Out of
Our Minds

It’s not just industrial competition. Rare earths are a critical chokepoint in accessing the modern world. China recently blocked the export of rare earths to western defense manufacturers, which could hurt the production of everything from bullets to fighter jets. The move was the latest tit-for-tat in the race for global supremacy between the world’s two largest economies. Because of that, the US has called these minerals critical to national security and made reviving a domestic rare-earths industry a priority. This has been a focus in both the current and former administrations.

Companies in the industry have gotten a boost from the attention. In July, the Department of Defense made a US$400 million equity investment in Las Vegas-based MP Materials. MP Materials, which owns the only fully operational rare-earths mine in the country—Mountain Pass in California—has signed contracts with the Department of Defense, General Motors, and Apple to supply magnets that incorporate rare earths. MP Materials’ stock, along with those of other rare earths miners operating in the US such as Lynas Rare Earths and Ramaco Resources, have risen sharply in recent months.

These efforts are an attempt to rebuild an industry that once existed in the US. Thirty years ago, about a third of all mined rare earths came from the US. Amid the globalization of the 1990s and 2000s, as China built up its capacity, the US and other countries ceded the territory. By 2011, China was mining 97% of all rare earths.

China is the leading supplier of ore, and it is also the biggest refiner of that ore. Mining is an industry where scale is paramount, the barriers to entry are high, the capital investments required are large, and the rivalry can be devastating. China’s domination of the industry will not be easy to offset, much less overtake.

Right now, even the most advanced cars sold commercially can’t take complete control of the vehicle. The driver is still driving—even if they aren’t holding the steering wheel at that moment, they are responsible if the car hits something. So, for now, car owners by law still need insurance. Eventually, though, some auto manufacturer will offer a model that does take complete control of the car’s operation away from the user (“driver” at this point ceases to be an accurate description of this person). This level of autonomy could shift liability for accidents.

Where that liability shifts to will be a question not just for the auto-insurance industry but for lawmakers and regulators as well. US autonomous-car company Waymo, for instance, self-insures its self-driving robotaxis. It’s possible that manufacturers become responsible for liability insurance for the cars they sell. Or maybe individual vehicle owners will still be responsible. It will all come down to regulations that haven’t even been written yet, which could slow down adoption.

Regardless, cars that drive themselves are not going to take over the nation’s roads any time soon. Mass-market autonomous vehicles could start hitting the market as soon as 2026 and could comprise 10% of the US auto fleet anywhere between 2035 and 2044, depending on adoption, Morningstar estimated. It will be years before self-driving cars make up even a small part of the nation’s auto fleet—and the number will not be growing very fast. It is not likely, therefore, to have a meaningful impact on Progressive, State Farm, and the other insurers in our investment horizon.

What will matter for auto insurers’ profitability in the years ahead is what matters now: access to data. It isn’t just car-crash frequency insurers care about. There are questions about crash severity, customer retention, changes in the probability and severity of a claim over time, as well as opportunities to cross-sell and upsell. These companies need to do complicated lifetime-value calculations on their customers when analyzing how much to spend to retain and acquire new ones. Having more data makes those calculations more accurate. The largest companies with the most customers and most data will have a marked advantage over their competitors. In this sense, autonomous vehicles are not a game-changer for insurers.

For now, the insurance industry can continue with business as usual, but with one eye on that day when autonomous vehicles take over the country’s roads.

The company behind Kroger’s RFID pilot program is Avery Dennison, which might surprise investors who are more familiar with its low-tech products such as mailing labels. Before Stan Avery invented self-adhesive labels, or stickers, nearly a century ago, labels had to be moistened with water to make them stick. Avery mailing labels later became a staple of office-supplies stores. Although the company sold its office- and consumer-products division to CCL Industries in 2013, adhesive technology remains the core of its business, used in everything from shampoo- and wine-bottle labels to barcode stickers, shipping labels, reflective films for traffic signage, customized apparel tags—and RFID labels.

Avery Dennison has been in the business of RFID technology since the early 2000s, but the technology didn’t immediately take off due to its high costs and a lack of readily available software to manage RFID data. In 2000, each RFID tag cost more than US$1, compared to 15 cents a decade ago and just 5 cents today. As technological advances and manufacturing scale brought down those costs, prominent retailers such as Macy’s and Marks and Spencer began using RFID-tagging systems, and it wasn’t until they demonstrated the benefits that other chains followed suit. In perhaps the most cutting-edge example, Avery Dennison customer Uniqlo uses RFID tags for all its merchandise, which has enabled the companies to create a touchless self-checkout system that doesn’t require the scanning of individual barcodes. Today, RFID labels generate just 10% of Avery Dennison’s total revenue, but sales of these products are growing significantly faster than the rest of the company’s businesses while also earning superior margins.

According to the company’s estimates, RFID technology has penetrated about 40% of the apparel market, where there are roughly 45 billion taggable units. The larger growth opportunity is in food, where there are over four times as many potential taggable units but very few companies are using RFID systems. In the past, an obstacle to using RFID labels for food was that the chips wouldn’t work under varying temperature and humidity conditions, preventing tagged products from being refrigerated or microwaved. Avery Dennison has since developed solutions to these problems, unlocking a new market.

As it pursues that growth opportunity, the company’s scale and vertical integration in RFID manufacturing are key competitive advantages. Inside an RFID tag are inlay components such as an antenna, microchip, and substrate to hold them together; the company has a 50% share of the RFID inlay industry and more than 1,500 patents to protect the design of its antenna and other tag properties. Its strength in this area has also supported continued investment in R&D to sustain its market leadership over the long run, such as developing designs suitable for customers in other markets, including logistics and health care.

Kroger’s initiative is important because, like the apparel industry, widespread use of RFID technology in the grocery industry will likely hinge on a few success stories among the big chains. The cost of RFID tags is also a key metric that investors should monitor; as that comes down, Avery Dennison can make a more compelling case to customers that the money saved from reducing their food waste and labor needs is worth the investment in an RFID system.

If Avery Dennison’s project for Kroger is successful, and the cost of the tags continues to fall, Kroger may be just the first of many grocers to switch to this method of tracking their perishables.

This is how we evaluate companies. The cornerstone of our analysis is the competitive-forces framework designed by Harvard University professor Michael Porter. Competitive strategy, as the field has come to be called, looks not just at a company’s internal fundamentals, but also compares them to its peers across its industry.

Consider Schneider and its competitor Eaton Corp. Schneider Electric makes equipment for electricity distribution and industrial controls and is based in Rueil-Malmaison, France. Eaton, which also makes equipment for electricity distribution, has its operations based in Ohio. Ostensibly, that would give Eaton a leg up over Schneider. But these companies have virtually the same manufacturing footprint and supply chain despite the different domiciles. Schneider has about 20 factories in the US. So it isn’t likely to lose any ground to Eaton, at least not due to tariffs.

An even better example is Techtronic and its competitor Stanley Black & Decker. Techtronic is based in Hong Kong and owns the Ryobi and Milwaukee brands. It was a pioneer in battery-powered, cordless power tools. Stanley is based in Connecticut and makes the Black & Decker, Craftsman, and DeWalt brands. Again, one company is based in the US and the other is based overseas. But both companies’ manufacturing operations are based overseas so the real issue is where overseas they are.

Stanley Black & Decker has about 24% of its power-tools manufacturing capacity in Mexico, and 18% in China. Techtronic also has production in Mexico and China. But it is already in the process of moving its China production out of the country, which it plans to complete this year, and importantly already has a substantial portion of its power-tools manufacturing in Vietnam. Stanley is also planning to move production out of China, but is further behind in that process. It will likely take one to two years to complete. Currently, Vietnam’s tariffs are at 10% while Mexico is at 25% for non-USMCA-compliant goods and China’s now at 30%. Therefore, Techtronic has an advantage in already having production in Vietnam and in being further along in moving its production out of China. The advantage could of course shift if the tariffs increase or decrease, but the analytical framework remains in place.

Looking at the tariff question this way doesn’t mean a company won’t be affected by them. The economic consensus is that the steep tariffs will hurt everybody. But examining that potential pain through the lens of competitive advantage provides a clearer picture of how companies will be affected, and which ones will come out stronger.

At the time, China’s biggest problems revolved around inefficient capital allocation, and the outdated development model that was unbalanced and unsustainable. So, a transformation of its economy was much needed. Made in China 2025, with a focus on ten key sectors including information technology, industrial machines and robots, biotech, transportation, and electric vehicles, was a plan to launch the country into the future. If they achieved even half of what they set out to do, I thought it would be a success. So, now that 2025 is here, how has it worked?

China has become the world’s biggest manufacturer of and market for electric vehicles. EV sales in China in 2025 are expected to total about 8 million cars; the US is second with only 1.4 million. BYD has become the largest EV manufacturer in the world and recently unveiled its latest game-changing innovations: a state-of-the-art advanced driver assistance system, or ADAS, free on most of its models and a five-minute charger. Not only are BYD’s cars comparable in quality to its competitors, they’re also better priced. Meanwhile, a high-powered prototype version of Xiaomi’s SU7 Ultra electric vehicle was one of the top preproduction cars at the famed Nürburgring race track in Germany. What’s really surprising is that Xiaomi is primarily known as a cellphone and appliances maker; it launched the SU7, its first model, in March 2024.

Chinese companies have also expanded rapidly in the biotechnology industry. Since 2007, when the first Chinese-developed drug underwent Phase II testing in the US—under the regulations of the FDA—China’s efforts have mushroomed. Between 2007 and 2023, there were 691 clinical trials of Chinese-developed drug candidates in the US; 82% of those trials occurred between 2019 and 2023. A Chinese biotech named Akeso released results in September 2024 for a cancer drug that was on par with the efficacy of Merck’s mega-blockbuster Keytruda. In 2019, none of the molecules licensed by large pharmaceutical companies came from China. By 2024, 30% did.

A key factor making all this possible is China’s long-term commitment in education and indigenous R&D. There is today a pipeline of young, educated researchers creating the theoretical breakthroughs that result in commercial products years or even decades later.

An emerging field where China has a commanding lead is humanoid robots. In 2023, the government issued a roadmap and aggressive timeline to force this field into commercialization. China planned for mass production in 2025 and maturation of the supply chain by 2027, with the goal of becoming an industry leader. China seems on track to achieve its goal, partly thanks to its success in localizing a critical component that had long been dominated by Japanese robotic makers, the compact precision gears needed in every robot joint. Thanks to a robust ecosystem and supply chain of components suppliers, originally built for the car industry and the home-appliance industries, companies such as Sanhua Intelligent Controls and Shuanghuan Driveline have evolved to become key players in the humanoid-robot ecosystem. Goldman Sachs estimated that by 2035 there could be annual sales of 1.4 million robots a year in a US$38 billion market. Much of that will be Chinese.

Looking back, I’d say China has in fact achieved well over half of its goals. China’s abilities have advanced from just producing other countries’ clothes and toys to competing directly in a range of critical, high-tech, and high-value industries, including industrial robots, electric vehicles, solar energy, biotech, and others.

