public administration’s continued reliance on fax machines became a national joke. Even the national soccer teams are being routinely beaten…
…Germany today is in the midst of another cycle of success, stagnation and pressure for reforms, said Josef Joffe, a longtime newspaper publisher and a fellow at Stanford University.
“Germany will bounce back, but it suffers from two longer-term ailments: above all its failure to transform an old-industry system into a knowledge economy, and an irrational energy policy,” Joffe said…
…Germany still has many strengths. Its deep reservoir of technical and engineering know-how and its specialty in capital goods still put it in a position to profit from future growth in many emerging economies. Its labor-market reforms have greatly improved the share of the population that has a job. The national debt is lower than that of most of its peers and financial markets view its bonds as among the world’s safest assets.
The country’s challenges now are less severe than they were in the 1990s, after German reunification, said Holger Schmieding, economist at Berenberg Bank in Hamburg.
Back then, Germany was struggling with the massive costs of integrating the former Communist east. Rising global competition and rigid labor laws were contributing to high unemployment. Spending on social benefits ballooned. Too many people depended on welfare, while too few workers paid for it. German reliance on manufacturing was seen as old-fashioned at a time when other countries were betting on e-commerce and financial services.
After a period of national angst, then-Chancellor Gerhard Schröder pared back welfare entitlements, deregulated parts of the labor market and pressured the unemployed to take available jobs…
… Private-sector changes were as important as government measures. German companies cooperated with employees to make working practices more flexible. Unions agreed to forgo pay raises in return for keeping factories and jobs in Germany…
… Booming exports to developing countries helped Germany bounce back from the 2008 global financial crisis better than many other Western countries.
Complacency crept in. Service sectors, which made up the bulk of gross domestic product and jobs, were less dynamic than export-oriented manufacturers. Wage restraint sapped consumer demand. German companies saved rather than invested much of their profits.
Successful exporters became reluctant to change. German suppliers of automotive components were so confident of their strength that many dismissed warnings that electric vehicles would soon challenge the internal combustion engine. After failing to invest in batteries and other technology for new-generation cars, many now find themselves overtaken by Chinese upstarts…
…BioNTech, a lauded biotech firm that developed the Covid-19 vaccine produced in partnership with Pfizer, recently decided to move some research and clinical-trial activities to the U.K. because of Germany’s restrictive rules on data protection.
German privacy laws made it impossible to run key studies for cancer cures, BioNTech’s co-founder Ugur Sahin said recently. German approvals processes for new treatments, which were accelerated during the pandemic, have reverted to their sluggish pace, he said…
…One recent law required all German manufacturers to vouch for the environment, legal and ethical credentials of every component’s supplier, requiring even smaller companies to perform due diligence on many foreign firms, often based overseas, such as in China…
…German politicians dismissed warnings that Russian President Vladimir Putin used gas for geopolitical leverage, saying Moscow had always been a reliable supplier. After Putin invaded Ukraine, he throttled gas deliveries to Germany in an attempt to deter European support for Kyiv…
…One problem Germany can’t fix quickly is demographics. A shrinking labor force has left an estimated two million jobs unfilled. Some 43% of German businesses are struggling to find workers, with the average time for hiring someone approaching six months.
Germany’s fragmented political landscape makes it harder to enact far-reaching changes like the country did 20 years ago. In common with much of Europe, established center-right and center-left parties have lost their electoral dominance. The number of parties in Germany’s parliament has risen steadily.
3. GLP-1 Drugs: Not a Miracle Cure for Weight Loss – Biocompounding
Weight loss drugs have been the talk of the town for the last couple of months. The weight loss drugs on the market are Wegovy, Ozempic from Novo Norodisk (NVO), and Mounjaro from Eli Lilly (LLY)…
…These drugs consist of a natural hormone called GLP1…
…GLP-1 drugs mimic the action of a hormone called glucagon-like peptide 1, a natural hormone produced by the body in the gut. When blood sugar levels start to rise after a meal, the body produces this hormone to achieve multiple functions as seen in the image above. By producing and administering this hormone as a therapeutic, the drug will elicit similar effects seen with the natural hormone…
…Apart from increasing insulin production, GLP-1 can also help regulate body weight. GLP-1 improves glycaemic control and stimulates satiety, leading to reductions in food intake and thus body weight. Besides gastric distension and peripheral vagal nerve activation, GLP-1RA induces satiety by influencing brain regions involved in the regulation of feeding, and several routes of action have been proposed. GLP-1 can also reduce gastric emptying, so you don’t feel hungry so fast.
