AI Wealth Truth (65): Why AI Chips Are Worth More Than AI Algorithms

I. In 2023, Nvidia's stock price tripled. A chip company became one of the most valuable companies in the world. Meanwhile, many AI application companies are still losing money. Why do the "shovel sellers" make more money than the gold miners?

II. This phenomenon has an economic name: infrastructure arbitrage.

III. The gold rush in history is a typical example. In 1848, gold was discovered in California. Thousands rushed to mine gold. Who made money in the end? Only a few lucky miners got rich. But those who sold shovels, jeans, and tents almost all made money.

IV. Levi Strauss started by selling jeans back then. He did not take the risk of mining gold, yet benefited from the gold rush. This is the "selling shovels" strategy.

V. What is the gold rush of the AI era?

VI. Countless companies are building AI applications. AI chat, AI art, AI writing, AI coding. Everyone wants to create the next killer app. But the success rate is low. Most AI startups will fail.

VII. What do all these companies need? Compute. GPU chips. Whether your AI app succeeds or fails, you need chips to train and run models. Chips are the shovels of the AI era.

VIII. Where does Nvidia sit?

IX. A monopoly position. Nvidia controls more than 80% of the AI training chip market. There are no real substitutes. If you want to train AI, you have to buy Nvidia chips. This is an absolute monopoly.

X. Demand exceeds supply. AI demand has exploded, but chip capacity cannot keep up. Customers have to wait in line. This gives Nvidia extremely strong pricing power. Scarcity pushes up prices and profits.

XI. Ecosystem lock-in. Nvidia's CUDA software ecosystem is the industry standard. Developers all use CUDA. Even if there are alternative chips, migration costs are high. Software ecosystems lock in hardware sales.

XII. What does this mean?

XIII. The application layer is high risk. Which AI product will succeed? No one knows. Investing in the application layer is high risk. Upside is high, but failure probability is high too. This is gold mining.

XIV. The infrastructure layer is lower risk. No matter which application succeeds, chips are needed. Investing in chips is investing in the entire AI industry. You do not have to pick the winner. This is selling shovels.

XV. This logic generalizes:

XVI. Cloud computing is also a shovel for AI. AI apps need to run in the cloud. AWS, Azure, and Google Cloud are all making money from AI. No matter which AI app succeeds, cloud platforms collect rent.

XVII. Energy is also a shovel for AI. Data centers need massive electricity. The AI boom pushes up energy demand. Power companies benefit too.

XVIII. Network equipment is also a shovel for AI. Data transmission needs high-speed networks. Switches, routers, fiber optics. The infrastructure layer benefits in every dimension.

XIX. What can ordinary people learn from this?

XX. 1. Identify the "shovels." In any new trend, ask: what will be needed regardless of which specific apps win? Those are the shovels. Investing in shovels is more robust than investing in miners.

XXI. 2. Understand the power of monopoly. Nvidia's excess profits come from monopoly. Monopoly means pricing power, and excess profits. Look for infrastructure companies with monopoly positions.

XXII. 3. Accept lower but more stable returns. Buying shovels will not make you rich overnight. But it will not bankrupt you either. For most people, this is a better risk-return profile. Stability is more sustainable than getting rich fast.

XXIII. Why is Nvidia worth more than OpenAI? (At one point Nvidia's market cap was more than 5x OpenAI's valuation.) Because OpenAI is still burning cash, while Nvidia is printing money. Shovel-selling profits are more certain than gold-mining profits.

XXIV. In the AI era, the biggest winners may not be AI application companies. They may be those selling GPUs, cloud services, and power. Whoever controls infrastructure controls the value chain. This is the rule of every technological revolution. In the railroad era, sellers of steel and coal made more money than railroad companies. In the internet era, sellers of network equipment and real estate (data centers) had sure wins. The AI era is no different. Only by seeing this rule clearly can you make rational choices in the AI boom.