What We’re Reading (Week Ending 03 September 2023) - 03 Sep 2023
Reading helps us learn about the world and it is a really important aspect of investing. The legendary Charlie Munger even goes so far as to say that “I don’t think you can get to be a really good investor over a broad range without doing a massive amount of reading.” We (the co-founders of Compounder Fund) read widely across a range of topics, including investing, business, technology, and the world in general. We want to regularly share the best articles we’ve come across recently. Here they are (for the week ending 03 September 2023):
1. China Reaches Peak Gasoline in Milestone for Electric Vehicles – Colin McKerracher
Earlier this month, Chinese oil giant Sinopec made a surprise announcement that mostly flew under the radar. It’s now expecting gasoline demand in China to peak this year, two years earlier than its previous outlooks.
The main culprit? The surging number of electric vehicles on the road…
…China has been the largest driver of global growth for refined oil products like gasoline and diesel over the last two decades. But EV adoption rates in China are now soaring, with August figures likely to show plug-in vehicles hitting 38% of new passenger-vehicle sales. That’s up from just 6% in 2020 and is starting to materially dent fuel demand.
Fuel demand in two and three-wheeled vehicles is already in structural decline, with BNEF estimating that 70% of total kilometers traveled by these vehicles already switched over to electric. Fuel demand for cars will be the next to turn, since well over 5% of the passenger-vehicle fleet is now either battery-electric or plug-in hybrid. The internal combustion vehicle fleet is also becoming more efficient due to rising fuel-economy targets.
Diesel demand for heavier vehicles will keep growing for a bit longer, but even there a seismic shift is underway. Electric, fuel cell and battery-swapping options have quickly climbed to 12% of light commercial vehicle sales and 4% to 5% of medium and heavy commercial vehicle sales. That heavy-duty figure is likely to climb to over 10% by 2025.
Combine all those segments, and BNEF expects total oil demand for road transport in China to peak late next year. Demand won’t drop off a cliff anytime soon — fleet turnover in the trucking segment in particular will take time — but it still marks a major shift for global oil demand patterns. It also has big implications for refiners that need to quickly adjust the mix of products they produce.
It also called out the effects China’s ride-hailing fleet is having on urban gasoline demand.
Vehicles used for ride-hailing in China are far more likely to be electric — their share is nearing 40% of the fleet — than those that are privately owned. Electric ride-hailing vehicles are also more productive than their gasoline-powered counterparts, accounting for 50% of the kilometers traveled on market leader Didi’s ride-hailing platform in December…
…The Sinopec announcement highlights how looking just at the fleet of vehicles can lead one to miss the full story with respect to energy impact…
…The speed that oil gets squeezed out of the transport mix depends on how fast countries like China switch over the number of kilometers traveled to electric — not just the number of cars and trucks.
2. Peasant Logic and the Russian GKO Trade – Joe Pimbley
Later in 1998, after Russia blew up, I attended a public risk management conference in Paris. And one of the speakers was Allen Wheat, CEO of Credit Suisse at the time. I didn’t know Wheat, but he impressed me as a blunt, direct-speaking guy. He talked about Credit Suisse’s version of the GKO trade. He didn’t mention a short position in a Russian-issued dollar bond, so maybe Credit Suisse didn’t bother with the credit risk hedge. But he talked about the GKO and rubles and the cross-currency forwards Credit Suisse executed with Russian banks…
…Interesting to me, Wheat’s story was not that he got to the bottom of this controversy and figured out what part of the loss owed to market risk and what part owed to credit risk. Wheat’s conclusion to his board of directors was that Credit Suisse had a problem with its “risk management philosophy.” It had market risk and credit risk silos when really risk management must be integrated. It’s unproductivee to distinguish market risk from credit risk if things are going to fall between the cracks and nobody’s going to take responsibility for understanding the complete risk picture.