However, apart from the positives GLP1 drugs also cause muscle loss, lessen bone density, and lower your resting metabolic rate.
A research paper published in 2019, reported the percentage of weight loss comprising fat mass vs the proportion comprising lean body mass in patients using the different GLP1 drugs…
…This means that while GLP1’s can help to reduce obesity, individuals using the drugs need to be mindful to preserve their lean mass which requires exercising regularly to ensure they limit the loss of lean mass and improve their basal metabolic rate.
4. An Interview with Daniel Gross and Nat Friedman about the AI Hype Cycle – Ben Thompson, Daniel Gross, and Nat Friedman
NF: I think one of the interesting trends that we’ve seen in the last six months that we weren’t seeing a year ago is basically the application of large models to things that were previously some form of human intellectual labor or productivity labor. So in a way, what they’re doing in these cases is the models are automating or replacing or augmenting some part of a company. They’re competing not with existing software products but with parts of companies.
An example of one that Daniel and I were just talking to recently, we won’t name the company, but they automate filing bids on public tenders for businesses that do business with the government in different jurisdictions, and the time savings of this is totally enormous for these companies, and the upside for them is huge. It’s replacing a raft of internal and external consultants who were doing copywriting and bid preparation and just lots of fairly mechanical but still nothing-to-sneeze-at intellectual labor that produced bid documents. There’s material revenue upside for being able to bid on more things and win more bids, and this company’s growing like crazy, like a weed, so that would be one example.
Another example, there’s a whole sector now of these avatar platforms where people are basically able to produce personalized videos of someone saying, “Hey Ben, I saw that you were interested in our product and I wanted to tell you a little bit about us” and being able to basically generate text, feed that into an avatar platform that generates a realistic video that’s customized and using that in advertising, using it in personal outreach, using it in training materials. There’s some competing with non-consumption here where some of those videos would never have been produced because it would’ve just been too costly, and there’s some like, “Hey, God, I used to have to spend a ton of time doing this, now I can do it quite quickly”. Another example that’s like that, and by the way, all of the avatar, I mean I can name some of those Synthesia, D-ID, HeyGen, they’re all doing great, all of these companies are growing really well.
Another similar category is optimizing e-commerce. There used to be an entire — there still is — an entire industry of consultants and experts and companies who know how to do the SEO around product titles and descriptions and make sure that you have an Amazon landing page that converts, and some of that knowledge and know-how is getting crystallized into models and agent-like tool chains, and the testing can now be done automatically and you can use language models to run this kind of thing. I think this is interesting because these are all niches that really weren’t happening six or nine months ago, and in every category I just mentioned, there’s a company that’s making or soon will be making tens of millions of dollars doing this productively, creating real economic value for their customers and in some cases competing with teams of people or consultants…
...Does this just confirm the thesis though that the most compelling aspects for AI are number one mostly in the enterprise? Again, because enterprises are going to think about costs in a, I hesitate to use the word rational, but in a traditionally rational way, “It’s worth this huge upfront investment because it will pay off X amount over Y axis of time” as opposed to consumers which are more about an experience and may not think about the lifetime cost or value of something, along with this data point where whoever has the data wins. Is that just the reality or is there still opportunities for new entrants in this space?
DG: I think the story of progress is one where things will often, I think, start off looking at the enterprise as a way to make the existing thing better, that idea that the first TV shows or cameras pointed at radio shows, the horseless carriage and all that sort of stuff. So I think there’s a lot of V1 AI, let’s just accelerate or automate a lot of the human interaction with text just because we can do text synthesis now with computers. But the native use cases that’ll come out I think slightly later are going to be consumer ones — those I think will be entirely different things that are not replacing a process that existed before, they’re doing something that was never possible before and so there are consumer experiences today that are not really like anything else on the Internet.
Well, the two that I had on here were that seemed to still have a lot of traction are still growing are Midjourney and Character.AI, which are completely novel experiences and about fantasy and imagination and something that couldn’t be done previously.
