36th Annual Graham & Dodd Event with Li Lu '96, Founder and Chairman of Himalaya Capital

Summary

• Li Lu defines wealth as ownership of global purchasing power, accumulated through businesses that create value over time. His framework combines Graham’s business ownership, market independence, and margin of safety with Buffett’s circle of competence and Munger’s advice to “fish where the fish are.” Himalaya’s implementation is to buy great businesses in dynamic economies at advantageous prices and hold long enough to participate in their value creation.
• AI’s usefulness does not remove its recurring costs, capital requirements, or need for human judgment. Li calls the technology “transformative, dangerous, disruptive, and costly” and rejects investment narratives culminating in an “IPO of a God.” He advocates keeping humans in the loop, particularly during recursive self-improvement. Santos suggests that skilled people checking AI output will capture scarce economic rents; Li broadly agrees, while emphasizing nuances.
• Market concentration can create opportunities among businesses overlooked by investors chasing the leaders. Santos puts Nvidia’s capitalization at about 15% of GDP, versus a historical 2–4% range for the largest company. Li contrasts roughly 14–15% annual S&P returns over 16 years since 2010 with approximately zero during 2000–2010: combined, the result returns to high single digits. His conclusion is that widening speculative enthusiasm can improve the opportunity set for investors who remain rational.
• China’s industrial upgrade has coincided with a severe contraction in household wealth and confidence. Li says typical middle-class families had 70% of net worth tied to homes carrying roughly 50% mortgage leverage; a subsequent nationwide property-price decline of about 40% reduced family wealth by 60–70%. Households consequently consume less, save more, and repay debt. He puts saving deposits at RMB165 trillion, about 120% of GDP, while domestic-demand growth has slowed to roughly zero.
• Li’s central question for investing in China is whether a business is a state objective or the cost of achieving one. “Being the cost of a state objective is no fun.” He argues semiconductor restrictions supplied domestic producers with the orders needed to fund improvement, while manufacturing advanced toward the higher end. He expects equities eventually to assume a greater role in household wealth as property’s share declines, but leaves the timing open.
• Li expects commercial interdependence to outlast the geopolitical order currently being reshaped. Leaving the common market sacrifices efficiency, specialization, and access to trading partners, while barriers redirect commerce through third countries. Semiconductor supply chains illustrate the dependence underpinning what he calls “strategic stability.” Countries can compete intensely while sharing an interest in preserving that stability.
• A value strategy requires patient partners and a temperament capable of surviving underperformance. Li warns that demanding good results every year, quarter, or month can turn an investor into a speculator. Downside protection starts with understanding the boundaries of one’s knowledge and expanding them gradually. His rule for compounding—“you don’t lose what you already have”—extends beyond capital to knowledge, relationships, wisdom, and health.

Deep dive

1. Business ownership and purchasing power anchor Li’s framework

• Li’s first three principles come from Benjamin Graham: a stock represents business ownership; the market exists to serve the investor; and an inherently unpredictable future requires a margin of safety. Buying shares is therefore equivalent to buying a company, while market quotations provide opportunities rather than instructions.

• Buffett’s addition is the circle of competence, especially “knowing the edge of it.” Munger’s is to “fish where the fish are”: crowded markets, industries, and businesses attract competition that averages away returns. Neglected opportunities can offer better fishing grounds, provided the investor understands what is inside that circle.

• Himalaya’s sixth principle defines wealth as “the percentage ownership of global purchasing power.” Li implements it by buying great businesses in dynamic economies at advantageous prices and holding long enough to share in their value creation. Santos connects the definition to a concern about preserving purchasing power if fiscal challenges prevent inflation from being controlled.

• Li separates investors, pricing arbitrageurs, speculators, and brokers or market makers. The market’s purpose is to channel savings into productive businesses; speculation and intermediation impose costs on that function. Buffett’s “church with a casino attached to it” captures the tension. Li describes Berkshire as a “gigantic cathedral” and Himalaya as an effort to demonstrate that similar churches can be built elsewhere.