Clearly, that’s a nice message, even if you wonder why Wheat didn’t work through the finger-pointing and hold people to account. Who can argue against an integrated approach to risk? But Wheat admitted he got chastised by his board when he presented that conclusion. The board said, “Allen, we think we understand what’s wrong here. It’s good to do all your analysis and get deep into the details, but at some point, you’re not seeing the big picture. You really need to use ‘peasant logic.’”
Wheat explained that “peasant logic” was the board’s term for what we might call “common sense,” but I like peasant logic better. The board said, “You people worry about how good your models are and you wonder about using two years of historical data or five years of historical data, and whether one is better than the other, and how much data you should have. We think you should have looked at the big picture and said, ‘messing around with 40% yields means there’s a lot of risk here. This is an unstable government and currency situation.’ We think you aren’t seeing the forest for the trees.”
So this was Wheat’s point: sometimes it’s good to forget the data and models and use peasant logic. In this case, if there are abnormal returns, there must be some abnormal risk…
…Then it came time for questions, and from the back of the room, someone had to shout out his question to be heard. And as soon as he started speaking, you could tell it’s a Russian accent and the guy is Russian. Being Russian lent authenticity to his remark, “You want historical data. I’ll give you 75 years of historical data. Russia has never honored any debt obligation.”
…Unfortunately, Wheat’s reaction was to be annoyed. Wheat didn’t say, “Wow, what a great way to look at this. Why are we trusting Russian debt?” And he also didn’t say, “That’s a great example of the peasant logic the board was trying to impress upon me.”
The Russian continued. “I work for Merrill Lynch and we did this trade also and lost a lot of money. Beforehand, I told them it was a terrible trade because of Russia’s history and they didn’t listen to me because I’m just a mathematician.” Wheat still hadn’t cottoned to the idea that the Russian was helping him make his point about peasant logic, so he said in a rather dismissive, sarcastic way, “Well, I wish we had you working for us, then we wouldn’t have lost money. Right?”
Now it’s easy in hindsight, when you know how something worked out, to say “Aha, I knew such and such.” But still, I thought the Russian added to Wheat’s remarks and his remarks really made Wheat’s point. This guy in the audience was demonstrating peasant logic. The traders put all these fancy complex pieces together and think they’re really smart, but what the heck are they doing lending money to a government that, to this guy who is closer to it than the rest of us, you shouldn’t trust?
3. Google Gemini Eats The World – Gemini Smashes GPT-4 By 5X, The GPU-Poors – Dylan Patel and Daniel Nishball
The statement that may not be obvious is that the sleeping giant, Google has woken up, and they are iterating on a pace that will smash GPT-4 total pre-training FLOPS by 5x before the end of the year. The path is clear to 20x by the end of next year given their current infrastructure buildout. Whether Google has the stomach to put these models out publicly without neutering their creativity or their existing business model is a different discussion…
…Access to compute is a bimodal distribution. There are a handful of firms with 20k+ A/H100 GPUs, and individual researchers can access 100s or 1,000s of GPUs for pet projects. The chief among these are researchers at OpenAI, Google, Anthropic, Inflection, X, and Meta, who will have the highest ratios of compute resources to researchers. A few of the firms above as well as multiple Chinese firms will 100k+ by the end of next year, although we are unsure of the ratio of researchers in China, only the GPU volumes.
One of the funniest trends we see in the Bay area is with top ML researchers bragging about how many GPUs they have or will have access to soon. In fact, this has become so pervasive over the last ~4 months that it’s become a measuring contest that is directly influencing where top researchers decide to go. Meta, who will have the 2nd most number of H100 GPUs in the world, is actively using it as a recruiting tactic.
Then there are a whole host of startups and open-source researchers who are struggling with far fewer GPUs. They are spending significant time and effort attempting to do things that simply don’t help, or frankly, matter. For example, many researchers are spending countless hours agonizing on fine-tuning models with GPUs that don’t have enough VRAM. This is an extremely counter-productive use of their skills and time.
