AI Wealth Truth (66): Why OpenAI's $7 Trillion Chip Plan Is a Power Game

I. In early 2024, reports said OpenAI CEO Sam Altman was planning a $7 trillion chip infrastructure project. The number is absurdly large, comparable to the combined GDP of many countries. Why would an AI company spend so much money on chips?

II. The answer is: compute is a strategic resource in the AI era. Whoever controls compute controls the future of AI. This is not business competition. This is a power game.

III. Let us understand the strategic meaning of compute:

IV. AI capability depends on compute. Larger models need more compute to train. One reason GPT-4 is stronger than GPT-3 is that it used more compute. Future AI may need exponentially more compute. Compute is the physical foundation of AI capability.

V. Today, compute is controlled by a few companies. Nvidia controls chips. Microsoft, Google, and Amazon control cloud computing. OpenAI depends on these companies. Dependence means being constrained.

VI. Controlling compute means controlling the AI value chain. If you control chip supply, all AI companies have to buy from you. You can decide who gets compute and who does not. That is enormous power.

VII. Why is this like an "oil war"?

VIII. In the 20th century, oil was the most important strategic resource. Countries and companies that controlled oil had huge geopolitical influence. Oil wars, embargoes, and price manipulation were tools of power. In the 21st century, compute may play the same role.

IX. Concentrated compute means concentrated AI capability. Only a few entities can train the strongest AI. Everyone else can only use the AI services those entities provide. Users become vassals.

X. Compute control means data control. Training AI needs data. Data is often generated while using AI. Those who control compute can obtain more data. More data trains better AI. A positive feedback loop reinforces control.

XI. Who is fighting for control of compute?

XII. Big tech. Microsoft, Google, Meta, and Amazon are investing massively in data centers and chips. They do not want to depend on Nvidia. They are developing their own AI chips. Vertical integration to control the whole supply chain.

XIII. Sovereign states. The US restricts exports of advanced chips to China. China is investing heavily in domestic chips. Chips become a geopolitical weapon. Compute competition is an extension of national competition.

XIV. New forces. OpenAI is trying to build its own chip supply chain. Other AI companies are considering similar strategies. They do not want to be held hostage.

XV. What does a $7 trillion plan mean?

XVI. Literally, it is unlikely to happen. The number is too big. Even spread over 10 years it is $700 billion a year. Global semiconductor industry revenue is only a little over $500 billion a year. This looks more like a vision or a negotiation strategy.

XVII. But it reflects real anxiety. OpenAI worries that compute supply will be controlled. It wants its own infrastructure. This is resistance to a dependent position.

XVIII. It points to the direction of future competition. AI competition will increasingly become infrastructure competition. Chips, data centers, energy, networks. Whoever controls these will control the AI era.

XIX. What does this mean for ordinary people?

XX. You are a vassal in this power game. The AI services you use depend on who wins the compute war. Your data flows to the winners. Your choices are limited to the options the winners provide. You do not get to participate; you only get to be affected.

XXI. Investment opportunities may be in infrastructure. Same logic as the previous chapter. Chips, energy, data centers. No matter who wins the application layer, the infrastructure layer benefits.

XXII. Political influence will shape the future of AI. Regulation, trade policy, antitrust policy. These will decide how compute is allocated. Politics is not external to AI; it is a core variable of AI development.

XXIII. Compute is not a neutral technical resource. It is power. Controlling compute means controlling who can develop AI and who cannot. Controlling what content can be generated and what cannot. Controlling what kind of future will arrive and what will not. This is a war in progress. You and I are spectators and pawns.

XXIV. The $7 trillion plan may not happen. But the anxiety and ambition it represents are real. The power structure of the AI era is being shaped. Once the balance of power stabilizes, change becomes hard. We are in this critical period. The eventual winners will set the rules of the AI era. And those rules will affect everyone's life.