But what is most impressive is that the culture created by the Made in China 2025 plan has resulted in a more innovative culture overall. The most recent high-profile example of China’s ability to innovate came in January when a little-known company named DeepSeek unveiled an AI chatbot that virtually matched the capabilities of western AI models and, just as critically, did so requiring only a fraction of the computing power. While DeepSeek has grabbed most of the attention, AI models from several other Chinese companies such as Alibaba have also scored well on industry benchmarks.

Looking at the horizon, one might expect more “DeepSeek moments” in other fields. SiCarrier Technologies, a Chinese semiconductor equipment maker founded in 2021, debuted a full series of new equipment, from etching to atomic-layer deposition, at the SEMICON China exhibition last month. It joined a long list of Chinese semiconductor-equipment makers that have been working to localize the entire supply chain in the face of US restrictions. While China’s chip-making capabilities have a long way to go to reach the level of industry leaders such as ASML on the most advanced kinds of chips, it has closed the gap in other areas of chip making. China has also unveiled a quantum computer that operated a million times faster than Google’s, and a fusion reactor called “artificial sun” that set a world record by operating for 17 minutes. That broke the record set by…the same reactor.

These achievements reflect China’s pivot from an innovation model led by private companies, which focused mainly on consumer technologies, to a state-directed, whole-of-nation approach, with a sharper focus on areas with national security implications. A key factor making all this possible is China’s long-term commitment in education and indigenous R&D. There is today a pipeline of young, educated researchers creating the theoretical breakthroughs that result in commercial products years or even decades later. More than 11 million Chinese students graduated from college last year, up from 7.1 million in 2014 and only 2.4 million in 2004. China produces 3.6 million graduates annually in STEM fields (science, technology, engineering, and mathematics), more than any other country. The US is third, with 820,000, out of a total of 4.1 million college graduates annually. As I warned in 2017, “just as older and less‐educated labor from China disrupted the low‐end of the global labor pool in past decades, the younger and better‐educated generation in China is now disrupting the middle class of the developed world.”

China’s gross domestic expenditure on R&D has surged in the past decade, according to the Organization for Economic Co-operation and Development, and is now neck-and-neck with the US. In terms of the results from R&D and basic scientific research, one indicator to watch is The Nature Index, which tracks articles in the top 145 natural-science and health-science research journals. It’s a good way to track the productivity of research institutions (mainly universities but also companies and research labs) conducting fundamental research. In 2014, China’s Nature Index ranking was a distant second to the US. In 2024, China surpassed the US as the top nation on the Nature Index, and nine of the top ten institutions were Chinese.

Housing affordability has been a significant challenge in Germany. After home prices soared 80% from 2012 to 2022, it didn’t take much for higher interest rates to tip the country’s housing market into a downturn. Yet, despite the downturn, Scout24 managed to increase sales by double digits in 2023 and 2024. How? In a weak housing market, sellers and agents are forced to increase spending on marketing to help houses sell. Scout24 also has an advantage given that its portal, ImmoScout24, captures nearly 70% of the money spent on digital advertising for real estate in Germany, which gives the company pricing power.

Source: Company disclosure, Harding Loevner analyst estimates. Data as of March 31, 2025.

The digital-payments industry is a broad umbrella, comprising everything from consumer services such as digital wallets and peer-to-peer payments to merchant-facing technology that forms the backbone of online shopping, such as merchant acquiring, payment gateways, and payment processing. On the consumer side, digital wallets securely store the same cards found in your physical wallet, while peer-to-peer payment apps such as Venmo are the kind you might use to settle a tab with a friend by wiring them your portion. On the merchant side, a merchant acquirer is an institution that enables retailers to accept credit-card payments. Retailers also need a payment gateway to securely transmit payment information, and a payment processor to authorize and confirm the funds transfer between the cardholder’s issuing bank and the retailer. PayPal is unique because it does all of those things—and nothing else.

Focus is crucial for industries in which consumer habits determine growth. Products and services that become easy routines for people generally see greater usage, resulting in higher revenues; they also have a better chance of holding onto their users even when other good options become available. But for a company to achieve this level of customer loyalty, it must understand its users’ needs and be able to anticipate how those needs might change. The odds of success are likely higher if the product or service is an area of strategic focus. It’s why Spotify, for example, has been able to out-innovate and build a more user-friendly service than Apple Music. Much like music streaming, payments are a small extension of Apple’s core business, while at PayPal, it is the core business. Chief Executive Officer Alex Chriss, who started in late 2023, said at a recent investor presentation that the company is focused on product improvements that should help differentiate PayPal and build habituation among users.

For example, it is applying artificial intelligence to the data it collects from merchants and a shopper’s financial institutions so it can flag which card in their digital wallet would provide the most rewards points or largest discount. PayPal can do this because it is a two-sided network, with both consumers and merchants on its platform. Apple Pay and Google Pay don’t directly partner with merchants or track purchases at the individual product level, so they’re more limited in their ability to offer personalized rewards.

Last year, PayPal also introduced Fastlane, which significantly speeds up the online checkout process by allowing a shopper to skip several steps. No password is needed—Fastlane uses other authentication methods—and the shopper’s billing and shipping information is already securely stored. They can go to their cart and complete the transaction in one click. This is similar to Shopify’s handy Shop Pay functionality. PayPal said in February that 25% of the shoppers using Fastlane were new to PayPal, and more than 50% were PayPal users whose accounts had been inactive.

PayPal is trying to catch up in NFC payments, too. In September, the company began airing a commercial in which Will Ferrell discovers PayPal’s tap-and-go and cash-back features while singing “I want to pay with you everywhere” to the tune of Fleetwood Mac’s “Everywhere.”

L’Oréal also benefited from a subdued competitive environment during the pandemic. Virus-related disruptions resulted in fewer new brands entering the market, and its rivals held off on new product launches. These conditions enabled L’Oréal to increase its sales at 1.7 times the rate of the industry. Now, its growth is settling back to its pre-pandemic level of around 1.2–1.3 times the industry rate.

But global beauty sales are still expected to increase 4.5% this year, compared with a rate of 5% for the years leading up to the pandemic. And L’Oréal is still outpacing the industry—the company’s sales climbed 5.1% in 2024.

Shares of L’Oréal command a sizable premium over other European Consumer Staples stocks, and that gap has widened in recent quarters. For example, Bloomberg data show that L’Oréal’s forward price-to-earnings ratio is more than 50% higher than the average for European staples. Some analysts focused on the European sector thus view L’Oréal’s valuation as relatively expensive and have turned bearish on its stock.

However, that narrow lens is less relevant to a global investor. When you expand an analysis of L’Oréal’s valuation to include US companies, it starts to look more interesting.

For a long time, the average price-to-earnings ratios for US and European Consumer Staples stocks tended to trace each other—until the end of 2023, when their valuations began to diverge. US Consumer Staples are now much more expensive than their European counterparts.

The similarity in the names isn’t a coincidence. That’s because, until 2000, the two Novos were one company. More than two decades after their corporate breakup, their headquarters, some labs, and production plants remain near one another (and both businesses continue to be controlled by a holding company for the Novo Nordisk Foundation).

“Novo” is Latin for new, and that’s what industrial enzymes were in the 1940s, when the original company introduced its first enzymatic products—one used to clean leather hides before they are tanned, and another for textile desizing, in which amylase breaks down starch agents to prepare fabrics for dyeing. The company was also a pioneer in synthetic biology, launching its first enzyme derived from genetically engineered microorganisms in 1987; lipase is still used today in detergents to help remove stains deposited by a greasy burger or pizza, and some of Novonesis’s biggest customers are Procter & Gamble and Unilever.

Enzymes are proteins that act as catalysts to speed up chemical reactions. For example, they are naturally found in the stomach and help break down food into tiny particles that can be converted into energy. In industrial settings, they are used as a substitute for harsh chemicals, which tend to require higher temperatures and pressure to carry out the necessary reactions. The use of enzymes also prevents unwanted reactions, as each type of enzyme binds only to specific molecules, thus leaving behind less waste than traditional chemicals that could give off hazardous byproducts. And industrial enzymes are a good business: While they comprise a small portion of a manufacturer’s overall cost of goods sold, the switching costs can be high, which gives enzyme suppliers pricing power over their customers. This is especially true for Novonesis, given that it is the dominant player in the oligopolistic niche market.

Another source of strength for Novonesis is its research-and-development strategy. The company has historically spent much more on R&D as a percentage of sales than competitors such as DuPont, DSM, and BASF, and that investment has produced a number of important products and thousands of patents over time. In 2024 alone, Novonesis introduced 45 new products, and 30% of its revenue comes from products launched in the past five years.

Novonesis recently acquired its Danish competitor, Chr. Hansen, which supplies bacteria cultures and probiotics used in products such as yogurts. (The US$12.3 billion deal closed last year.) There are product-level advantages to putting the companies together. For example, adding protein to yogurt changes the texture and taste, but Novonesis can now combine products to better address those issues. Management has said it sees similar synergistic opportunities around food preservation for baked goods.

Revenue growth had been bumpy for Novonesis over the past decade, but recently it has gained positive momentum. Pro-forma sales climbed 5% in 2024 (8% when excluding the negative effects of currencies and acquisitions), and the company projects organic sales growth of 5% to 8% this year. The long-term outlook also looks attractive because there remains a significant opportunity for enzymes to displace chemical additives in industrial production processes. The clearest example is detergents, as producers emphasize products that are more water efficient, enable washing machines to run at lower temperatures, and use fewer harsh chemicals. An expanding middle class in emerging markets could also increase the demand for better household-cleaning products.

It’s also at lower US tariff risk than some other European companies, given that Novonesis has R&D and production facilities in the US, and most of its demand there is fulfilled by local production.

It may not be one of the companies shaking up the splashy market for weight-loss solutions, but Novonesis happens to possess a lot of the traits that investors admire about its better-known sibling: cutting-edge research that allows it to succeed in high-growth markets. For Novonesis, that growth strategy is beginning to bear fruit.

Each of these risks have sunk banks in recent years. During the 2008 financial crisis, many banks failed because of credit risk when borrowers defaulted on their mortgages. In 2023, Silicon Valley Bank, Signature Bank, and then First Republic Bank all collapsed when duration risk fears sparked bank runs (to be clear, they were not the only banks with underwater assets, they were the only banks to suffer a run). And liquidity risks often come up when there are already other problems, such as the bank run at Washington Mutual that turned long-term issues into an immediate crisis and the largest bank failure in US history.

Regulators create rules to keep banks and their management teams from taking too much risk, but it is only a partial solution. No amount of regulation can force banks to make only good loans. So, one idea that’s been getting some attention recently is that maybe banks shouldn’t make loans at all. A “narrow bank” would take in deposits and provide payment services but wouldn’t make loans or take credit risk. Instead, it would hold all the cash it brought in as deposits at the central bank or buy short-term government bonds, meaning that there is basically no credit risk. There is almost no duration risk because the assets are short term or even overnight, like the deposits. And there is no liquidity risk because the assets are cash (or as liquid as non-cash assets can be).