These startups and open-source researchers are using larger LLMs to fine-tune smaller models for leaderboard style benchmarks with broken evaluation methods that give more emphasis to style rather than accuracy or usefulness. They are generally ignorant that pretraining datasets and IFT data need to be significantly larger/higher quality for smaller open models to improve in real workloads.
Yes, being efficient with GPUs is very important, but in many ways, that’s being ignored by the GPU-poors. They aren’t concerned with efficiency at scale, and their time isn’t being spent productively. What can be done commercially in their GPU-poor environment is mostly irrelevant to a world that will be flooded by more than 3.5 million H100s by the end of next year. For learning, experimenting, smaller weaker gaming GPUs are just fine…
…While the US and China will be able to keep racing ahead, the European startups and government backed supercomputers such as Jules Verne are also completely uncompetitive. Europe will fall behind in this race due to the lack of ability to make big investments and choosing to stay GPU-poor. Even multiple Middle Eastern countries are investing more on enabling large scale infrastructure for AI.
Being GPU-poor isn’t limited to only scrappy startups though. Some of the most well recognized AI firms, HuggingFace, Databricks (MosaicML), and Together are also part of this GPU-poor group. In fact, they may be the most GPU-poor groups out there with regard to both the number of world class researchers per GPU and the number of GPUs versus the ambition/potential customer demand. They have world class researchers, but all of them are limited by working on systems with orders of magnitude less capabilities. These firms have tremendous inbound from enterprises on training real models, and on the order of thousands of H100s coming in, but that won’t be enough to grab much of the market.
Nvidia is eating their lunch with multiple times as many GPUs in their DGX Cloud service and various in-house supercomputers. Nvidia’s DGX Cloud offers pretrained models, frameworks for data processing, vector databases and personalization, optimized inference engines, APIs, and support from NVIDIA experts to help enterprises tune models for their custom use cases. That service has also already racked up multiple larger enterprises from verticals such as SaaS, insurance, manufacturing, pharmaceuticals, productivity software, and automotive. While not all customers are announced, even the public list of Amgen, Adobe, CCC, ServiceNow, Accenture, AstraZeneca, Getty Images, Shutterstock, Morningstar, Evozyne, Insilico Medicine, Quantiphi, InstaDeep, Oxford Nanopore, Peptone, Relation Therapeutics, ALCHEMAB Therapeutics, and Runway is quite impressive.
4. Making Sense Of The China Meltdown Story – Louis-Vincent Gave
It is impossible to turn to a newspaper, financial television station or podcast today without getting told all about the unfolding implosion of the Chinese economy. Years of over-building, white elephants and unproductive infrastructure spending are finally coming home to roost. Large property conglomerates like Evergrande and Country Garden are going bust. And with them, so are hopes for any Chinese economic rebound. Meanwhile, the Chinese government is either too incompetent, too ideologically blinkered, or simply too communist to do anything about this developing disaster.
Interestingly, however, financial markets are not confirming the doom and gloom running rampant across the financial media…
…At Gavekal, we look at bank shares as leading indicators of financial trouble. When we see bank shares break out to new lows, it is usually a signal that investors should head for the exit as quickly as possible. This was certainly the case in 2007-08 in the US. Between February 2007 and July 2008 (six weeks before the collapse of Lehman Brothers), banks shares lost -60% of their value…
…Now undeniably, Chinese bank shares have not been the place to be over the past few years. Nonetheless, Chinese bank shares are still up a significant amount over the last decade. And this year, they have not even taken out the low of 2022 made on October 31st following the Chinese Communist Party congress. To be sure, the chart below is hardly enticing, even if the slope of the 200-day moving average is positive. Still, Chinese bank shares do not seem to be heralding a near-term financial sector Armageddon…
…China is the number one or two importer of almost every major commodity you can think of. So, if the Chinese economy were experiencing a meltdown, you would expect commodity prices to be soft. Today, we are seeing the opposite. The CRB index has had a strong year so far in 2023, and is trading above its 200-day moving average. Moreover, the 200-day moving average now has a positive slope. Together, all this would seem to point towards an unfolding commodity bull market more than a Chinese meltdown…
…Jacques Rueff used to say that exchange rates are the “sewers in which unearned rights accumulate.” This is a fancy way of saying that exchange rates tend to be the first variable of adjustment for any economy that has accumulated imbalances. On this front, the renminbi has been weak in recent months, although, like Chinese equities, it has yet to take out October’s lows.