DG: Yeah, it’s sort of funny, they told us the robots are going to be really good at blue collar jobs and really terrible at human psychology — that it’ll be the final realm of the human-to-human connection. Of course, it turns out the robots are fantastic at psychology and have zero dexterity for doing actual labor. But Character.AI is a good example and there’s now a bunch of these new kinds of native behavior, and it’s always interesting to ask of these behaviors. So you’re talking to an agent all day on Character, I find the good question to be asking is, “What were you doing previously?” as a way to figure out what this actually is, and the share of time that’s usually being taken is from gaming or social media. It’s really hard, I think, to forecast, to look at the iPhone and to forecast Uber or to look at the Internet and forecast even something like Amazon bots. They’re usually going to be, I think, consumer experiences. Those are the ones that are going to be the really disruptive stuff and the enterprise I think will get a lot of the obvious. We had a person here and now maybe we have a person in a co-pilot model.
That’s kind the trade-off of there being a top-down decision maker that thinks about things like lifetime value.
DG: They’ll do the rational thing.
They’re only going to do the obvious things.
DG: Yeah, and I think if businesses get disrupted by AI in any way, it will be something around a totally native, ideally a different user interface, an acceptance of a customer experience that’s a bit worse, which is usually your Clayton Christensen sort of downmarket disruption, but scales much more. I was actually thinking the companies that are trying to build, “We’re going to do your high-end legal work with AI”, I’m not exactly sure when that’ll work because the models still have this issue with hallucinating things and making things up. Whereas the low end, I was going to call a lawyer for $500 an hour to ask a particular question about my apartment lease, but instead I’m going to talk to legal GPT, that stuff I think will probably be much more impactful…
…There’s an aspect here — one of the questions with the coding bit is Stack Overflow and sites like that have taken the biggest hit, but is this a sustainable future? I think this is a broader question about do we run out of data on the Internet. Is there going to be a data manufacturing industry?
NF: There is already. I think this is the secret story just beneath the surface of what’s happening. Everyone knows about the GPUs, you got to have the GPUs, they’re very expensive, we’re talking about the Nvidia supply chain. All of us know about CoWoS and wafer packaging and Ajinomoto Films and all these things.
But the other key input is data and the readily available tokens you can scrape off the Internet are quickly exhausted, and so there is currently happening beneath the surface, a shadow war for data where the largest AI labs are spending huge amounts of money, like huge amounts of money to acquire more valuable tokens, either paying experts to generate it, working through labeling companies like Scale AI or others. There’s a new crop of startups in that space as well and we think more is going to happen there and it’s going to be a really interesting space to watch.
So there’s a way in which you need these really high IQ, high-value tokens in order to train your models, and the average piece of data you scrape off a random website kind is equal to all the other data that you have, but you’ll pay extra for really valuable training data, and so people are producing it. I don’t know the exact numbers, but I’ve heard rumors that Google is spending a billion dollars this year on generating new training data, and if you’re going to spend billions and billions on your CapEx to build out your GPU training clusters, spending some fraction of that or maybe an equal amount in generating data, which is a kind of CapEx as well kind of makes sense. Someone told me the other day experts are the new GPUs and so there’s this wave of spending on experts who are going to generate tokens that can be valuable.
Then of course the secondary question there is what the legal regime will ultimately be for training. We’re operating in the US, UK, and in Europe under this fair use regime now where it’s fair use for you to scrape text off the Internet as long as it’s public and you’re not going through paywalls or user walls to get it and then you can in aggregate train machine learning models on it. That’s kind of the bright letter of the law, but people don’t always feel good about that and so will the law change, will there be a kind of DMCA for AI? And which way will it cut? I think we don’t know yet and so there may be a war for data in more ways than one over the next couple of years…
… For the record, Nvidia’s results are going to come out in about 12 hours, so we don’t know what’s going to happen yet, but one of the most interesting questions broadly speaking is what is going to happen in the GPU space? Nvidia — do they have a moat, is it going to be a sustainable advantage? Obviously, they have a double advantage right now, in that they have the best hardware and they have CUDA, but there’s massive efforts on both sides to take that away. Can they build up a sustainable advantage that will persist?