To my knowledge, none of the tens of thousands of banks around the world operate this way. In part, that is because the regulators who are supposed to keep the current banking system safe don’t seem to trust the idea. In February 2024, for example, the Fed rejected the application of a bank called The Narrow Bank, which had been trying to secure a banking license for nearly a decade. Indeed, some have even argued that narrow banks actually make the system less secure. Because a narrow bank would be so secure, depositors, during a panic, might move their money away from conventional, risk-taking banks to narrow banks, exacerbating the crisis.

Today, the closest company to a narrow bank anywhere in the world isn’t even a bank. Wise is a financial-services firm with more than 11 million customers that started as an international money transfer service. Wise lets customers leave their money inside of the service as a pseudo deposit, similar to keeping money in Venmo or Cash App. Many customers who travelled frequently or did a lot of international business realized it was more convenient to leave their money with Wise than constantly shift it in and out of local banks. Today, Wise holds about £16 billion in deposits on behalf of its customers, investing the money in government bonds and other low-risk assets (see chart). And in fiscal 2024, it made £360 million in net interest income on those deposits, compared to £1 billion from its main business.

For large depositors, Wise could be safer than a traditional FDIC-insured bank even though it is not one itself. A small business in the US with US$5 million deposited at a bank is insured only up to US$250,000 by the FDIC.

Unlike banks, Wise does not lend out its customers’ money but instead holds it in cash or cash-like, safe assets.

Source: Wise PLC. and Harding Loevner analysis.

The investigation began after Chedraui, the third-largest retail chain in Mexico, accused Walmex of abusing its market power, using its size to coerce suppliers into giving it better prices and terms than they could to competitors, especially smaller ones. A Reuters story amplified the claims of pressuring suppliers, finding that some of those suppliers removed their products from Amazon as a result of Walmex’s pressure. Cofece’s investigation concluded that Walmex abused its position over the course of 13 years and broke anti-monopoly laws in the process.

The common assumption with big retailers such as Walmex is that their power and profits are a result of their scale and ability to lower prices on products that will attract and keep customers. And that is true and well understood. But what is not well understood is how its scale gives the company that power. Retailers don’t make their money on consumers. They make it on suppliers.

There is an intricate game between retailers, suppliers, and customers. Customers have bargaining power with retailers, because they will go wherever prices are the best. Therefore retailers have little bargaining power with their customers—but can have a lot of bargaining power with suppliers. Successful retailers become the conduit to customers and therefore gain an advantage over their suppliers. Retailers use that bargaining power to force terms from suppliers that advantage the retailers, and offset the retailers’ weak bargaining power against fickle buyers. What Cofece concluded was that Walmex was abusing that power.

The Cofece ruling imposed a ban against some of Walmex’s practices, such as retaliating against suppliers based on their contracts with other retailers. That should raise the bargaining power of suppliers against Walmex. The impact, though, is hard to predict. If the ruling stands, it would effectively reduce Walmex’s competitive advantages, which would hurt the bottom line. That won’t be an issue if it wins on appeal, though it will likely be several years for that to play out.

In the meantime, there are two wrinkles. For one thing, Cofece, in its oversight role, will determine whether the company is violating the ruling. For example, the ruling does not appear to ban volume-based discounts. But if Walmex demands a 50% discount from a supplier on, say, 10 million units of an item, will that constitute a violation? For now, only Cofece gets to answer that question. Moreover, the ruling applies specifically and only to Walmex. In effect, Cofece is raising the bargaining power of suppliers…against Walmex only.

That is the real reason Walmex is appealing the ruling rather than just paying the small fine and moving on. Because ultimately what makes a retailer successful is the leverage it holds over its suppliers, not its customers.

Although the techniques employed by DeepSeek are not entirely new, the company’s execution is considered extraordinary for how it significantly reduced training and inference costs. The company reportedly not only used fewer and less-powerful chips but also trained its model in a matter of weeks. There is considerable debate regarding the accuracy of the roughly US$6 million training cost figure that has been reported as well as speculation that DeepSeek may have utilized rival models such as Meta Platforms’ Llama, Alibaba’s Qwen, and OpenAI’s GPT-4 in a process known as distillation to create a student model. Therefore, it may be that the true training cost of DeepSeek’s model is understated.

There is the risk that more efficient AI models may reduce the growth rate of demand for computing power and, in turn, the pace of investment in AI infrastructure. But while investors seem to have equated improved efficiency with a diminished need for AI infrastructure, the history of technology shows that better performance and lower costs typically lead to wider adoption and faster growth over time. AI could follow the same trajectory.

Among the likely beneficiaries of cheaper, more efficient AI are the software providers building applications on top of these models. As more users discover the advantages of AI, increased demand for user-facing software also helps the cloud-services companies that provide the computing power and data storage. However, the commoditization of large language models may negatively affect creators of these models, particularly those that have eschewed open-source technology.

Also, greater computing efficiency may lead to higher volumes of cloud data and thus the need for more data centers and electrical power over time. This would benefit infrastructure providers on the condition that higher demand more than offsets price declines.

The field of AI is still evolving, and further advances in the technology and the methods for developing it are likely. Innovation is always disruptive, but on the whole, DeepSeek’s breakthrough would seem to be a good sign for the industry.

Chinese consumers have, through necessity, become very price conscious. Pingti, which translates literally to “horizontal replacement,” refers to a consumer buying a cheaper product that does the same thing as a more expensive one. Trading down, as we might say in the US. In the language of Porter forces that we use to analyze industry dynamics at Harding Loevner, it is new entrants matching the functionality of incumbents such as Adidas at better prices, leading to a more intense rivalry between companies.

And pingti is easy to do. There are websites, such as 1688.com (part of Alibaba) or Made-in-China.com, originally set up for manufacturers and exporters to do business with their wholesale buyers. Since most of the manufacturer supply chains are in China, it is therefore relatively easy for consumers to find cheaper versions of products from retailers such as Adidas, Li Ning, and others. All a consumer needs do is find an item they like in a store, look at the label to see who manufactured it, and go find the lower-priced, off-brand versions online.

“Customer mindsets have changed considerably in Mainland China in the post-COVID era,” Pan Ning, the chief executive of the Chinese division of Japan’s Uniqlo, said in July. “We are seeing a new set of consumer values centered on pingti.” The trend has been especially sharp among the younger generations, he said. Given its focus on low-priced goods, Uniqlo could also be seen as a beneficiary of the trend, which would help explain why its performance held up better in China than some other retailers.

Adidas has been adapting. The company has been expanding aggressively into lower-tier cities, where it is selling more-affordable versions of its products. It has also, as much as possible, tried to organize its Chinese operation as if it was a local company. The company has teams that are focused on discerning what local buyers want, and building products based on those insights. Once they come up with those ideas, the product designs are approved by the China team, not the home office back in Germany. The efforts appear to be yielding results. Its sales in China rose 9% in the third quarter of 2024, to about US$1 billion, its best quarter since the beginning of 2022.

While pingti was sparked by the economic downturn, it may outlast the eventual recovery. Chinese consumers have become more sophisticated. They no longer view price and brand as the sole markers of quality. Increasingly, they are seeing that lower-cost products can have the same essential quality. A good example of this is the Japanese brand Mont-bell, which has been opening new stores and expanding its presence in China. Mont-bell is a well-known replacement for Arc’teryx, the premium brand owned by Anta.

This dynamic doesn’t have a catchy name in the US, but pingti has been visible here: off-price retail stores operated by TJX such as TJ Maxx and Marshalls saw sales rise after the 2008 financial crisis. Even as the economy recovered after the crisis, those new customers stayed loyal and that outcome may hold in China, too. For that reason, manufacturers and retailers in China may have to account for the trend even after the economy eventually rebounds.

A related metric is the ratio of net property, plant, and equipment (PPE) to gross PPE. A high ratio indicates the outgoing CEO heavily invested in the company’s facilities and equipment. A declining ratio suggests that the outgoing CEO invested little to maintain or upgrade them—or as some might say, management was “sweating the assets.” The latter scenario constrains the next CEO’s ability to implement significant changes. Think of it as the difference between buying a well-maintained house and one needing extensive repairs: The same renovation budget would achieve far more in the well-maintained house, while at the fixer-upper the money would need to be used for necessary but less exciting projects.

Take Intel as an example. Once the leading chip designer and manufacturer, the US company has fallen behind rivals such as NVIDIA and TSMC. Its CEO was pushed out in December, and shareholders are hopeful that his replacement can restore Intel to its prominence. The challenge is that Intel is not easily changeable. Chipmaking is intricate and capital intensive, and the process relies on specialized tools, machinery, and facilities that last for many years. According to its filings, Intel’s manufacturing assets can remain in service for up to eight years, with some facilities lasting over two decades. Despite recent capital investments, the net-to-gross PPE ratio hasn’t dramatically improved, suggesting legacy infrastructure still defines the firm’s capabilities. A quick pivot is tough.

The other question to ask about an incoming CEO is whether change at the company would be a good or bad thing. Generally speaking, change is more welcome for a business that is in a position of weakness, such as Intel, than one that is in a position of strength, such as Schneider Electric. Schneider’s competitive edge in providing electrical solutions continues to be a source of sustained profitable growth, particularly as it sees heightened demand from data-center customers. While the industrials company replaced its CEO in November, investors are unlikely to want drastic changes. Fortunately, with Schneider’s long-lived assets providing operational consistency, the new CEO has limited room—or need—to meaningfully change things.

Consider the potentially very different implications of a new CEO for a company such as Intel compared with a company such as Schneider:

The concept behind autonomous vehicles is simple—cars that are controlled by onboard computers and sensors rather than a human driver—but the technologies behind making them work are far more complex. Autonomous vehicles replace the human driver with a highly advanced system of hardware and software. The most promising ones incorporate a system called LiDAR (Light Detection and Ranging). LiDAR systems flash a laser and record the time it takes that light to hit an object and return to measure distance and the shapes of objects (very similar to radar, which uses radio waves in the same way). A LiDAR system can measure data points down to the centimeter. An AV’s operating system takes the information provided by the LiDAR system to map out the car’s path and then control the steering, acceleration, and braking.

The concept of self-driving cars has been around for a long time. A man named Francis Houdina experimented with a radio-controlled car all the way back in 1925, driving it down New York’s Fifth Avenue. In the 1950s, a famous one-off concept car called the Golden Sahara utilized a rudimentary, LiDAR-like self-driving system. A team at Carnegie Mellon built its first autonomous vehicle, dubbed Navlab, in 1986. Much like humanoid robots, the problem wasn’t the technology but the cost; it was too expensive to be commercially feasible. Recent advances in hardware and software have driven the costs down and given these companies a real chance at making and selling AVs at scale and profitably.