That is against the US dollar. Against the yen, the currency of China’s more direct competitor, Japan, the renminbi continues to grind higher and is not far off making new all-time highs. And interestingly, in recent weeks, the renminbi has been rebounding against the South Korean won.
This is somewhat counterintuitive. In recent weeks, oceans of ink have been spilled about how China is the center of a developing financial maelstrom. Typically, countries spiraling down the financial plughole do not see their currencies rise against those of their immediate neighbors and competitors…
…In other words, a range of data points seems to indicate that Chinese consumption is holding up well. This might help to explain why the share prices of LVMH, Hermès, Ferrari and most other producers of luxury goods are up on the year. If China really was facing an economic crash, wouldn’t you expect the share prices of luxury good manufacturers to at least reflect some degree of concern?…
…Staying on the US treasury market, it is also odd how Chinese government bonds have outperformed US treasuries so massively over the past few years. Having gone through a fair number of emerging market crises, I can say with my hand on my heart that I have never before seen the government bonds of an emerging market in crisis outperform US treasuries. Yet since the start of Covid, long-dated Chinese government bonds have outperformed long-dated US treasuries by 35.3%.
In fact, Chinese bonds have been a beacon of stability, with the five-year yield on Chinese government bonds spending most of the period since the 2008 global crisis hovering between 2.3% and 3.8%. Today, the five-year yield sits at the low end of this trading band. But for all the negativity out there, yields have yet to break out on the downside…
…While the Chinese government debt market has been stable, the pain has certainly been dished out in the Chinese high yield market. Yields have shot up and liquidity in the Chinese corporate bond market has all but evaporated. Perhaps this is because historically many of the end buyers have been foreign hedge funds, and the Chinese government feels no obligation to make foreign hedge funds whole. Or perhaps it is because most of the issuers were property developers, a category of economic actor that the CCP profoundly dislikes.
Whatever the reasons, the Chinese high yield debt market is where most of the pain of today’s slowdown has been—and continues to be—felt. Interestingly, however, it seems that the pain in the market was worse last year than this year. Even though yields are still punishingly high, they do seem to be down from where they were a year ago…
…Why the sudden drumbeat about collapsing Chinese real estate and impending financial crisis when the Chinese real estate problem has been a slow-moving car crash over the past five years, and when, as the charts above show, markets don’t seem to indicate a crisis point?
At least, markets outside the US treasury market don’t seem to indicate a crisis point. So could the developing meltdown in US treasuries help to explain the urgency of the “China in crisis” narrative?…
…Basically, US treasuries have delivered no positive absolute returns to any investor who bought bonds after 2015. Meanwhile, investors who bought Chinese government bonds in recent years are in the money, unless they bought at the height of the Covid panic in late 2021 and early 2022. This probably makes sense given the extraordinary divergence between US inflation and Chinese inflation.
None of this would matter if China was not in the process of trying to dedollarize the global trade in commodities and was not playing its diplomatic cards, for example at this week’s BRICS summit, in an attempt to undercut the US dollar (see Clash Of Empires). But with China actively trying to build a bigger role for the renminbi in global payments, is it really surprising to see the Western media, which long ago gave up any semblance of independence, highlighting China’s warts? Probably not. But the fact that the US treasury market now seems to be entering a full-on meltdown adds even more urgency to the need to highlight China’s weaknesses.
A Chinese meltdown, reminiscent of the 1997 Asian crisis, would be just what the doctor ordered for an ailing US treasury market: a global deflationary shock that would unleash a new surge of demand and a “safety bid” for US treasuries. For now, this is not materializing, hence the continued sell-off in US treasuries. But then, the Chinese meltdown isn’t materializing either.