NF: For the next couple of years, it’s Nvidia and it’s TPU and those are the only players that are really viable.
Google’s Tensor Processing Unit.
NF: Yeah, it’s a huge strategic asset for Google. I mean, they’re the only company basically that has an independent, not fully independent because obviously they overlap when it gets down to the fabs, and some other parts of the supply chain but they’re not subject to Jensen allocating them H100s. They can just kind of allocate their own and by all accounts, their TPU v5, they’re producing in absolute record numbers.
Easier to deal with TSMC than to deal with Jensen is what you’re saying.
NF: Yeah, I mean, at least they don’t have that one Jensen choke point. I mean, Jensen right now is dealing with overwhelming demand and limited supply, and so he’s having to very carefully allocate GPUs, and it’s sort of a very central resource distribution mechanism and allocation mechanism. It’s kind of wild. So even if you say, “Oh, AMD’s chips are going to be as good,” they’re just not going to produce them in numbers that matter next year and so I think my take is, there’s only two players for the next couple of years that matter, and my take is also that we will be supply-constrained, because there will be more AI applications that take off and need huge inference capacity, and there will be more people trying to train large models.
Is there a hype cycle aspect where we actually look back in a few years, and there were way too many GPUs bought and produced, and we actually end up with an overhang? Basically what happened with Nvidia last year, but at a 100x, a 1000x scale and that actually ends up being a huge accelerant for AI, because you end up with super cheap inference because you have all this depreciated GPUs that were bought up in 2023 and 2024, and then it all crashed. Actually going back to the dot-com bubble and all the fiber that got laid by companies that immediately went out of business.
NF: You might have a dark fiber, “How many shortages are not followed by a glut?” is always the interesting question. They usually do get followed by a glut and I think one scenario in which that happens is I’m a very strong believer in scaling laws for these big general reasoning models. Essentially, the more training data and the more flops you put in, you’re just going to get a better and better model out, and we’ve seen this now over several orders of magnitude, it’s just incredibly consistent. We saw it with GPT-1 and GPT-2, and GPT-3, and now GPT-4, and we’ll see it I think with GPT-5. So, it’s possible that there’s some escape velocity that occurs where a few labs are the only ones who can afford to train the GPT-5 or GPT-6 equivalent models, and all of the startups and businesses that were getting essentially a sub-scale amount of GPU, unless they were doing something incredibly domain specific, those are no longer needed. So, you’ll have I don’t know, three or four companies that afford to train the $10 billion model, and that’s actually a limited number of GPUs.
5. Respect and Admiration – Morgan Housel
This isn’t universal, but there are cases when people’s desire to show off fancy stuff is because it’s their only, desperate, way to gain some sense of respect and admiration. They don’t have any wisdom, intelligence, humor, empathy, or capacity for love to gain people’s respect. So they rely on the only remaining, and least effective, lever: Look at my car, beep beep, vroom vroom…
…My guess is that if your favorite comedian, or actor, or athlete turned out to be broke, you wouldn’t care. It wouldn’t impact how much you admire them, because you admire them for talents that money can’t buy.
Even when Amazon was huge and successful, Jeff Bezos used to drive a Honda Accord. Today he has a $500 million yacht. Is he respected and admired more for it? Not in the slightest. He could ride a Huffy bike and people would consider him the greatest entrepreneur of our era, because he is. Steve Jobs didn’t have any furniture. It didn’t matter. He’s a genius. He’s Steve Jobs. Material stuff makes no difference when you’re respected and admired for internal traits…
…Once you see people being respected and admired for reasons that have nothing to do with the stuff they own, you begin to wonder why you have such a strong desire for those possessions. I tend to view material desire as a loose proxy for the inverse of what else you have to offer the world. The higher my desire for fancy stuff, the less real value I have to offer.
Disclaimer: The Good Investors is the personal investing blog of two simple guys who are passionate about educating Singaporeans about stock market investing. By using this Site, you specifically agree that none of the information provided constitutes financial, investment, or other professional advice. It is only intended to provide education. Speak with a professional before making important decisions about your money, your professional life, or even your personal life. We currently have a vested interest in Alphabet (parent of Google), Amazon, and TSMC. Holdings are subject to change at any time.