They’re not there quite yet. Israeli company Mobileye has dominated the market for driver-assisted software, but has struggled to sell to auto makers more advanced systems that enable higher levels of autonomous driving. Tesla earlier this year delayed the initial launch of its robotaxis (the company dubbed them “cybercabs”) from July to October, and the October “launch” was just the reveal of a prototype. Production vehicles won’t be available for sale until the first half of 2025 at the earliest (and Elon Musk is notorious for his unmet timelines). Just getting the cars onto the road is not sign of success, either. A Cruise robotaxi hit and dragged a pedestrian in San Francisco last year; subsequently California regulators rescinded the test permit they’d given the company just months earlier. Stories like that highlight what could end up being a material obstacle: regulators who will need to be convinced these cars are safe before they approve any more of them for public roads, or suspend ones that have already been approved.

The AV competition splits pretty cleanly between two kinds of companies, auto and software companies, and in this competition, experience making cars is less important than experience making software. Because the key part of the vehicle is the autonomous operating system, companies such as Baidu and Alphabet are on equal footing with Tesla and GM.

The fact that software companies such as Alphabet and Baidu are already such prominent players in this market proves the point. The ”picks and shovels” of the AV industry won’t be the car makers, we think, it will be the software makers. Whoever builds the best operating system—which means one that can both be completely effective and pass muster with the regulators who will allow it on the roads—will be able to sell it to any auto maker. The car itself will almost become a commodity while the operating system will be the key differentiator. It will be analogous to mobile phones. There are really only two operating systems: Apple’s iOS and Alphabet’s Android. Autonomous vehicles are likely to repeat that development. We imagine the landscape for AV operating systems will end up looking like the landscape for mobile-phone operating systems. Which means this competition is virtually all or nothing, where only one or two companies will end up with the dominant system that everybody else is forced to use.

EV makers are beginning to highlight their use of CATL batteries as a way to differentiate their cars from the competition. Companies such as Zeekr have been advertising their use of CATL batteries, much like PC makers used to do with “Intel Inside” stickers to differentiate themselves in a crowded, intensely competitive market. Like Intel’s chips in the 1990s, CATL’s batteries have consistently pushed the envelope in technological advances, including industry-leading charging speed and density, which enable total ranges of more than 600 miles. These advanced batteries, called Shenxing and Qilin, now account for roughly a third of its EV battery shipments. Earlier this year, CATL started offering special service stations exclusively for owners of cars with the Shenxing batteries.

In August, CATL took that idea—that the battery in your car matters, more than you think—one step further. It opened a showroom in Chengdu, in Sichuan province, displaying nearly 100 models of EVs from 50 different car brands that use CATL batteries. The only major car maker not featured is BYD, a vertically integrated EV maker that has insisted on producing its own batteries, effectively becoming CATL’s main competitor.

As if to illustrate just how important an EV battery really can be, a Mercedes EV caught fire in Incheon, South Korea, just a week before CATL’s showroom launched. Thankfully no one was hurt, but the fire ended up damaging or destroying dozens of cars in an underground garage. After that, Korean regulators started requesting automakers disclose their battery suppliers.

The car in question had a battery from Farasis, a Chinese semi-captive supplier to Mercedes and one of a long tail of unprofitable, substandard rivals to CATL, which has the best safety record in the industry among major makers. CATL sets aside the most reserves for quality/warranty payouts (a conservative accounting approach), yet uses the least amount of those reserves in practice.

Batteries are not as commoditized as many think. These recent events could lead more consumers, and investors, to that conclusion.

Additionally, in September, Sony pulled its shooter game Concord off the market two weeks after it was released, saying the game “didn’t land the way we’d intended.” Like Ubisoft, the Japanese conglomerate issued refunds. There is also speculation that Grand Theft Auto VI, which has long been in development at Take-Two Interactive, may get delayed until 2026—13 years after the previous installment in the series was released.

News such as this tends to restart the debate over whether gaming is just a bad industry for investors. In fact, the industry structure seems challenging. The barriers to entry are somewhat low, and the number of substitutes and competitors—including new entrants in China and South Korea—are both high, factors that generally limit a company’s ability to generate sustainably high profits. But the charts above do not depict such a terrible situation. Rather, there may be some weak players, but there are clearly others that have managed to become more profitable and stable over time.

One way some publishers have done this is by building diversified franchises, which leaves them less prone to revenue lulls and limits the impact that a failed release has on the company’s overall finances. Electronic Arts, for example, has a number of successful franchises, including FIFA, Madden, Apex Legends, and The Sims. Activision Blizzard, in addition to Call of Duty, has World of Warcraft, Candy Crush, and Overwatch, among others—a collection Microsoft found so attractive that it acquired the company for about US$70 billion last year (Microsoft owns Xbox). Other diversified players include Tencent and NetEase, both based in China.

Another way that some in the industry have generated more consistent revenue and profits is through add-on digital content. Twenty years ago, video-game publishers were essentially consumer-goods companies. Their games were a physical product, and manufacturing, shipping, and selling these discs carried a significant cost. Playing them was also a one-off experience. Digital technology changed all this, with players increasingly downloading games rather than purchasing new or used discs. This has allowed game developers and publishers to boost the revenue generated by a game after it’s released by offering so-called expansion packs for online purchase, as well as virtual cosmetics to enhance a player’s avatar and virtual currencies. For example, Call of Duty: Black Ops 6 was released October 25, but you can almost guarantee that at some point, perhaps in six months or so, an expansion pack with additional content will become available. Extending the useful life of each game in this way relieves the pressure on developers to quickly crank out the next one. In the second quarter of 2023, Activision Blizzard’s last earnings report before its sale to Microsoft was completed, the company said that downloadable content and micro-transactions accounted for 63% of its net bookings.

The industry’s earnings consistency also has to do with the other types of companies involved in gaming: console makers, such as Microsoft and Sony, as well as businesses that sell tools and services to developers and publishers. One example is Keywords Studios, a provider of game-ready art and quality-assurance testing. Gaming platforms and toolmakers capture a portion of the industry’s revenues, which means they benefit from the overall growth without taking on the risk of any single game.

The recent challenges at Ubisoft and other companies reflect two industry trends that are worth watching. One is that game development has slowed. It’s unclear whether this is due to increased remote work, developers missing deadlines because they’re trying pack more interactive features into a game, pushback against the industry’s overtime habits, or perhaps all the above. In any case, these would seem to be temporary challenges.

Another is that some gamers seem to be spending less money on the hobby, which describes a change in user habits that could be more permanent. However, there’s not enough data to know if that’s the case. Instead, what we can gather from 30 years of financial data is that some gaming companies are built to ride out the rough patches, and so it would be a mistake for investors to write off the whole industry.

Daifuku, based in Japan, is a leading provider of cleanroom-automation solutions such as the monorail described above, which TSMC, Samsung, Intel, and others depend on to keep chip production on track. Each monorail extends 10 kilometers (6.2 miles) and is equipped with about 1,000 autonomous vehicles, called front opening unified pods (FOUPs). Typically, multiple cleanrooms operate together, with several monorail systems connected to one another. Each FOUP can travel 200 meters (656 feet) per minute. This speed is important because the photoresist light-sensitive polymers will degrade if the process takes too long. Dynamic routing, or the ability of the FOUPs to find the best path depending on the wafer’s condition while avoiding collision with other vehicles, is one of Daifuku’s key competitive strengths.

Because the company has only one significant rival, Murata Machinery, Daifuku enjoys strong bargaining power over buyers. (Samsung has a subsidiary that also does some of this work, but exclusively for its parent company.) And although the broader cleanroom-automation market is largely split between Daifuku and Murata, fabrication facilities, or fabs, working on the most cutting-edge technology—chips with features two and three nanometers small that are highly sensitive to contamination and degradation and require more precision—primarily turn to Daifuku. To create smaller transistors and other components, fabs not only must limit airborne and aerosol particles but also minimize machine vibrations and have other critical environmental controls. Daifuku is distinguished by its ability to engineer large, sophisticated projects such as these. Rather than focus on manufacturing all the disparate parts that go into an automation system, the company’s core competency is designing ways for various solutions to work together to address each customer’s increasingly complex needs.

Cleanroom automation isn’t Daifuku’s only business, although it may soon be its biggest. For much of the company’s 87-year history, its largest source of revenue has been factory and warehouse conveyor systems, including those used today by Amazon.com, Home Depot, and Walmart to swiftly transport boxes around facilities that can span several football fields. Its elevator-like systems of storage racks can also organize and retrieve crates stacked multiple stories high (a common setup in Asia, where land is limited, and warehouses tend to be tall rather than wide). Daifuku’s systems are also used in automotive plants to hoist cars through assembly lines, and in airports to move baggage.

Given the strength of its main businesses, and the increasing interest among companies, consumers, and regulators for more recycled material in products, TOMRA seemed primed for profitable growth as the go-to equipment supplier in this “circular economy.” From 2017 through 2021, the company compounded its earnings at a 13% rate. At a capital markets day in 2022 it promised more of the same—to boost revenues 15% per year while expanding margins.

But since that capital markets day, which coincided with turnover in the executive team, the company’s execution has faltered. Earnings this year are expected to be lower than they were three years ago. TOMRA was hurt by inflation, as existing contracts prevented the company from increasing prices to offset rising costs. The company was also the victim of a severe cyberattack, which cost US$20 million to remediate. In the food division, the company had to undergo a drastic restructuring after a series of acquisitions led to a disjointed operation.

World-class industrial businesses often have an ethos that focuses their work force and strategy on what’s most important—think Danaher’s famous Danaher Business System, Parker Hannifin’s Win Strategy, or Honeywell’s Accelerator program. The companies attribute their historical success to these systems because they focus decision makers on organic growth, profitability, and return on invested capital. We’d like to see TOMRA’s management take a similar approach and pay more attention to concrete financial metrics as opposed to buzzwords like “circular economy” or “food security.” Talking about cash-flow return on investment and other hard numbers with investors often encourages a more dedicated focus on them.

In the first quarter of this year, two of TOMRA’s three divisions lost money, yet on its earnings call with investors, the management team hardly referenced the dismal results. In fact, in the opening 30 minutes of the one-hour call, the CEO made scant reference to TOMRA’s lackluster financial results. The second-quarter results showed some promise; revenue for its reverse-vending machines rose 15%, an encouraging sign. Earnings, however, for the first six months fell from 2023’s levels.

While TOMRA has endured a challenging period, it still has a dominant position in a growing market, and some mid- to long-term drivers of growth are still intact. Jurisdictions from Quebec to Tasmania are planning to launch deposit programs this year and next, albeit others are rethinking plans; France for instance cancelled a planned bottle-recycling program. But Austria, Poland, Hungary, and Romania have also emerged as new markets. Additionally, the technology the company has developed for sorting recyclables can be leveraged for other uses; the company expanded into metal recycling in 2006 and mining in 2008 as well as food sorting.