5. Why China’s economy ran off the rails – Noah Smith
This is a pretty momentous happening, since a lot of people had started to believe — implicitly or explicitly — that China’s economy would never suffer the sort of crash that periodically derails all other economies. That was always wrong, of course, and now the bears are coming out for a well-deserved victory lap…
…Anyway, OK, here is my quick story of what happened to China. In the 1980s, 90s, and early 2000s, China reaped huge productivity gains from liberalizing pieces of its state-controlled economy. Industrial policy was mostly left to local governments, who wooed foreign investors and made it easy for them to open factories, while the central government mostly focused on big macro things like making capital and energy cheap and holding down the value of the currency. As a result, China became the world’s factory, and its exports and domestic investment soared. As did its GDP.
At this time there were also substantial tailwinds for the Chinese economy, including a large rural surplus population who could be moved to the cities for more productive work, a youth bulge that created temporarily favorable demographics, and so on. China was also both willing and able to either buy, copy, or steal large amounts of existing technology from the U.S. and other rich countries.
Meanwhile, during this time, real estate became an essential method by which China distributed the gains from this stupendous economic growth. It was the main financial asset for regular Chinese people, and land sales were how local governments paid for public services.
Then the 2008 financial crisis hit the U.S., and the Euro crisis hit Europe. The stricken economies of the developed nations were suddenly unable to keep buying ever-increasing amounts of Chinese goods (and this was on top of export markets becoming increasingly saturated). Exports, which had been getting steadily more and more important for the Chinese economy, suddenly started to take a back seat:..
… The government told banks to lend a lot in order to avoid a recession, and most of the companies they knew how to shovel money at were in the real estate business in some way. That strategy was successful at avoiding a recession in 2008-10, and over the next decade China used it again whenever danger seemed to threaten — such as in 2015 after a stock market crash.
Maybe China’s leaders were afraid of what would happen to them if they ever let growth slip, or maybe they didn’t really think about what the costs of this policy might be. In any case, China basically pivoted from being an export-led economy to being a real-estate-led economy. Real-estate-related industries soared to almost 30% of total output.
That pivot saved China from recessions in the 2010s, but it also gave rise to a number of unintended negative consequences. First, construction and related industries tend to have lower productivity growth than other industries (for reasons that aren’t 100% clear). So continuing to shift the country’s resources of labor and capital toward those industries ended up lowering aggregate productivity growth. Total factor productivity, which had increased steadily in the 2000s, suddenly flatlined in the 2010s:..
…This productivity slowdown probably wasn’t only due to real estate — copying foreign technology started to become more difficult as China appropriated all the easier stuff. Nor was productivity the only thing weighing on China’s growth — around this same time, surplus rural labor dried up. Anyway, put it all together, and you get a slowdown in GDP growth in the 2010s, from around 10% to around 6% or 7%:
But 6-7% is still pretty darn fast. In order to keep growth going at that pace, China had to invest a lot — around 43% of its GDP, more than in the glory days of the early 2000s, and much more than Japan and Korea at similar points in their own industrial development.
Only instead of deploying that capital efficiently, China was just putting it toward increasingly low-return real estate. The return on assets for private companies collapsed:
Much of this decline was due simply to the Chinese economy’s shift toward real estate; if you strip out real estate, the deterioration in the private sector looks much less severe…
…So even as the pivot to real estate was adding to a long-term slowdown in China’s growth, it was also generating a bubble that would eventually cause an acute short-term slowdown as well. If there’s a grand unified theory of China’s economic woes, it’s simply “too much real estate”.
Disclaimer: None of the information or analysis presented is intended to form the basis for any offer or recommendation. We currently have a vested interest in Adobe, Alphabet (parent of Google), and Meta Platforms. Holdings are subject to change at any time. Holdings are subject to change at any time.