2. AI needs human control and continuing financial discipline

• Asked whether AI displaces equity analysts, Li first challenges its investment narrative. He traces the claimed progression through reasoning, coding, autonomous agents, recursive self-improvement, AGI, and ASI, culminating in a man-made god. Investors are almost invited into the “IPO of a God.” His categorical objection: either technology does not reach that destination, or humanity unites to prevent it.

• His provisional definition is a machine using brute compute and clever algorithms to establish contextual correlations across human-generated data and its derivatives. It synthesizes knowledge and expertise, but he says it does not itself change the physical transformations underlying energy, transport, or manufacturing. “We know how it works and how to make it better, but we don’t quite understand why it works.”

• Li’s red line is recursive self-improvement without a human in the loop: independently generated goals would move the system further from human control. Employment, emotional dependency, and cyber, biological, and nuclear risks also enter the calculation. AI will always have incremental costs. Capital needs, profitability, and societal tolerance should therefore shape development; he expects boom-and-bust cycles, potentially several.

• Santos’s proposed implication for analysts is that the scarce resource becomes the skilled person checking AI output. Someone who understands a domain and communicates effectively independently of AI may capture the rents. Li broadly agrees—“somewhere along the line”—while explicitly leaving room for “different aspects and nuances.”

3. Concentration can improve the opportunity set elsewhere

• Santos’s challenge is concentration: the largest company’s capitalization historically ran at 2–4% of GDP in his chart, versus about 15% for Nvidia now. Li answers with an unnamed research lab burning tens of billions, promising to burn hundreds of billions more, and citing a 10% chance of killing everyone, yet seeking at least $100 billion at a $2 trillion valuation. He treats that combination as irrational and unsustainable.

• Li’s grounding comparison is the S&P: roughly 14–15% annual returns over 16 years since 2010, against approximately zero or slightly negative returns during 2000–2010. Across 26 years, that becomes high single digits. He connects the outcome to around 2% real GDP growth, 2–3% inflation, and listed companies doing somewhat better than the broader economy. Long-run returns average out to high single digits.

• When capital crowds into the leaders, otherwise attractive businesses can trade cheaply. Li sees that bifurcation in both American and Chinese markets. A technological wave expands the “margin of imagination,” intensifying speculation and exaggeration. Remaining rational is difficult, he acknowledges, but the resulting dispersion can offer better long-term opportunities. Investors need not let the market dictate their decisions.

4. China’s industrial upgrade carries a heavy household cost

• Li reads the last decade as a deliberate attempt to escape the middle-income trap. Made in China 2025 targeted a shift from low- and middle-end manufacturing toward the middle and high end. Trump-era trade restrictions and subsequent Biden-era technology restrictions reinforced that direction. By 2021, Beijing doubled down, adding technological self-sufficiency in bottlenecks including semiconductors, aircraft engines, and operating-system code.

• Property is the cost side of that choice in Li’s account. At the 2021 peak, a typical middle-class family held 70% of net worth in its home, carrying roughly 50% mortgage leverage. He puts the subsequent nationwide price decline at about 40% and the associated fall in family wealth at 60–70%. The response: consume less, save more, and repay debt.

• Li says domestic-demand growth has consequently slowed toward zero, while saving deposits have risen to about RMB165 trillion, or 120% of GDP. Two to three years of roughly flat or slightly declining prices have further weakened confidence. The household experience therefore looks very different from the industrial progress he describes.

• Annual new sales in real estate, he says, are down about 70%; the roughly RMB10 trillion figure is also the amount spent annually on manufacturing upgrades and improving technology efficiency, with the divergence around 2021. His development model is sequential: an industry promoted as an objective in one decade can become a cost in the next. Property once delivered urbanization and middle-class wealth, then absorbed the adjustment.

5. Policy objectives shape China’s industrial opportunities

• Li says every Made in China 2025 objective has been hit, citing progress in EVs, batteries, storage, new energy, and heavy machinery. He calls US-led semiconductor sanctions a “godsend”: domestic orders generate revenue, profit, R&D, better products, and more orders. He describes mature-node equipment and fabrication as almost self-sufficient and advanced nodes as getting close, with a few missing ingredients; continued progress, he argues, could bring self-sufficiency.