The company says its machines collect more than 46 billion used beverage containers every year, and that collection rates in markets with deposit programs are higher than those with curbside-collection plans. TOMRA has managed to boost revenues steadily over the past decade, which points to a degree of stability. The new management’s missteps combined with the changing interest-rate environment and countries altering recycling plans have sown doubt. There is still a business and environmental case to be made for TOMRA. What we’re looking for are signs that the new management team is balancing those two things more effectively.

Perhaps because consumers in the US and elsewhere are accustomed to seeing “made in China” tags on their clothing, there is a misconception that China was always the domain of fast fashion—cheap, rapidly produced apparel that mimics catwalk trends. Traditionally, though, production and shipping lead times were longer in China, making it a more suitable supplier for brands such as Nike than retailers trying to quickly recreate viral looks. Leading fast-fashion retailers such as Inditex, the Spanish parent of Zara, and the British e-ecommerce site ASOS source only about 30% to 40% of their merchandise from China and often turn to countries such as Spain, Portugal, Morocco, and Turkey for styles requiring a quicker turnaround.

Shein and Temu, however, mostly rely on Chinese suppliers, and they’re able to offer low prices in part because of the method they use to get their products from those suppliers to shoppers. By shipping individual orders directly to customers, they avoid US import duties, which are waived if a shipment’s value is less than US$800. US lawmakers have called this an unfair advantage over American retailers, many of which do pay import duties to bring in inventory made overseas. Duties, particularly in countries with high duties, can be a meaningful component of costs, although their avoidance isn’t the only reason Shein and Temu can offer low prices. Another is the efficiency of their supply chains and logistics operations.

But while these relative newcomers have been popular with some consumers, there doesn’t appear to be much risk to high-quality retail companies. One reason is that the retail market pie is so large, which means the rivalry dynamics are such that many different strategies can succeed at the same time. For example, even after getting off to a gangbuster start in the US, Temu’s gross merchandise value represents just 1% of the US e-commerce market, putting it significantly behind Amazon at 38% and even Walmart at 6%. Its customer retention rate is also lower, suggesting that some shoppers may be drawn to the novelty of scoring a deal but aren’t becoming repeat customers. The Sensor Tower data below shows the proportion of Apple iPhone users who open the Temu app at least once a month after downloading and how that compares to other retail apps; the large drop-off in Temu visitors is true among Android users as well:

Tinubu had success turning around the economy in Lagos as its governor, and he wasted no time in tackling the country’s sclerotic economy. On the day of his inauguration, he abruptly ended a decades-old subsidy paid to oil companies to keep fuel prices low that cost the government US$10 billion a year, more than it spent on education or health care. He reorganized the county’s foreign-exchange system and abandoned the naira’s currency peg—it has fallen about 70% against the dollar since—and took other steps geared toward improving the economy and boosting tax revenue.

Since ending the subsidy, the price of oil (which is a strong source of revenue in Nigeria) has stabilized and credit-quality issues that plagued the energy sector have improved. Credit trends, for now at least, are stable. Two currency devaluations have removed a lot of the concerns about the previous monetary regime. So far, the reforms seem to be well received by outside observers. In early May, the ratings agency Fitch upgraded the country’s outlook to positive from stable, based on the initial success of Tinubu’s reform package.

There is good reason for investors to pay attention. Nigeria’s demographics are compelling: it has the sixth-largest global population, one that is young and fast-growing. Demographic tailwinds can be powerful over longer periods of time as penetration of banking and insurance services naturally expand as countries develop. Even amid the travails of recent years the country’s high-quality banks have fared well. Zenith and Guaranty Trust consistently post a return on assets north of 3%, compared to the more typical 1-2% earned by most banks around the globe.

Recent earnings might seem to indicate the hoped-for inflection point has arrived, and investors may also be enticed by the fact that Zenith and Guaranty Trust trade well below book value. But the earnings are somewhat misleading since both banks posted large and unsustainable hedging gains due to the naira devaluation. Nevertheless, their operating performance demonstrates their relative stability amid the volatility induced by the reform package.

A larger argument against the idea that an inflection point’s been reached is the reality that unwinding the mistakes of the past 8 to 10 years is going to be extremely difficult. Tinubu’s reforms, while necessary, have caused lots of pain and economic hardship. The currency devaluation feeds straight into inflation, which is running north of 30%. That, combined with the removal of the fuel subsidy and later an electricity subsidy as well, was too hard a pill for the majority of the mostly poor population to swallow. It seems likely that the next couple of years will be rocky and there may be credit-quality issues for the banks since credit stresses typically lag economic shocks.

Whatever the short-term pain, though, we do believe that the long-term picture is still positive, which is why we remain resolutely interested in Nigerian companies, not just the banks. The policy changes this new government has made will hopefully put the economy on a better growth trajectory and make the society more prosperous. In fact, we’d go so far as to say that Nigeria is lucky to have elected President Tinubu at this critical time. He appears to be the right man to have taken over the country given its precarious economic state. After his success in Lagos, it’s reasonable to have hope for his administration’s ability to turn the country around. But the path to better days ahead will likely be turbulent.

Fast fashion has become a roughly US$120 billion industry and is projected to grow to US$185 billion by 2027. Shein, H&M, Temu, Zara, and others have built worldwide businesses by creating cheap clothing in bulk as quickly as possible to respond to changing fashion trends, especially for a growing global middle class with more disposable income. The crux of the problem is that fast fashion tends to encourage what many consider an excessive amount of consumption. Traditional clothing companies might release new lines four times a year; fast-fashion companies do so weekly or even daily. The clothes are designed to be relatively cheap and worn for a short period of time before being discarded, sometimes after only seven or eight wears. Consumers are then expected to buy the next hot thing. The result? People are buying more clothes, an estimated 60% more, but keeping them only half as long as they would a more durable, higher-quality piece of clothing.

Those dynamics exacerbate the environmental impacts of an industry that is inherently polluting to begin with. Conventional textile mills generate about one-fifth of the world’s industrial water pollution. About 20-35% of all the microplastics in the world’s oceans come from microfibers shed by clothes as they are washed. The industry also produces veritable mountains of unsold clothing, which get burned or thrown into landfills strewn around the Global South. (Shein, for one, has argued that its business model better matches supply with demand and produces less unsold clothing).

Costco’s results for the quarter were healthy, but they also contained signs that industry dynamics are beginning to shift. For example, given that overall inflation was flat across the retailer’s product categories, growth in same-store sales was dependent on increased customer traffic.

At Harding Loevner, we believe that whether a business thrives largely depends on the competitive structure of its industry, which we evaluate using Michael Porter’s Five Forces framework. NVIDIA has an estimated 95% share of the market for AI chips. Its graphics processing units (GPUs), which were originally used to power video games, have been adapted to become the key engine of AI computation. The customer base for these chips is highly concentrated and primarily consists of “hyperscalers”—companies such as Alphabet, Amazon.com, Meta, and Microsoft that operate large networks of data centers with immense computing power and storage needs. More than half of NVIDIA’s data-center revenue for its fiscal quarter that ended in January came from large cloud-services providers, and one company alone accounted for 19% of last year’s total company revenue.

From the Porter perspective, a key factor in the bargaining power of buyers is their ability to backward integrate. All of the aforementioned customers have this capability, which threatens to diminish their reliance on NVIDIA. In our view, it’s only a matter of time before this manifests. In fact, Amazon, Google, and Microsoft each have unveiled custom data-center chips as potential substitutes for NVIDIA’s general-purpose GPUs. Because each server in a data center performs only a few specific functions that are repeated over and over, these custom chips, called application-specific integrated circuits (ASICs), are tailored to just the tasks the companies need them to perform.

Industry rivalry is also likely to increase. As the frenzied pace of spending on AI infrastructure eventually eases, companies will begin to look for less expensive options from new suppliers, which may include NVIDIA’s own customers.

NVIDIA is a high-quality company led by skillful managers. Its revenue more than doubled over the past fiscal year, and net profit rose fivefold. But the high potential for increased rivalry and backward integration among key customers suggest that it is unlikely NVIDIA can continue growing at these rates with these margins while maintaining its grip on 95% of the AI-chip market. The fragility of its competitive structure limits our valuation tolerance, which is why we exited the stock in the first quarter after holding it for more than five years.

While it is possible that NVIDIA will continue to exceed increasingly aggressive consensus earnings forecasts for some time, we own other more attractively priced businesses that allow us to invest in AI’s potential. For example, Microsoft and other providers of software and services have the advantage of being able to monetize generative-AI technology without regard for whose GPUs or ASICs are used in the future. Microsoft’s customers also don’t appear to have the ability or interest to try to backward integrate; in fact, AI functionality such as its Copilot chatbot are helping to lock in users. In general, as the applications for AI evolve, value is likely to shift from hardware, such as NVIDIA’s GPUs, to the software and services built upon this infrastructure.

Read more about our software and services holdings in the first quarter Global Equity report.

While US chip designer NVIDIA gets most of the AI press, TSMC manufactures virtually all of the high-performance AI chips designed by NVIDIA. The Taiwan-based company uses some of its most sophisticated manufacturing technologies for these chips, enabling them to be at the forefront of transistor density, speed, and energy efficiency. AI chip customers also increasingly depend on TSMC’s innovative capabilities in integrating and packaging the logic, memory, and input/output components involved. TSMC’s CEO recently projected that in four years about 20% of this company’s total revenue would come from AI-specific chips.

Samsung Electronics has received less attention from investors but is also carving out a niche for itself in AI-related semiconductor manufacturing. Samsung anticipates a rebound in their memory chip prices in 2024, in part due to the expected proliferation of on-device AI, a memory-hungry implementation, and is investing for longer term in the next generation of high-bandwidth memory chips integral to AI applications, promising to leapfrog competitors in performance. Samsung Electronics is also positioning itself to compete with TSMC in manufacturing advanced logic chips, using its capability of designing and manufacturing memory, logic, and chip packaging to unlock new opportunities for its foundry business.

Infineon, widely recognized for its strength in automotive and power semiconductors, is less directly exposed to AI themes but does have products supporting growth data centers and AI computing whose high-power demands is a good match for Infineon’s switched mode power supplies and voltage regulator designs that power high performance systems.

During 2023, Amazon overhauled the sprawling US logistics network that supports its online-retail operations, switching from a national system for filling and routing orders to a regional setup. The core change was to partition the US into eight main regions, assigning each region dedicated fulfillment warehouses, sorting centers, last-mile delivery hubs, and vehicles. By limiting the area each facility serves, the company doesn’t need as many connecting points along its distribution routes, and trucks are filled closer to their capacity and travel shorter distances. This helps Amazon save money on fuel, fleet upkeep, labor, and other overhead, as well as potentially reduce its environmental impact.

Amazon’s latest financial report in February revealed that fulfillment costs, which had soared to 16.4% of net sales in 2022, shrank to 15.8% of net sales in 2023 as the new system was implemented. As the chart below shows, it was only the second time in 14 years that this metric’s long ascension was interrupted. This time, we view it as an inflection point, and so the downward trend should continue.

Given the opaque relationship between the Chinese government and private companies, the exact terms of the exchange are unknown; Baidu’s move could result in the US putting the company on the restricted list that blocks US investments in companies that could be supporting China’s military. However, there is a plausible market-based reason for the donation: the quantum computer was expensive to build and maintain and the effort wasn’t profitable.

At a meeting with analysts in January, Baidu’s chief financial officer, Rong Luo, explained that the kind of “moonshot research” the quantum computer represented was too expensive and didn’t add to the company’s bottom line. Donating it helped cut costs and foster goodwill with the government. It was the kind of pragmatic move that the company had gotten away from in recent years.

Like their global peers, Chinese internet companies enjoyed years of growth and strong cash flows. That allowed them to expand beyond their core operations to invest in new business formats and technologies, such as Alibaba in offline retail and Tencent in fintech. They usually focused on creating scale even if the business itself was initially unprofitable. That goal depressed margins in the sector. But as growth has slowed, Baidu, Alibaba, Tencent, Netease, and other Chinese tech companies have been refocusing their operations on turning a profit.

At Baidu, this included bringing in Luo, who has a PhD in management science from Peking University and had been the CFO at TAL Education Group. During his tenure there, Luo played an important role in TAL’s transformation from operation-driven to data- and technology-driven, and from offline to online.

When Luo arrived at Baidu two years ago, some senior management didn’t recognize the importance of return-focused performance metrics. That has changed. Expenses have been cut. Research and development spending as a percentage of sales, which had been about 20% (and more than its peers), has been reduced. The company was making investments in innovative technologies, such as quantum computing, but there wasn’t likely to be a return on these investments in the foreseeable future. The effectiveness of projects is now measured on the return on investment and efficiency. Thus, the decision to give away the computer.

Now the company is focused on the parts of its business that are or can be profitable: its core ad business, cloud computing, and AI. In the highly competitive business of search, Baidu released an AI-based advertising platform called QingGe. The new platform essentially examines a users’ interactions with a chatbot and generates a new ad specifically tailored to them. The goal is to create a better advertising platform, with the goal of improving the click-through rate by 10x. Hitting that goal would add several billion in incremental RMB in 2024, the company said.

That’s one example of how management is employing AI, a crucial test. In 2017, Baidu was the leader among its peers in AI. It had Andrew Ng, the former head of Google Brain, as its chief scientist and signaled its intentions with its slogan “All in AI.” But Ng left the company in 2017.

Despite his departure, Baidu was the first among Chinese companies to launch a chatbot in 2019, shortly after OpenAI released ChatGPT. That chatbot, called Ernie, is not as advanced as ChatGPT 4.0 but is more advanced than its Chinese peers and many Chinese companies are using it to train their own more specialized models. In December, Baidu reported Ernie had more than 100 million users.

Even with a technological lead over its peers, Baidu hasn’t yet shown it can make AI a profit center the way some US-based companies have. It is testing monetization models, such as a subscription service, and fusing it into its search business, but neither investors nor Baidu itself can be certain how the AI efforts will affect the company’s profits (the company does not provide revenue projections for its AI business). What investors do know is that the company’s criteria for the success of that effort and everything else it does will be measured on the bottom line.

One might expect that a decrease in the number of babies would be a major problem for a company like Unicharm, which dominates the Japanese market for disposable diapers. But by taking advantage of other demographic trends that have opened new markets for the company, Unicharm has seen its overall business grow, even as its core market shrinks.

Unicharm was founded in 1961 as an insulation and building materials supplier and later began manufacturing feminine sanitary pads. In the following decades, Unicharm relied on its expertise in processing non-woven fabrics and absorbent materials made from petroleum-based polymers to expand into other markets beyond feminine care, especially diapers.

Unicharm grew to dominate the market for baby diapers in Japan, with an estimated 41% retail market share in 2023. But the product segment has become a challenging, low-margin business, partly because consumers have not been swayed to pay a premium for baby diapers with special features, and the decreasing number of babies being born has meant that Unicharm is fighting for its portion of a shrinking pie. Meanwhile in neighboring China, once a key market for the company’s baby diapers, stiff competition from local manufacturers has driven down retail prices.

While Unicharm says it will continue to make baby diapers, because of the positive and happy associations consumers attach to baby products, the company has increasingly focused on market segments that offer higher margins (if not the same positive feelings) such as adult incontinence products.

Japan now has the highest proportion of elderly citizens of any country in the world. About 29% of Japanese are over age 65, while more than 10% are 80 or older. As the population has gotten older, Unicharm’s adult products have become a larger part of its business. Adult diapers constitute about 33% of the company’s net sales in Japan today, up from an estimated 25% in 2018. Overall, the company derived one-third of its sales and 40% of its profits from Japan last year.

Adult incontinence products are a high-margin business because consumers are willing to pay a premium for products with special features, such as improved fit to prevent leakage. Furthermore, the company’s marketing campaigns have been successful at creating demand for its light and mini-incontinence pads by associating them with an active lifestyle.

The sort of quality companies that we look to invest in have, among other characteristics, strong management teams who can respond to changing market and industry dynamics. Unicharm’s ability to leverage its core competency in absorbent materials in different markets is an example of how managing change can help a business thrive over time. And Unicharm might even find a way to help address Japan’s aging demographics—it’s recently introduced a new feminine hygiene product that monitors ovulation to help women plan their pregnancies.

It’s an expansive deal, beginning in January 2025 and running for ten years. Netflix will air Raw in the US and internationally, and will have the rights to all WWE shows outside the US. It also may be a harbinger of what’s to come in the world of streaming services and sports. Over the next several years, the contracts for broadcast rights to the NBA, NFL, NHL, and MLB in the US will come up for renewal, as well as the domestic and US rights to top soccer leagues such as the Premier League (UK), La Liga (Spain), Seria A (Italy) and Bundesliga (Germany). The streaming services will be serious contenders in the bidding for all of them.

Netflix has previously declined to jump into sports. “We’re not anti-sports, we’re pro-profits,” co-CEO Ted Sarandos said last year. Making a major foray into sports, even if it’s a scripted one, may be a signal that Netflix is seeing—or at least planning for—a ceiling in terms of how many new subscribers it can attract without sports.

The biggest problem with sports programming is that while it does the things a broadcaster wants it to—steadily bring in dedicated viewers—over time the rights can become addictive for companies. If a sport increases in popularity, that improves profitability and hopefully leads to good returns on the investment for the rights holder. However, those attractive returns also lead to a higher renewal contract, which resets profitability at a lower level even when competitive dynamics are benign.

The problem for Netflix is that the sports-rights competition is not benign. Netflix will be competing with Paramount, Peacock, and Disney, as well as big-tech companies that have completely different strategic goals. For tech companies like Amazon and Apple, for example, “success” in terms of sports programming is more about e-commerce or hardware sales than subscriber profitability.

Netflix has spent roughly US$17 billion a year on content since 2021, a level it plans to maintain for the foreseeable future. The company said that the WWE deal would not add to its content budget, meaning it expects to spend less elsewhere. On one level, then, this is not an expensive bet. WWE already has a fanatical global audience and NBC is paying nearly US$300 million annually just for the US rights. Maybe US$500 million a year for global rights is a bargain.

Given Netflix doesn’t have any legacy networks to protect like some of its competitors do, it sidesteps a lot of the messy problems in the industry. But Netflix has its own set of pressures with which to contend. In 2023, Netflix’s subscriber base rose to 260 million, which was a record. Its growth rate, however, is slowing. In the past two years, the subscriber base is up 17%. In the two years prior to that, it rose 33%. Netflix used to be valued and treated like a tech stock. These days it is valued and treated more like a media company.

But one Japanese company in the Information Technology (IT) sector has taken an unusual approach to reinvigorating the labor market. SHIFT, which provides software-testing services, is poaching and re-training workers from other industries, turning them into IT pros. As Prime Minister Fumio Kishida advocates for the reskilling of workers and for companies to increase wages, SHIFT’s system is an example of how companies could help put a dent in Japan’s entrenched model of lifetime employment, helping the country dig out of a decades-long deflationary environment.

Among its labor-market challenges, Japan faces a major shortage of IT engineers. Of the 75 million people in the country’s working population, only about one million are IT professionals. That’s a lower proportion than in other developed countries such as the US and Germany. But SHIFT estimates that 10% of Japan’s non-IT workers aspire to join the space.

Software testing is also one of the most labor-intensive processes in software development, as it entails the time-consuming work of manually looking for bugs within code and making sure an application works as intended. For firms that struggle to find experienced programmers to begin with, having those expensive programmers do testing doesn’t make sense. Some of SHIFT’s biggest customers are consulting firms that would rather have their own high-paid employees focus on their main business of consulting while outsourcing something like software testing to save money. For example, the average salary at Nomura Research Institute (NRI), an IT-consulting firm, is double the average at SHIFT:

Perhaps most intriguingly, NVIDIA is a leading company in the biggest tech-driven trend to come along in years, artificial intelligence. It makes the chips that power AI-based computer systems, and it sells the software and services companies need to operate their own AI programs. And NVIDIA was a big company before AI took off.

As if to emphasize the point, this week the company posted third-quarter revenue of US$18.1 billion—tripling from a year ago and setting a company record. Earnings were also a record at US$10 billion.

NVIDIA meets all our core criteria for a quality growth company: It has a sustainable business and a competitive advantage, solid management, and financial strength. Over the last six years, its revenue has grown from nearly US$7 billion in fiscal 2017 to just under US$27 billion in fiscal 2022.

But as NVIDIA reaches dizzying valuations—it currently trades at 118 times earnings—we can’t help thinking of another good technology company whose stock once commanded nose-bleed premiums: Cisco Systems.

Cisco makes the networking equipment that enables much of the internet. As the World Wide Web was taking off in the mid-1990s, Cisco Chief Executive John Chambers was telling people a new era of computing was coming, and the company’s hardware played a big part in making that new era real. Between 1995 and 2000, its revenue surged 850%, from about US$2 billion to US$19 billion.

Cisco stock did even better, growing an incredible 3,800%, from US$2 in January 1995 to US$79 in March 2000. The title of its 1999 annual report exhorted investors to “Capture the Momentum.” Many took that advice. In March 2000, at the height of the dot-com bubble, Cisco became the most valuable company in the world, with a market capitalization of more than US$500 billion.

Then the dot-com bubble burst. Cisco shares fell 88%, dropping from US$79 to a low of US$9.50 two years later. That drop wasn’t the result of any large change in the company’s performance. Revenue was nearly US$19 billion in fiscal 2000, US$22 billion in 2001, and about US$19 billion in 2002. But the premium that investors were willing to pay for the company disappeared as the hype around the web faded.

Over the next two decades, Cisco kept growing. Its sales in 2022 were nearly US$52 billion, more than double the US$19 billion it hit in 2000. Yet its stock has never traded as high as it did on March 1, 2000. Today it’s at roughly US$53. On a total return basis, it took 20 years for Cisco stock to recover from the dot-com crash.