• AI is another equalizer in his account: code generation helps address missing software layers, including those serving PCs, servers, and Nvidia’s CUDA system. Asked to distinguish American and Chinese AI, he contrasts Silicon Valley’s question—“what else can this technology do?”—with Chinese companies’ greater emphasis on consumer and enterprise customers. He explicitly says Chinese firms also share the frontier curiosity; the distinction is emphasis.

• An audience member asks whether the stock market itself is an objective or a cost. Li leaves the date open but argues equities must eventually take a greater role as property shrinks within family and national balance sheets: they grow with—and help drive—the economy. On rules changing overnight, his answer is to study state objectives, history, and the record. He applies that requirement to the United States and other countries too.

6. Commercial interdependence supports strategic stability

• Santos contrasts the openness surrounding Himalaya’s 1997 founding with today’s less generous international atmosphere. Li distinguishes a changing political order from a more durable economic order. The common global market consolidated after the Soviet Union’s collapse; withdrawal sacrifices efficiency, technological improvement, specialization, and access to resources. He sees Washington reconsidering the costs of global stewardship, without successfully withdrawing from those responsibilities so far.

• Trade barriers can redirect commerce through intermediaries. Li points to American trade with Mexico plus ASEAN as offsetting declining direct US–China trade. His semiconductor illustration is that producing countries and leading executives “all use chopsticks,” emphasizing the dependence of the AI buildout on those suppliers. His instruction is correspondingly blunt: countries betting heavily on that technology must maintain workable relationships with them.

• Li says the US and Chinese presidents find it urgent to meet four times in eight months, and remarks that a meeting’s main accomplishment can be scheduling the next one. The transcript does not make clear how many of those meetings have already happened. “Stability is of strategic importance”: China pursues semiconductor self-sufficiency, America hopes for AI breakthroughs, and each side believes time favors it. Their competition coexists with an interest in preserving the common market and avoiding disruption.

7. Li applies value principles broadly and dismisses crypto

• Asked whether value investing works for growth and technology businesses, Li answers that “all intelligent investment is value investment.” A declining business returning cash, a rapidly growing company, and a portfolio of low-probability bets all require understanding what is received relative to price. The common mathematics is discounting future cash flows, insofar as they can be understood and estimated accurately; application depends on competence.

• On energy and climate, Li identifies energy and transportation as foundational layers of the economy. Current AI constraints run “from chips to electricity to labor.” He acknowledges climate implications but does not develop a specific directional investment argument from them.

• His crypto judgment is categorical: “I think eventually it will disappear on its own.” He calls it “fundamentally bogus,” arguing that currency requires an enforcement mechanism for settlement and a store of wealth. He associates crypto with illicit transactions and favors banning it. He nevertheless treats collecting it, like paying a particular price for art, as a personal preference; that does not establish societal value in his account.

8. Patient capital allows knowledge and trust to compound

• An investor two years into practicing value investing asks about career risk during underperformance. Li warns against wanting both long-term value creation and favorable results every year, quarter, month, day, or hour: “Pretty soon you become a speculator.” The choice is whether the market serves or instructs you. For investors unable to maintain that distinction, he recommends indexing.

• Building a business around the approach requires like-minded capital that will stay. The necessary temperament accepts being right because logic and facts support the conclusion, independently of others’ agreement. Li says some people fit that psychology and others should not force it. That also limits the ambition to gather the biggest fund or maximize assets under management.

• Downside protection starts with the boundary of what can be known. In his twenties, Li began with easier-to-understand balance sheets, gradually moving into more complex businesses as competence grew. “If you can compound your knowledge in a sustainable fashion, eventually you can compound your capital as well.” Each decade’s learning, he says, can leave him almost a different person.

• Li credits “so many lucky breaks” rather than treating the record as proof of genius. Munger’s “web of deserved trust” anchors his advice on partners and relationships. The compounding rule is “you don’t lose what you already have,” extended to knowledge, trust, wisdom, and health. Continued engagement may prolong productive and health spans, he suggests, even if it does not prolong lifespan.