So, are there lessons we can learn from Cisco as we look at NVIDIA? Just as Chambers argued 25 years ago that a new era of computing was at hand, Huang has been talking up AI. Just this month he said “In the last 40 years, nothing has been this big. It’s bigger than PC, it’s bigger than mobile, and it’s gonna be bigger than the internet, by far.” The enthusiasm for AI’s future has led to very high valuation multiples for NVIDIA, although not as high as Cisco reached in 2000.

The possibilities are mind-blowing. The DRC, with a population of 112 million and situated almost precisely in the middle of the continent, is the second-largest country in Africa by landmass and the 11th largest in the world. The Congo River, which winds through the country and is the second largest in the world by discharge volume, has the potential to produce 100 gigawatts of clean electricity, of which less than 3% is currently being harnessed.

With the river and its tributaries coursing through the countryside, the DRC has an estimated 80 million hectares of arable land, enough to produce food for two billion people. The DRC also has abundant mineral resources. A recent UN study estimated its untapped mineral reserves were worth US$24 trillion. The DRC holds 46% of the world’s cobalt reserves, 23% of the world’s lithium reserves, and 4% of the world’s copper reserves. These are all commodities essential in the effort to enable the global transition to green energy.

Property Deleveraging

Disney and Charter clashed over a new cable agreement. Charter complained about the high fees it paid for ESPN, FX, and other Disney channels while Disney was funneling premium content to its own Disney+ streaming service. Eventually, Disney’s channels disappeared from Charter’s Spectrum TV service, and Spectrum stood firm: it was prepared to give up Disney’s networks, seemingly forever.

It didn’t come to that, though, and the terms of the agreement don’t sound monumental: Disney+ will now be an option for Charter subscribers for an additional fee. Disney gets new eyeballs, but hands over a chunk of its fees to Charter, which had faced a seemingly inevitable revenue decline from its cable business due to the rise of streaming. The agreement offers a reprieve from that.

However, the terms of the resolution show that Disney as a supplier had far less bargaining power with its customer than it had assumed. It was forced to make material concessions that will affect its transition to a full-streaming model. Make no mistake, Disney backed down, and that loss is going to hurt its bottom line in the long run. Moreover, if the world’s largest entertainment company couldn’t use its muscle to get what it wants, then all the other networks are going to face the same dilemma when their carriage agreements come up for renewal. Disney’s capitulation could signal a major shift in the entertainment industry.

For decades, “content is king” was the mantra of the media industry. While cable companies were the primary distributor of content made by studios, the ultimate pricing power was always in the hands of the studios, especially studios like Disney that owned must-see kids, sports, and movie content. A cable provider could not attract subscribers without these must-have channels. As a result, even though there would be heated negotiations over fees, ultimately cable providers would yield to the studios’ demands.

The studios’ leverage has been weakened by the very thing they are focused on: streaming. Streaming subscribers often cherry pick services, signing up for a short time if there’s a must-see program and then canceling. For Disney and the others, there is a fundamental mismatch between costs and revenue duration. Streaming platforms need a constant flow of new programming to limit subscriber churn, but that pushes operating costs up into the billions. Not all of them can afford it.

It’s possible Charter perceived this, or just felt like it had nothing to lose. It was willing to take a loss of subscribers upfront by terminating its Disney arrangement rather than bleed subscribers over a number of years if it didn’t challenge Disney, given either course would likely lead to the same terminal point, just in different time frames. Charter’s hard stand showed the flaw in the Disney plan.

The content king may not be dead but it definitely seems like he holds less power.

Ozempic being used as a substitute for a different medical treatment goes beyond Novo Nordisk selling the same drug under different names. Ozempic could also serve as a replacement for people contemplating bariatric surgery, since it stands to reason that anybody considering surgery would likely be willing to try a less-invasive treatment. Recently, Intuitive Surgical noted a slowdown in the growth of bariatric surgical procedures because of the increasing use of drugs such as Ozempic. “Some customers have indicated that they are seeing increased patient interest in weight-loss drugs,” Intuitive chief financial officer Jamie Samath said in July.

Given that obesity is the most important reversible factor contributing to sleep apnea, there’s a chance Ozempic and similar drugs could become a treatment for that, too. Eli Lilly, for instance, is running a study of its type 2 diabetes medication Mounjaro, which is in the same drug class as Ozempic/Wegovy, as a possible treatment for sleep apnea. Results are due in March 2024.

That could be a problem for a company like ResMed, which makes sleep-apnea treatments with its CPAP (continuous positive airway pressure) therapy. These are machines that force air through the patient’s nasal passages, enabling better sleep. For patients, the issue may go beyond just which treatment is more effective.

Cost will be a big factor. Medicare and private insurers cover Ozempic for its commercially approved use, diabetes, but not for weight loss or other off-label uses. That could be important because the cost of Ozempic can be considerable (without insurance, it can cost about $1,000 a month). For sleep apnea patients, ResMed products may end up delivering the same or better results at a much lower price.

ResMed CEO Michael Ferrell in fact seems to welcome the new attention, since sleep apnea tends to be an underdiagnosed condition. Ozempic and similar drugs could bring potential patients “into the funnel,” he said on a conference call, while issues like cost and side effects will limit any hit to sales caused by the drugs. “That’s good for us, too,” he said.

The big-picture view suggests that Meta is building a hub for user-generated content (UGC) discovery and distribution that aims to go head-to-head with the likes of YouTube and TikTok. In the world of unlimited content, the power shifts from creators to distributors and curators. In this scenario, the ultimate constraint on demand for content is the totality of leisure man-hours; Meta’s goal is to optimize how to fill those hours with the most compelling content.

Generative AI could also help transform Meta’s main money-maker, its advertising business. The digital-ad market comprises roughly 70% of the overall advertising market, and is dominated by two firms, Meta and Alphabet. Finding a way to expand its market share has been a stumbling block for Meta. However, with these new AI tools, Meta may have found a way around that constraint. It may not be able to take much more share of the advertising market, but it could start taking share of the advertising-creation market. On the call, Zuckerberg specifically mentioned the strides made by its Advantage+ generative AI tools that allows advertisers to supply just an image, from which the AI will then develop an entire ad campaign. For small companies that can’t afford to hire an advertising agency, this could be a powerful alternative and a lucrative market for Meta. Its Advantage+ ad program has the potential to take share from creatives and even advertising agencies by customizing, altering, or outright creating content for advertising, rather than just targeting or measuring it.

ChatGPT, the chatbot that helped spark the market’s AI enthusiasm, is an important innovation because it can digest large amounts of text (hence the term large language model), communicate in natural (human) language, and generate sophisticated responses. Based on the Transformer architecture, a technique first introduced by Google in 2017, ChatGPT demonstrates the advances that have been made in AI that open the door to a wider set of business uses. But while natural language models recently produced epiphanies among lay chief executives and investors regarding AI, some tech companies were already investing in such capabilities and are being rewarded for that foresight. NVIDIA has been the biggest beneficiary this year in terms of its stock run and projected revenue gains; however, our other holdings, such as Adobe, Microsoft, Salesforce, ServiceNow, Synopsys, and TSMC, also appear among the possible beneficiaries. More companies—including, perhaps, some not yet in existence—will certainly join the ranks over time.

While it is still early, it’s evident that these companies see generative AI as transformative to their businesses and something upon which they can build new revenue models. Additionally, they are turning to AI to boost internal productivity, enhance existing customer offerings, and improve the quality and efficiency of customer interactions.

Most notably, Microsoft was able to gain an immediate leadership position in generative AI by making a US$10 billion investment in OpenAI, the company behind ChatGPT, earlier this year. Microsoft’s Bing search engine has since introduced ChatGPT into its web index data—a collection so large that it is rivaled by the dataset of only one other business in the world, Alphabet’s Google. Data are the feedstock of AI models, and an AI-enhanced search engine trained on so much data may attract more users to Bing, allowing Microsoft to sell more ads on the service. Microsoft is also adding generative AI to other products, including the Azure cloud service, enabling business customers who use Azure to easily integrate OpenAI models to glean more insights from their data and automate functions such as certain IT tasks. These added capabilities should motivate more businesses to migrate their data to the cloud and make Azure more competitive with Amazon.com’s AWS and Google Cloud.

The beneficiaries of demand for generative AI aren’t limited to traditional IT-sector companies. Data centers used to train AI models require up to ten times more power than typical data centers, thus requiring more-powerful equipment and backup power. Given the amount of heat they generate, new liquid-cooling solutions will be needed as well. This creates an opportunity for Schneider Electric, which has been developing innovative data-center equipment solutions for many years.

In the meantime, NVIDIA has emerged as the unrivaled global leader in providing the technologies at the center of the AI arms race. Due to an explosion of demand related to generative AI and LLMs from across its customer base, NVIDIA projects that data-center revenue for its fiscal second quarter ending in July will surge to US$11 billion. Not only is that more than double last quarter’s total, but the forecast also shattered the average analyst estimate that called for about US$7 billion.

Our investments in NVIDIA and Schneider are reflective of how we are thinking through the many unknowns and approaching portfolio structure in this environment. Through our fundamental framework, we can appreciate the broad excitement for AI, but we also remain conscious of valuations and thoughtful about diversification, recognizing that it’s unlikely anyone can predict today the biggest long-term winners.

In the first three months of 2023, P&G raised prices a whopping 10% year over year, following a series of large price hikes implemented throughout 2022:

By “not doing macro,” he means that we try not to allow our judgments about macroeconomic variables—GDP growth, inflation, and real interest rates—or geopolitical events to dictate our views on individual companies. By “macro does us,” he means that when the market’s risk tolerance and underlying assumptions change because of unexpected shifts in the macroeconomic environment, the consequential price movements can dominate a portfolio’s periodic absolute and relative returns. Although the injury may be only temporary, it is hard to avoid getting swept up in the general fervor. That’s a problem if it leads to reflexive and hasty reactions. It is precisely to avoid getting whipsawed in this way that we devote much of our efforts to restraining our inherent behavioral biases. But even with the sturdiest of behavioral guardrails designed to curb our responsiveness, the sudden jump in portfolio volatility and tracking error feels no less jarring.

Our investment approach centers on analysis of the prospects for specific companies and the industries in which they operate. As a result, the portfolios we construct are a mosaic of company-centric views, with the final picture coming into focus only after all of the pieces are assembled. Sometimes our bottom-up investment process leads us to sidestep systemic issues. In the years before the global financial crisis, for instance, we became disenchanted with the traditional banking industry. We didn’t like how the increased price transparency that came with the migration of services online diminished banks’ bargaining power over their borrowers and depositors, or how rising levels of consumer debt portended that growth could be weaker, and rivalry and risk-taking fiercer. That was enough to lead us largely to steer clear of banking stocks. Although in hindsight our portfolio positioning appeared to anticipate the subsequent dislocations, in fact we had no overarching view on systemic financial stability.

There is no question it would be nice to have clear foresight on GDP, inflation, and real rates. Like it or not, economic growth is the lifeblood of industrial economies, and, despite its ever-shifting relationship to equity returns, is closely associated with aggregate earnings. Similarly, inflation and real rates are both barometers and agents of economic transformation that always could and frequently do alter the path of economic growth. And there is strong reason to believe that macro-level dislocations are likely to be an order of magnitude greater than the mispricings that occur at the security level. Given the periodic importance of such dislocations, this raises the question: Why don’t we attempt to shape our portfolios more explicitly by directly forecasting economic variables or geopolitical events? The question is particularly vexing given the current importance of the inflation outlook for equities.

Tetlock’s conclusion was that expert predictions about geopolitical crises were no better than guesses. The only contribution that expertise seemed consistently to confer was a perverse boost in confidence regarding one’s (ineffective) forecasts.

The standard response typically trotted out is that forecasting is exceptionally hard, or as the Danish physicist Niels Bohr is alleged to have quipped, “Prediction is very difficult, especially about the future.” Nowhere is this more true than with geopolitical events, which by all accounts appear to defy anyone’s ability to anticipate them with anything approaching consistency. The political scientist Philip Tetlock tackled this issue head-on in a multidecade study described in his 2015 book, Superforecasting: The Art and Science of Prediction. Tetlock’s conclusion was that expert predictions about geopolitical crises were no better than guesses. What’s more, the only contribution that expertise seemed consistently to confer was a perverse boost in confidence regarding one’s (ineffective) forecasts.

The record for macroeconomic forecasting is not quite as wretched; at least there are frameworks and models on which to hang one’s thinking. But it’s still one of those endeavors where you’re doing very well if you’re right a little more often than you’re wrong. Even so, it is not as though the ground-level forecasting of cashflows, business prospects, and competitive forces is easy. So perhaps the real question is why we consider the latter sensible but the former a fool’s errand, at least for fundamental equity investors such as us.

The answer in large part comes down to the size of the opportunity set, or the number of times you get to apply your investing edge. Even the most skilled forecasters, whatever their forecasting game, have but the tiniest of edges and so the surest way to increase their chances of success is to apply that minute edge as many times as possible. In a global investment universe, there are roughly 8,000 equity securities, each operating in its own industry and geography with their own sets of return drivers, compared with a relative handful of forecastable macroeconomic variables. Given equal forecasting skill, you are going to have a far higher likelihood of some overall success by applying that skill across many securities rather than over a few economic statistics. Even allowing for the fact that not every security’s return is entirely idiosyncratic, there are still far more independent and durable drivers of individual security returns than there are of macroeconomic trends, which may allow you to get the micro right without so much as taking a swing at the macro.

Even if you were one of the few hyper-skilled and hyper-accurate macro forecasters, a portfolio of stocks would be a poor way to capitalize on views about inflation or economic growth. Although there’s a relationship between the macroeconomy and stock returns, that relationship is neither simple nor determinate. In practical terms, stocks are a terribly inefficient way to express a view on macroeconomic variables. Better to bet on currencies, yields curves, and commodity prices directly, all of which are far more closely tethered to the outlook for growth, inflation, and real rates.

It’s not just that there are better, more precise, and more levered ways to express macroeconomic views. It’s also that trying to do so with stocks risks erasing the hard-won company-level insights that are the linchpin of our portfolios.

And it’s not just that there are better, more precise, and more levered ways to express such views. It’s also that trying to do so with stocks risks erasing the hard-won company-level insights that are the linchpin of our portfolios. All the companies in which we invest have track records of successfully generating cash and reinvesting it wisely. In many cases these companies have survived wars, recessions, pandemics, inflation, deflation, and geopolitical shocks. Sacrificing those financially valuable fundamental attributes in a most likely vain attempt to time a particular economic cycle not only presupposes a preternatural ability to tie economic outcomes to individual security returns but also risks the long-term health of the portfolio.

We don’t do macro, so by default we allow macro to do us. There are, though, ways in which we can protect against developments that result in sudden changes in risk aversion. One is to diversify—events that damage the outlook in one industry or part of the world may have no impact, or even a beneficial one, on stock prices elsewhere. That said, diversification cannot work during times of systemic crisis, when correlations between geographies, industries, sectors, and individual securities converge. That’s where our reliance on a company’s strength comes in. Two hallmarks of a company’s quality are the ability of its management to prepare for a wide range of outcomes and whether it has the financial strength to survive the worst possible operating conditions.

Although we can’t estimate the probability of market-moving events, we can think about the magnitude and range of potential outcomes so we may more fully understand our exposures and ensure we are sufficiently diversified to protect against them. For example, before Russia’s invasion of Ukraine, many people thought about a range of outcomes that included war versus no war or disruption to energy supplies. But, given prior Western responses, few considered the potential for sanctions that would freeze all Russian assets and render them worthless, at least for the time being. Now, as we think about the financial market implications if China were to invade Taiwan, we must consider the possibility that Chinese assets could be similarly impaired.

So, what do we do about it? We certainly aren’t going to try to parse Chinese troop movements or overturn our investment theses on the dozens of companies, not only in China but also throughout the global supply chain, that could be impacted by what at this point must still be considered a very low-probability event. On the other hand, thinking long and hard about the potential risks to supply lines, revenues, or the corporate structures of portfolio companies and what further levels of diversification might be in order is very much in our wheelhouse.

As policymakers picked up where they had left off, they found themselves facing stakes heightened by the pandemic: stagnating incomes, weak consumer confidence, and a growing demographic crisis as birthrates continue to decline. These challenges may have accentuated their top priorities, ones that have been repeatedly highlighted in official policy statements over the last few years: innovation, rule of law, culture, the environment, and social harmony.

The fact is that ever since Deng Xiaoping initiated the initial series of capitalist overhauls in the 1980s, China has undergone multiple periods of reform. These changes cut a wide swath across economic activity and drastically curtailed certain targeted sectors. They were painful in their time, creating mass unemployment and fueling social discontent. Ultimately, they laid the groundwork and helped sustain several decades of nearly uninterrupted growth.

Previous reforms were far less visible to foreign observers because they barely touched the companies widely held by global investors at the time. For example, the coordinated supply-side reforms of 2015, undertaken in part to reduce chronic pollution, shuttered roughly one-fifth of China’s steel capacity (equivalent to Japan’s entire steel output) in under two years. Air quality improved dramatically, while bankruptcies almost tripled as many marginal producers were killed off. But not many foreign investors owned marginal steel producers, preferring to own faster growing companies such as Alibaba and New Oriental. Likewise, the anti-corruption campaigns that began in 2013 may have ushered in a more sustainable business environment, but they were terrible for liquor makers, whose products had become popular high-priced gifts to lubricate business deals and lobbying efforts. Kweichow Moutai, producer of its fiery namesake liquor, saw its sales growth plummet in 2014 and 2015, but the company was not nearly as widely owned externally as Tencent is today.

Much of the focus of late has been on one policy priority: common prosperity. Redolent of China’s collectivist past (the term was first used by Mao in 1950), the phrase frightens some foreign investors who are unsure which companies’ prosperity will be sacrificed at the altar of the commons. Yet policymakers have been clear: their focus is on growing middle-class disposable income, not “robbing the rich to help the poor,” according to Han Wenxiu, executive deputy director of the General Office of the Central Financial and Economic Affairs Commission. This overt aversion to a European-style welfare model may seem contradictory for a party that still pays lip service to its Marxist roots. But the reality is that China systematically underinvested in education, health care, and other social spending—especially in rural areas—as it sought to catch up economically with more developed economies. Until now, policymakers have done little in the way of redistribution; indirect taxes, which generally serve to widen income inequality, still represent two-thirds of fiscal revenue. With China coming into its own, we should expect its practices to converge with those in more advanced economies, including some form of income and wealth redistribution.

To my mind, these regulations are reminiscent of the US Progressive Era of the late 19th and early 20th centuries, epitomized by Theodore Roosevelt’s Square Deal.

In practice, the government’s targets for common prosperity—judging from recent policies and the detailed roadmap for its first pilot program in Zhejiang, the richest province in China and home to Alibaba—are education, health care, and housing. In these pivotal areas, structural impediments have exacerbated inequalities over time, producing a set of challenges that would be very familiar, for example, to residents of California. One of the more draconian national policy shifts, which recently consigned much of the private after-school tutoring business to the non-profit sector, does not go as far as South Korea’s complete ban of private tutoring in the 1980s.¹ In each country, the reforms were designed to ease the burden on parents who spend up to thousands of dollars each month coaching their children on how to pass exams. (To put this cost in perspective, the Chinese city with the highest average annual per capita disposable income in 2020 was Shanghai at $11,000.) Likewise, China’s recent online regulations covering antitrust, data security, and the safety of minors are similar to the concerns of consumer advocates everywhere.

To my mind, these regulations are reminiscent of the US Progressive Era of the late 19th and early 20th centuries, epitomized by Theodore Roosevelt’s Square Deal. It was not an easy time to invest and was marked by muscular antitrust interventions, the inception of a progressive income tax, and the appearance of the first federal consumer and environmental protections. Certain industries faced a permanently higher level of regulation with which they had been unfamiliar. But many companies thrived, and the reforms arguably laid the foundation for a century of growth that shaped the American economy into the largest in the world today, home to the largest number of globally competitive companies.

Structural changes of this magnitude will inevitably shake up competitive forces, buffeting the outlook for growth and strength of free cash flow generation for many businesses—but not all of them in negative ways. If China’s reforms succeed in improving middle-class disposable income while opening more opportunities for more people and still ensuring that the country remains a meritocracy, the government will have set the stage for more sustainable end demand for many industries. It’s a tall order, but one notable advantage enjoyed by Chinese policymakers today is the benefit of a century of hindsight observing which policies worked—and which did not—in the countries that have tried them.

The fact that filtering information is the long-term investor’s task, of course, doesn’t prevent most from feverishly pursuing every bit of information they can about the companies they follow. Egged on by the 24/7 news cycle, internet chat rooms, and by data providers to know their companies in as much depth as possible, investors often seek to become experts in every aspect of each company’s business. It’s dubious whether this helps them make better estimates of what a company’s shares are worth. Instead, the evidence suggests that, beyond a certain amount, each extra piece of information has zero marginal impact on forecast accuracy. The impact remains positive, however, on the investor’s confidence in their forecast.

The evidence suggests that, beyond a certain amount, each extra piece of information has zero marginal impact on forecast accuracy. The impact remains positive, however, on the investor’s confidence in their forecast.

In one unpublished study, described in Winning Decisions by J. Edward Russo and Paul J. H. Schoemaker, people who handicap horse races were asked to predict, based on increasing amounts of information, which horse would win a race. They found that, beyond a very limited amount of data, forecast accuracy did not improve with more data, but the handicappers’ confidence in their forecasts continued to grow with each extra bit of information.

Knowing more makes you feel better, in other words, but doesn’t help your decision-making. In investing, feelings and actual results arguably are even more disconnected. After all, feeling good about a decision is rarely a help in successful investing—to have a good outcome, an investment view must be different from that of most other investors, and standing apart from the herd almost always leads to discomfort.